CN118010103B - Intelligent monitoring method and system for equal-thickness cement soil stirring wall in severe cold environment - Google Patents
Intelligent monitoring method and system for equal-thickness cement soil stirring wall in severe cold environment Download PDFInfo
- Publication number
- CN118010103B CN118010103B CN202410425088.5A CN202410425088A CN118010103B CN 118010103 B CN118010103 B CN 118010103B CN 202410425088 A CN202410425088 A CN 202410425088A CN 118010103 B CN118010103 B CN 118010103B
- Authority
- CN
- China
- Prior art keywords
- mutual information
- features
- learning machine
- machine model
- extreme learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 102
- 238000000034 method Methods 0.000 title claims abstract description 53
- 239000004568 cement Substances 0.000 title claims abstract description 22
- 238000003756 stirring Methods 0.000 title claims description 69
- 239000002689 soil Substances 0.000 title claims description 24
- 230000006870 function Effects 0.000 claims abstract description 118
- 238000012545 processing Methods 0.000 claims abstract description 38
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 29
- 238000002156 mixing Methods 0.000 claims abstract description 28
- 238000005457 optimization Methods 0.000 claims abstract description 28
- 230000005540 biological transmission Effects 0.000 claims abstract description 15
- 230000009467 reduction Effects 0.000 claims abstract description 15
- 230000008859 change Effects 0.000 claims abstract description 13
- 238000006073 displacement reaction Methods 0.000 claims description 33
- 239000011159 matrix material Substances 0.000 claims description 29
- 239000002131 composite material Substances 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 26
- 238000001914 filtration Methods 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000002790 cross-validation Methods 0.000 claims description 10
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000010355 oscillation Effects 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 4
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 description 6
- 238000005259 measurement Methods 0.000 description 6
- 238000010276 construction Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004140 cleaning Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- QVFWZNCVPCJQOP-UHFFFAOYSA-N chloralodol Chemical compound CC(O)(C)CC(C)OC(O)C(Cl)(Cl)Cl QVFWZNCVPCJQOP-UHFFFAOYSA-N 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000013450 outlier detection Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention provides an equal-thickness cement mixing wall intelligent monitoring method and system in a severe cold environment, which relate to the technical field of cement mixing wall monitoring and comprise the steps of setting one or more sensors, respectively acquiring data acquired by the one or more sensors based on a wireless transmission mode, processing the acquired data to obtain monitoring data, carrying out self-adaptive decomposition on the monitoring data, extracting characteristics of different time scales to obtain a plurality of candidate characteristics, carrying out reduction and optimization processing on the plurality of candidate characteristics by utilizing a combined mutual information minimization method based on combined mutual information to construct a target characteristic subset, constructing a target multi-core extreme learning machine model based on different types of kernel functions, and processing the target characteristic subset by utilizing the target multi-core extreme learning machine model to predict the performance change trend of a cement mixing wall so as to obtain a prediction result.
Description
Technical Field
The invention relates to a cement mixing wall monitoring technology, in particular to an intelligent monitoring method and system for a cement mixing wall with equal thickness in a severe cold environment.
Background
The civil engineering construction is carried out in the alpine region, and especially when the foundation is reinforced by using the cement soil stirring wall with equal thickness, the double challenges of extreme climate and complex geological conditions are faced. Traditional cement-soil mixing wall monitoring methods rely on physical sampling and manual observation, are long in time consumption and high in cost, are difficult to implement in severe high-cold environments, and cannot meet the requirement of real-time monitoring. In addition, the conventional monitoring technology has significant limitations in processing and analyzing large-scale monitoring data, and is difficult to effectively extract key information from complex monitoring data, thereby affecting the accuracy and practicality of monitoring results.
Disclosure of Invention
The embodiment of the invention provides an intelligent monitoring method and system for an equal-thickness cement soil stirring wall in a severe cold environment, which can solve the problems in the prior art.
In a first aspect of an embodiment of the present invention,
The intelligent monitoring method for the equal-thickness cement soil stirring wall in the alpine environment comprises the following steps:
Setting one or more sensors, respectively acquiring data acquired by the one or more sensors based on a wireless transmission mode, and processing the acquired data to obtain monitoring data, wherein the sensors comprise one or more of a temperature sensor, a humidity sensor, a strain sensor, a displacement sensor and an inclination sensor, and the acquired data comprise one or more of temperatures of different depths of a stirring wall, temperature differences inside and outside the stirring wall, humidity of different depths of the stirring wall, a strain value, a displacement value and an inclination value;
Performing self-adaptive decomposition on the monitoring data, extracting features with different time frequency scales, obtaining a plurality of candidate features, performing dimension reduction and optimization processing on the plurality of candidate features by utilizing a joint mutual information minimization method based on joint mutual information so as to construct a target feature subset, wherein the joint mutual information represents the dependence and the association degree between the features;
And constructing a target multi-core extreme learning machine model based on different types of kernel functions, and processing the target feature subset by using the target multi-core extreme learning machine model to predict the performance change trend of the cement-soil mixing wall so as to obtain a prediction result, wherein the performance of the cement-soil mixing wall comprises one or more of strength, stability, durability and structure.
In an alternative embodiment, one or more sensors are provided, data collected by the one or more sensors are respectively obtained based on a wireless transmission mode, and the collected data are processed to obtain monitoring data, which includes:
temperature sensors are respectively arranged at the top, the middle and the bottom of the stirring wall in a layered arrangement mode, so that temperatures of different depths of the stirring wall are collected to serve as monitoring data, and the temperature sensors are symmetrically arranged at the inner side and the outer side of the stirring wall, so that temperature differences between the inner side and the outer side of the stirring wall are collected to serve as monitoring data;
a plurality of humidity sensors are respectively arranged at the top, the middle and the bottom in the stirring wall in a layered arrangement mode so as to collect humidity of different depths of the stirring wall, and weighted average is carried out on data collected by the plurality of humidity sensors to be used as monitoring data;
Strain sensors are respectively arranged at the middle part and the bottom of the stirring wall in an embedded arrangement mode, strain values in the vertical direction and the horizontal direction of the stirring wall are collected to be used as monitoring data, 4 strain sensors are respectively arranged at the center and the edge of a monitoring section to form a strain monitoring array, and the strain values are collected to be used as monitoring data;
And filtering the displacement value acquired by the displacement sensor and the inclination value acquired by the inclination sensor respectively by adopting a self-adaptive Kalman filtering algorithm, wherein filtering parameters are dynamically adjusted to eliminate noise and abnormal interference, and the filtered displacement value and the filtered inclination value are used as monitoring data.
In an alternative embodiment, the adaptive decomposition includes a variational modal decomposition VMD, adaptively decomposing the monitored data and extracting features of different time-frequency scales to obtain a plurality of candidate features, including:
carrying out optimization solution on the monitoring data by utilizing a variational modal decomposition VMD model to obtain K intrinsic mode function IMF components, wherein each intrinsic mode function IMF component represents a local oscillation mode of the corresponding monitoring data, the variational modal decomposition VMD model comprises a first objective function and constraint conditions, and the formula of the first objective function of the variational modal decomposition VMD model is as follows:
；
Wherein u k is the kth intrinsic mode function IMF component, w k is the center frequency corresponding to the kth intrinsic mode function IMF component, K is the identity of the intrinsic mode function IMF component, K is the total number of the intrinsic mode function IMF components, t is time, u k (t) is the kth intrinsic mode function IMF component at time t, delta (·) is the dirac function, Representing a first derivative of time t, j being a preset parameter;
The formula of constraint conditions of the variational modal decomposition VMD model is as follows:
；
Wherein z (t) is the monitored data at time t;
Reconstructing the intrinsic mode function IMF components into m-dimensional space vectors for each intrinsic mode function IMF component, calculating the probability of pattern matching based on the distance between the space vectors and a preset similarity threshold, and calculating the sample entropy of each intrinsic mode function IMF component based on the probability ratio of the pattern matching to obtain a plurality of candidate features, wherein the formula is as follows:
；
Wherein SampEn k represents the sample entropy corresponding to the kth intrinsic mode function IMF component, m represents the spatial dimension, r represents the preset similarity threshold, and B m (r) represents the probability of pattern matching in the spatial dimension m and the preset similarity threshold r.
In an alternative embodiment, performing a dimension reduction and optimization process based on the joint mutual information using a joint mutual information minimization method for a plurality of candidate features to construct a target feature subset, includes:
Determining a selected feature subset from a plurality of candidate features;
calculating mutual information quantity between the plurality of candidate features and the tag feature based on joint probability distribution of the plurality of candidate features and the tag feature and edge probability distribution of the plurality of candidate features and the tag feature;
Calculating the conditional mutual information quantity between the selected feature subset and the tag feature based on the joint probability distribution of the selected feature subset and the tag feature and the edge probability distribution of the selected feature subset and the tag feature under the preset condition;
Taking the difference value between the mutual information quantity between the plurality of candidate features and the tag features and the conditional mutual information quantity between the selected feature subset and the tag features as the joint mutual information quantity among the selected feature subset, the plurality of candidate features and the tag features, and carrying out optimization solving with the aim of minimizing the joint mutual information quantity so as to carry out dimension reduction and optimization processing on the plurality of candidate features;
And iteratively executing the steps of determining the selected feature subset from the plurality of candidate features and the following steps until the number of the target feature subset reaches a preset feature number threshold or meets a preset performance requirement, so as to obtain the target feature subset.
In an alternative embodiment, the formula for calculating the amount of mutual information between the plurality of candidate features and the tag feature is as follows:
；
Wherein I (X k; C) represents the mutual information quantity between a plurality of candidate features and the tag features, X k represents the candidate features, C represents the tag features, X k is the value of a single candidate feature, C i is the value of a single tag feature, and p () represents a probability distribution function;
The formula for calculating the conditional mutual information quantity between the selected feature subset and the tag features is as follows:
；
Wherein I (X j; c|Y) represents the amount of conditional mutual information between the selected feature subset and the tag features, X j represents the selected features in the selected feature subset, Y is a given condition, X j represents the value of a single selected feature in the selected feature subset, and Y represents the value of the given condition;
the formula for calculating the amount of joint mutual information between the selected feature subset, the plurality of candidate features, and the tag feature is as follows:
；
Wherein I (X k;Xj; C) represents the amount of joint mutual information between the selected feature subset, the plurality of candidate features, and the tag feature;
the formula for optimization solution with the aim of minimizing the amount of the joint mutual information is as follows:
；
Wherein, Representing an optimization objective that minimizes the amount of joint mutual information, F represents a set of multiple candidate features, and S represents a selected subset of features.
In an alternative embodiment, constructing the target multi-core extreme learning machine model based on different types of kernel functions includes:
And combining based on a Gaussian kernel function and a linear kernel function to obtain a composite kernel function, wherein the Gaussian kernel function has the following formula:
；
Wherein K RBF (·) represents a Gaussian kernel function, n represents a sample of a target feature subset, n' represents a sample of a non-target feature subset, σ is a width parameter of the Gaussian kernel;
the formula of the linear kernel function is as follows:
；
Wherein K linear (·) represents a linear kernel function;
taking the composite kernel function as the basis of an initial multi-core extreme learning machine model, and randomly initializing parameters from an input layer to a hidden layer of the initial multi-core extreme learning machine model to construct the initial multi-core extreme learning machine model, wherein the parameters from the input layer to the hidden layer comprise weights and bias values;
Training an initial multi-core extreme learning machine model by using a first training sample, optimizing parameters of a composite kernel function and the node number of a hidden layer based on a second objective function, replacing the composite kernel function and the node number of the hidden layer in the initial multi-core extreme learning machine model by the optimized composite kernel function and the node number of the optimized hidden layer, and determining an intermediate multi-core extreme learning machine model;
And training the intermediate multi-core extreme learning machine model by using the second training sample, and optimizing parameters from an input layer to a hidden layer of the intermediate multi-core extreme learning machine model based on a third objective function to obtain the target multi-core extreme learning machine model.
In an alternative embodiment, the second objective function is formulated as follows:
；
Wherein L (·) represents a loss function, H represents the number of nodes of the hidden layer, Representing the true value in the d-th cross validation,/>Representing the predicted value output by the initial multi-core extreme learning machine model in the D-th cross validation, D represents the total times of the cross validation,
The formula of the third objective function is as follows:
；
Wherein HE is the hidden layer output matrix, β is the parameters of the input layer to the hidden layer, N is the second training sample, λ is the preset parameters;
The hidden layer output matrix is determined by calculating the similarity degree between each second training sample and the hidden layer node by using a composite kernel function, and the formula is as follows:
；
wherein HE represents a hidden layer output matrix, and alpha is a preset parameter;
the parameters from the input layer to the hidden layer are calculated by using a least square method based on the hidden layer output matrix, and the formula is as follows:
；
Wherein, For the mole-Penrose pseudo-inverse of the hidden layer output matrix HE, Y N is a sample prediction result of a second training sample output by the middle multi-core extreme learning machine model;
processing the target feature subset by using the target multi-core extreme learning machine model to predict the performance change trend of the cement-soil mixing wall to obtain a prediction result, wherein the method comprises the following steps of:
and processing the target feature subset by using a composite kernel function of the target multi-core extreme learning machine model to obtain a hidden layer output matrix of the target feature subset, and processing the hidden layer output matrix of the target feature subset by using parameters from an input layer to a hidden layer of the target multi-core extreme learning machine model to obtain a prediction result.
In a second aspect of an embodiment of the present invention,
The utility model provides an equal thickness cement soil stirring wall intelligent monitoring system under severe cold environment, include:
The first unit is used for setting one or more sensors, respectively acquiring data acquired by the one or more sensors based on a wireless transmission mode, and processing the acquired data to obtain monitoring data, wherein the sensors comprise one or more of a temperature sensor, a humidity sensor, a strain sensor, a displacement sensor and an inclination sensor, and the acquired data comprise one or more of temperatures of different depths of a stirring wall, temperature differences inside and outside the stirring wall, humidity of different depths of the stirring wall, a strain value, a displacement value and an inclination value;
The second unit is used for carrying out self-adaptive decomposition on the monitoring data and extracting features with different time frequency scales to obtain a plurality of candidate features, carrying out dimension reduction and optimization processing on the plurality of candidate features by utilizing a joint mutual information minimization method based on joint mutual information so as to construct a target feature subset, wherein the joint mutual information represents the dependence and the association degree between the features;
And the third unit is used for constructing a target multi-core extreme learning machine model based on different types of kernel functions, processing the target feature subset by using the target multi-core extreme learning machine model so as to predict the performance change trend of the cement mixing wall and obtain a prediction result, wherein the performance of the cement mixing wall comprises one or more of strength, stability, durability and structure.
In a third aspect of an embodiment of the present invention,
There is provided an electronic device including:
A processor;
a memory for storing processor-executable instructions;
Wherein the processor is configured to invoke the instructions stored in the memory to perform the method described previously.
In a fourth aspect of an embodiment of the present invention,
There is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method as described above.
In the embodiment of the invention, by arranging a plurality of sensors such as temperature, humidity, strain, displacement, inclination and the like, the monitoring data of the stirring wall in different depths and different aspects can be comprehensively collected, and sufficient data support is provided for subsequent performance prediction; by utilizing a self-adaptive decomposition and combined mutual information minimization method, key features are extracted from mass monitoring data, redundancy and noise are removed, an optimal target feature subset is obtained, and the accuracy and the efficiency of subsequent prediction are improved; the multi-core extreme learning machine model is adopted to model and predict target characteristics, and a plurality of different kernel functions are constructed to better fit complex nonlinear relations, so that the generalization capability and robustness of prediction are improved, the current performance state of the stirring wall can be predicted, the change trend of performance in a future period of time can be predicted, and the multiple aspects of strength, stability, durability and the like are provided for quality control and safety early warning; the monitoring data are collected in a wireless transmission mode, and an extreme learning machine model with higher calculation efficiency is adopted, so that real-time online prediction of the performance of the stirring wall can be realized, abnormal change trend can be found in time, the method can be applied to performance monitoring and prediction of the cement soil stirring wall under different engineering environments, and the method has good applicability and popularization.
Drawings
FIG. 1 is a flow chart of an intelligent monitoring method for an equal-thickness cement-soil mixing wall in a severe cold environment according to the embodiment of the invention;
fig. 2 is a schematic structural diagram of an intelligent monitoring system for an equal-thickness cement-soil mixing wall in a severe cold environment according to an embodiment of the invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The technical scheme of the invention is described in detail below by specific examples. The following embodiments may be combined with each other, and some embodiments may not be repeated for the same or similar concepts or processes.
Fig. 1 is a flow chart of an intelligent monitoring method for an equal-thickness cement-soil mixing wall in a severe cold environment according to an embodiment of the invention, as shown in fig. 1, the method includes:
S101, setting one or more sensors, respectively acquiring data acquired by the one or more sensors based on a wireless transmission mode, and processing the acquired data to obtain monitoring data, wherein the sensors comprise one or more of a temperature sensor, a humidity sensor, a strain sensor, a displacement sensor and an inclination sensor, and the acquired data comprise one or more of temperatures of different depths of a stirring wall, temperature differences inside and outside the stirring wall, humidity of different depths of the stirring wall, a strain value, a displacement value and an inclination value;
The temperature sensor takes the thickness and the temperature gradient of the cement soil stirring wall into consideration, and adopts a layered arrangement mode. And temperature sensors are respectively arranged at the top, the middle and the bottom of the stirring wall, and the temperature changes at different depths are monitored. Meanwhile, temperature sensors are symmetrically arranged on two sides of the stirring wall, and the temperature difference between the inside and the outside of the stirring wall is monitored;
The humidity sensors are distributed in the stirring wall, and one humidity sensor is respectively distributed at the top, the middle and the bottom of the stirring wall in a layered distribution mode to monitor humidity changes at different depths; arranging a plurality of humidity sensors at each monitoring point, fusing measurement data of the plurality of sensors in a weighted average mode, removing abnormal values, and improving the confidence coefficient of measurement results;
The strain sensor adopts an embedded layout mode and is embedded in the concrete in the pouring process of the stirring wall. And in consideration of the stress distribution characteristics of the stirring wall, strain sensors are mainly arranged at the middle part and the bottom of the stirring wall and monitor the strain in the vertical and horizontal directions. 4 strain sensors are distributed on each monitoring section and are respectively positioned at the center and the edge of the section to form a strain monitoring array; the measuring principle of the vibrating wire type strain sensor is that the natural frequency of the vibrating wire is collected by dialing the vibrating wire, and the strain value is obtained by conversion according to the frequency-strain relation; by adopting a frequency measurement algorithm based on Fast Fourier Transform (FFT), the accuracy of frequency measurement is improved;
The displacement sensor selects a laser displacement sensor and an inclination sensor, and based on a self-adaptive Kalman filtering algorithm for data anomaly detection, a statistical model of data is built by carrying out statistical analysis on displacement and inclination data, so that automatic identification and rejection of anomaly data are realized; then, the displacement and inclination data are fused by utilizing a self-adaptive Kalman filtering algorithm, and the parameters of a filter are dynamically adjusted to realize optimal estimation; the displacement and inclination data after filtering can effectively eliminate measurement noise and abnormal interference, and improve the signal-to-noise ratio and reliability of the data;
taking the factors of complex construction site environment, difficult wiring and the like into consideration, the invention adopts a wireless transmission mode, such as Zigbee, loRa, NB-IoT and the like, and acquires the original data of each monitoring parameter in real time;
In the embodiment, by arranging sensors of different types such as temperature, humidity, strain, displacement, inclination and the like, various physical parameters of the stirring wall in the construction and operation processes can be collected in a multi-dimensional and omnibearing manner, so that the state of the stirring wall is comprehensively monitored; by adopting the wireless transmission technology to collect the monitoring data, the problems of difficult and easy damage of wired transmission wiring can be overcome, the wiring cost is reduced, the flexibility and reliability of data transmission are improved, and the remote monitoring and the real-time monitoring are convenient to realize.
In an alternative embodiment, one or more sensors are provided, data collected by the one or more sensors are respectively obtained based on a wireless transmission mode, and the collected data are processed to obtain monitoring data, which includes:
temperature sensors are respectively arranged at the top, the middle and the bottom of the stirring wall in a layered arrangement mode, so that temperatures of different depths of the stirring wall are collected to serve as monitoring data, and the temperature sensors are symmetrically arranged at the inner side and the outer side of the stirring wall, so that temperature differences between the inner side and the outer side of the stirring wall are collected to serve as monitoring data;
a plurality of humidity sensors are respectively arranged at the top, the middle and the bottom in the stirring wall in a layered arrangement mode so as to collect humidity of different depths of the stirring wall, and weighted average is carried out on data collected by the plurality of humidity sensors to be used as monitoring data;
Strain sensors are respectively arranged at the middle part and the bottom of the stirring wall in an embedded arrangement mode, strain values in the vertical direction and the horizontal direction of the stirring wall are collected to be used as monitoring data, 4 strain sensors are respectively arranged at the center and the edge of a monitoring section to form a strain monitoring array, and the strain values are collected to be used as monitoring data;
Filtering the displacement value acquired by the displacement sensor and the inclination value acquired by the inclination sensor respectively by adopting a self-adaptive Kalman filtering algorithm, wherein filtering parameters are dynamically adjusted to eliminate noise and abnormal interference, and the filtered displacement value and the filtered inclination value are used as monitoring data;
optionally, processing the collected data to obtain the monitoring data further includes:
Preprocessing the collected original data, including data cleaning, outlier detection, data normalization and the like. The data cleaning is to remove noise, repeated values, missing values and other interference factors in the original data; the abnormal value detection is to identify and reject abnormal measuring points which deviate from the normal range obviously; the data normalization is to unify the data with different dimensions to the same dimension, so that the subsequent analysis is convenient;
And fusing the preprocessed multi-source monitoring data, comprehensively utilizing the data of different sensors, and improving the reliability and accuracy of the monitoring data. For example, for temperature monitoring, data of temperature sensors with different depths can be fused to obtain the temperature distribution of the whole stirring wall; for deformation monitoring, the deformation state of the stirring wall can be comprehensively estimated by fusing the data of sensors such as strain, displacement, inclination and the like;
Key characteristic parameters such as temperature gradient, humidity average value, strain average value, displacement rate, inclination angle and the like are extracted from the fused monitoring data, and can reflect key information of the performance of the stirring wall, so that a basis is provided for subsequent state evaluation and prediction;
Filtering the extracted characteristic parameters by adopting a filtering algorithm, further eliminating residual noise and interference, and improving the smoothness and stability of data, wherein the filtering algorithm comprises Kalman filtering, low-pass filtering, wavelet filtering and the like;
the filtered characteristic parameters are subjected to standardization processing and are converted into unified dimensionless indexes, so that comparison and analysis between different monitoring points in different periods are facilitated, wherein the standardization method comprises z-score standardization, min-max standardization and the like;
Optionally, the normalized characteristic parameters can be used as final monitoring data to be transmitted to a monitoring platform in real time for subsequent applications such as visual display, trend analysis, state evaluation, predictive early warning and the like.
In the embodiment, through collecting various monitoring parameters such as temperature, humidity, strain, displacement, inclination and the like, comprehensive evaluation of various states of the stirring wall can be realized, the working performance and the health state of the stirring wall can be comprehensively mastered, and a reliable basis is provided for early warning analysis and decision optimization; the displacement and inclination data are filtered by adopting the self-adaptive Kalman filtering algorithm, and the high-frequency noise and abnormal interference in the monitoring data can be effectively eliminated by dynamically adjusting the filtering parameters, so that the reliability and the accuracy of the monitoring data are improved; through comprehensive adoption of various layout modes such as layered layout, embedded layout, symmetrical layout and the like, the layout positions and the number of the sensors can be optimized, the cost is reduced while the monitoring requirement is met, and the monitoring economy and the monitoring practicability are improved.
S102, carrying out self-adaptive decomposition on the monitoring data, extracting features with different time frequency scales, obtaining a plurality of candidate features, and carrying out dimension reduction and optimization processing on the plurality of candidate features based on joint mutual information by using a joint mutual information minimization method so as to construct a target feature subset, wherein the joint mutual information represents the dependence and the association degree between the features;
Wherein the adaptive decomposition comprises a variational modal decomposition VMD, which is an adaptive signal analysis method with the core purpose of decomposing a complex monitoring signal into a predetermined number of natural modal functions (INTRINSIC MODE FUNCTIONS, IMFs);
And organizing monitoring data such as temperature, humidity, strain, displacement, inclination and the like according to time sequence to form a multi-dimensional time sequence data set, and carrying out self-adaptive decomposition on the multi-dimensional time sequence data by utilizing a VMD algorithm.
In an alternative embodiment, adaptively decomposing the monitored data and extracting features of different time-frequency scales to obtain a plurality of candidate features, including:
carrying out optimization solution on the monitoring data by utilizing a variational modal decomposition VMD model to obtain K intrinsic mode function IMF components, wherein each intrinsic mode function IMF component represents a local oscillation mode of the corresponding monitoring data, the variational modal decomposition VMD model comprises a first objective function and constraint conditions, and the formula of the first objective function of the variational modal decomposition VMD model is as follows:
；
Wherein u k is the kth intrinsic mode function IMF component, w k is the center frequency corresponding to the kth intrinsic mode function IMF component, K is the identity of the intrinsic mode function IMF component, K is the total number of the intrinsic mode function IMF components, t is time, u k (t) is the kth intrinsic mode function IMF component at time t, delta (·) is the dirac function, Representing a first derivative of time t, j being a preset parameter;
The formula of constraint conditions of the variational modal decomposition VMD model is as follows:
；
Wherein z (t) is the monitored data at time t;
Reconstructing the intrinsic mode function IMF components into m-dimensional space vectors for each intrinsic mode function IMF component, calculating the probability of pattern matching based on the distance between the space vectors and a preset similarity threshold, and calculating the sample entropy of each intrinsic mode function IMF component based on the probability ratio of the pattern matching to obtain a plurality of candidate features, wherein the formula is as follows:
；
Wherein SampEn k represents a sample entropy corresponding to a kth intrinsic mode function IMF component, m represents a space dimension, r represents a preset similarity threshold, and B m (r) represents a probability of pattern matching under the space dimension m and the preset similarity threshold r;
Optionally, the monitoring data z (t) is discretely sampled into a time sequence { z1, z2,., zN }, and z (t) is decomposed to obtain K eigenmode components IMF: { u 1(t), u2(t),..., uk (t) }, where each u k (t) reflects the oscillation mode of z (t) at the corresponding frequency w k;
Setting the number K of IMF components to be decomposed, randomly initializing center frequency w k of each IMF component u k (t), and setting balance parameter lambda and time step tau of the VMD model;
Calculating K Wiener filters according to the current w k for each time point t j, and filtering z (t) by using a filter bank to obtain a new estimated value of u k (t);
Calculating a Gaussian smoothing filter according to the current u k (t), and carrying out Gaussian filtering on the z (t) to obtain a new estimated value of the z (t);
Calculating new center frequency of each IMF component u k (t) by utilizing Hilbert transformation, calculating the difference value between the sum of the current IMF components and z (t), and calculating constraint penalty term values according to lambda;
Calculating a first objective function value according to the result obtained in the steps, converging, and outputting K IMF components u k (t) finally obtained;
Optionally, performing matrix reconstruction on each IMF component u k (T), taking m delay components of u k (T) to form an mx 1 vector e (i) = [ u k(i),uk(i+1),...,uk (i+m-1) ]ζ, arranging all e (i) in time sequence to form an mx (N-m+1) matrix Ak, where ak= [ e (1), e (2), e (N-m+1) ];
determining a similarity threshold r and a distance measure G (a, b), presetting a similarity tolerance threshold r (generally taking 0.2 times of u k (t) standard deviation), and selecting a distance measure, such as Euclidean distance or maximum absolute distance;
Constructing a standard pattern matrix Y, and constructing an m multiplied by 1 pattern matrix Y= [ ym, ym, & gt, ym ] by taking a median vector ym of Ak as a reference pattern;
Calculating the distance G ij = D between each vector in Ak and each vector in Y (Ak (: i), Y (: j)), and obtaining a (N-m+1) x 1 distance matrix Gk;
Calculating a matching probability statistic Bm (r) = (Nm (r))/(N-m+1), wherein Nm (r) is the number of vector pairs with a distance smaller than r, and Bm (r) is the matching probability of the modes Ak and Y within a radius r;
calculating sample entropy SampEn (m, rk) = -In (Bm (r)) of the kth IMF component u k (t), wherein the larger the sample entropy value, the higher the uncertainty of u k (t);
Repeating the steps to obtain sample entropy of all K IMF components, thereby obtaining a plurality of candidate features.
In this embodiment, complex signals can be decomposed into a plurality of IMF components with different characteristics through VMD adaptive decomposition, so as to extract multi-scale time-frequency characteristics, the VMD decomposition can automatically find local oscillation modes with smaller scale and weaker intensity in data, which is helpful for revealing hidden internal structures, the IMF components are mapped into vector space, the complex uncertainty degree of the IMF components is quantized by using sample entropy, and the adaptive entropy characteristics based on pattern matching can automatically reflect rules and random characteristics of the data on different time-frequency scales; the time scale range and granularity level of the extracted features can be controlled by adjusting the space dimension m and the similarity threshold r, so that the extracted features are suitable for the requirements of different application tasks, and have strong flexibility; meanwhile, by comprehensively utilizing sample entropy of a plurality of IMF components, the robustness of the method can be improved, and the sensitivity to noise and abnormal values can be reduced.
In an alternative embodiment, performing a dimension reduction and optimization process based on the joint mutual information using a joint mutual information minimization method for a plurality of candidate features to construct a target feature subset, includes:
Determining a selected feature subset from a plurality of candidate features;
calculating mutual information quantity between the plurality of candidate features and the tag feature based on joint probability distribution of the plurality of candidate features and the tag feature and edge probability distribution of the plurality of candidate features and the tag feature;
Calculating the conditional mutual information quantity between the selected feature subset and the tag feature based on the joint probability distribution of the selected feature subset and the tag feature and the edge probability distribution of the selected feature subset and the tag feature under the preset condition;
Mutual information measures the correlation between the feature and the target variable, and the larger the value is, the stronger the dependence between the feature and the target variable is;
Taking the difference value between the mutual information quantity between the plurality of candidate features and the tag features and the conditional mutual information quantity between the selected feature subset and the tag features as the joint mutual information quantity among the selected feature subset, the plurality of candidate features and the tag features, and carrying out optimization solving with the aim of minimizing the joint mutual information quantity so as to carry out dimension reduction and optimization processing on the plurality of candidate features;
And iteratively executing the steps of determining the selected feature subset from the plurality of candidate features and the following steps until the number of the target feature subset reaches a preset feature number threshold or meets a preset performance requirement, so as to obtain the target feature subset.
In an alternative embodiment, the formula for calculating the amount of mutual information between the plurality of candidate features and the tag feature is as follows:
；
Wherein I (X k; C) represents the mutual information quantity between a plurality of candidate features and the tag features, X k represents the candidate features, C represents the tag features, X k is the value of a single candidate feature, C i is the value of a single tag feature, and p () represents a probability distribution function;
The formula for calculating the conditional mutual information quantity between the selected feature subset and the tag features is as follows:
；
Wherein I (X j; c|Y) represents the amount of conditional mutual information between the selected feature subset and the tag features, X j represents the selected features in the selected feature subset, Y is a given condition, X j represents the value of a single selected feature in the selected feature subset, and Y represents the value of the given condition;
the formula for calculating the amount of joint mutual information between the selected feature subset, the plurality of candidate features, and the tag feature is as follows:
；
Wherein I (X k;Xj; C) represents the amount of joint mutual information between the selected feature subset, the plurality of candidate features, and the tag feature;
the formula for optimization solution with the aim of minimizing the amount of the joint mutual information is as follows:
；
Wherein, Representing an optimization objective that minimizes the amount of joint mutual information, F representing a set of a plurality of candidate features, S representing a selected subset of features;
In the embodiment, by minimizing the joint mutual information of the candidate feature and the selected feature to the tag, the redundancy of the candidate feature and the selected feature can be reduced to the greatest extent when a new feature is selected, meanwhile, the correlation with the tag is reserved, and the multi-source heterogeneous feature is effectively fused; the heterogeneous candidate feature set can be processed, and the method is suitable for processing heterogeneous features extracted by fusing multi-source monitoring data, and has remarkable dimension reduction effect; through iterative minimization of the joint mutual information, candidate features with weak correlation on the tag or large redundancy with other features can be gradually removed, so that feature dimensions are effectively reduced, and the calculation complexity of subsequent learning is reduced; through the statistical regularity and discriminant of mutual information measurement, a high-quality feature subset which is most relevant to the tag and least relevant to each other can be automatically selected from the redundant candidate feature sets, and important guarantee is provided for the accuracy and efficiency of subsequent learning analysis tasks.
S103, constructing a target multi-core extreme learning machine model based on different types of kernel functions, and processing the target feature subset by using the target multi-core extreme learning machine model to predict the performance change trend of the cement-soil mixing wall so as to obtain a prediction result, wherein the performance of the cement-soil mixing wall comprises one or more of strength, stability, durability and structure.
In an alternative embodiment, constructing the target multi-core extreme learning machine model based on different types of kernel functions includes:
And combining based on a Gaussian kernel function and a linear kernel function to obtain a composite kernel function, wherein the Gaussian kernel function has the following formula:
；
Wherein, K RBF (·) represents a Gaussian kernel function, n and n' are samples respectively, and σ is a width parameter of the Gaussian kernel;
the formula of the linear kernel function is as follows:
；
Wherein K linear (·) represents a linear kernel function;
taking the composite kernel function as the basis of an initial multi-core extreme learning machine model, and randomly initializing parameters from an input layer to a hidden layer of the initial multi-core extreme learning machine model to construct the initial multi-core extreme learning machine model, wherein the parameters from the input layer to the hidden layer comprise weights and bias values;
The method comprises the steps of firstly introducing training data, setting Gaussian kernel parameters sigma, the number G of hidden layer neurons and regularization coefficients, respectively calculating Gaussian kernels and linear kernels, combining the Gaussian kernels and the linear kernels into a composite kernel matrix, initializing an input layer to the hidden layer parameters by using smaller uniform random numbers, calculating a hidden layer output matrix according to an activation function, and finally obtaining an initial multi-core extreme learning machine model, wherein the model parameters comprise weights from the input layer to the hidden layer, bias values of the hidden layer and the composite kernel matrix;
Training an initial multi-core extreme learning machine model by using a first training sample, optimizing parameters of a composite kernel function and the node number of a hidden layer based on a second objective function, replacing the composite kernel function and the node number of the hidden layer in the initial multi-core extreme learning machine model by the optimized composite kernel function and the node number of the optimized hidden layer, and determining an intermediate multi-core extreme learning machine model;
training the intermediate multi-core extreme learning machine model by using the second training sample, and optimizing parameters from an input layer to a hidden layer of the intermediate multi-core extreme learning machine model based on a third objective function to obtain a target multi-core extreme learning machine model;
The first training sample set is firstly guided into a model of an initial multi-core extreme learning machine, a second objective function is constructed, the second objective function is set to be a weighted sum for minimizing a prediction error and model complexity, the prediction error can be measured by adopting a loss function such as a mean square error, and the model complexity can be represented by a composite kernel function parameter and a norm of the number of hidden layer nodes;
Optimizing the composite kernel function parameters, such as optimizing the width parameters of the Gaussian kernel by network searching, cross validation and other methods, and optimizing the weight parameters of the linear kernel by gradient descent, least square and other methods;
Then, the performance index of the current model is evaluated on the verification set by initializing the hidden layer node number, the hidden node number is gradually increased, the evaluation is repeated, and when the performance index is not improved any more, the hidden layer node number is stopped to select the corresponding optimal hidden layer node number;
The optimized middle multi-core extreme learning machine model is assembled through parameters such as the number of the optimized composite cores and hidden nodes;
secondly, a second training sample set is imported, an intermediate multi-core extreme learning machine model is trained, a third objective function is built, on the verification set, input layer parameters are optimized to the hidden layer through gradient descent and other optimization algorithms to minimize the third objective function, and then the input parameters are updated, so that the target multi-core extreme learning machine model is finally obtained;
In this embodiment, the influence of feature redundancy and noise can be effectively reduced by combining mutual information minimization criteria, the correlation between features in the optimized feature subset and the labels is ensured to be stronger by minimizing the condition mutual information difference between the selected feature subset and the unselected feature set, the feature subset has enough information characterization capability on the labels by the termination condition including two constraints of feature quantity and performance, the feature space dimension is reduced by removing redundancy and irrelevant features, the training and prediction calculation cost of the subsequent machine learning model is reduced, the feature quality after noise reduction optimization is higher, the model generalization performance based on the optimized features is enhanced, and the performance and robustness of the subsequent machine learning model are effectively improved.
In an alternative embodiment, the second objective function is formulated as follows:
；
Wherein L (·) represents a loss function, H represents the number of nodes of the hidden layer, Representing the true value in the d-th cross validation,/>Representing the predicted value output by the initial multi-core extreme learning machine model in the D-th cross validation, D represents the total times of the cross validation,
The formula of the third objective function is as follows:
；
Wherein HE is the hidden layer output matrix, β is the parameters of the input layer to the hidden layer, N is the second training sample, λ is the preset parameters;
The hidden layer output matrix is determined by calculating the similarity degree between each second training sample and the hidden layer node by using a composite kernel function, and the formula is as follows:
；
wherein HE represents a hidden layer output matrix, and alpha is a preset parameter;
the parameters from the input layer to the hidden layer are calculated by using a least square method based on the hidden layer output matrix, and the formula is as follows:
；
Wherein, For the mole-Penrose pseudo-inverse of the hidden layer output matrix HE, Y N is a sample prediction result of a second training sample output by the middle multi-core extreme learning machine model;
processing the target feature subset by using the target multi-core extreme learning machine model to predict the performance change trend of the cement-soil mixing wall to obtain a prediction result, wherein the method comprises the following steps of:
processing the target feature subset by using a composite kernel function of the target multi-core extreme learning machine model to obtain a hidden layer output matrix of the target feature subset, and processing the hidden layer output matrix of the target feature subset by using parameters from an input layer to a hidden layer of the target multi-core extreme learning machine model to obtain a prediction result;
for example, for new input data x, its predicted output Y N can be calculated by the following formula:
wherein H (x) is the hidden layer output calculated from the input x and training data, Is the output layer weight.
In the embodiment, by introducing a plurality of different types of kernel functions, a high-efficiency and robust health diagnosis framework is constructed, so that the characteristics related to the cement mixing wall performance change can be automatically mined from complex heterogeneous monitoring data efficiently, large-scale monitoring data can be processed efficiently, accurate and real-time monitoring results can be provided in complex alpine environments, and the safety and reliability of the application of the cement mixing wall in alpine regions are greatly improved. In addition, the embodiment of the invention has stronger adaptability and robustness, can stably work in different alpine environments, and provides a new and effective monitoring means for civil engineering construction in alpine regions.
Fig. 2 is a schematic structural diagram of an intelligent monitoring system for an equal-thickness cement-soil mixing wall in a severe cold environment according to an embodiment of the present invention, as shown in fig. 2, the system includes:
The first unit is used for setting one or more sensors, respectively acquiring data acquired by the one or more sensors based on a wireless transmission mode, and processing the acquired data to obtain monitoring data, wherein the sensors comprise one or more of a temperature sensor, a humidity sensor, a strain sensor, a displacement sensor and an inclination sensor, and the acquired data comprise one or more of temperatures of different depths of a stirring wall, temperature differences inside and outside the stirring wall, humidity of different depths of the stirring wall, a strain value, a displacement value and an inclination value;
The second unit is used for carrying out self-adaptive decomposition on the monitoring data and extracting features with different time frequency scales to obtain a plurality of candidate features, carrying out dimension reduction and optimization processing on the plurality of candidate features by utilizing a joint mutual information minimization method based on joint mutual information so as to construct a target feature subset, wherein the joint mutual information represents the dependence and the association degree between the features;
And the third unit is used for constructing a target multi-core extreme learning machine model based on different types of kernel functions, processing the target feature subset by using the target multi-core extreme learning machine model so as to predict the performance change trend of the cement mixing wall and obtain a prediction result, wherein the performance of the cement mixing wall comprises one or more of strength, stability, durability and structure.
In a third aspect of an embodiment of the present invention,
There is provided an electronic device including:
A processor;
a memory for storing processor-executable instructions;
Wherein the processor is configured to invoke the instructions stored in the memory to perform the method described previously.
In a fourth aspect of an embodiment of the present invention,
There is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method as described above.
The present invention may be a method, apparatus, system, and/or computer program product. The computer program product may include a computer readable storage medium having computer readable program instructions embodied thereon for performing various aspects of the present invention.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (8)
1. An intelligent monitoring method for an equal-thickness cement soil stirring wall in a severe cold environment is characterized by comprising the following steps:
Setting one or more sensors, respectively acquiring data acquired by the one or more sensors based on a wireless transmission mode, and processing the acquired data to obtain monitoring data, wherein the sensors comprise one or more of a temperature sensor, a humidity sensor, a strain sensor, a displacement sensor and an inclination sensor, and the acquired data comprise one or more of temperatures of different depths of a stirring wall, temperature differences inside and outside the stirring wall, humidity of different depths of the stirring wall, a strain value, a displacement value and an inclination value;
Performing self-adaptive decomposition on the monitoring data, extracting features with different time frequency scales, obtaining a plurality of candidate features, performing dimension reduction and optimization processing on the plurality of candidate features by utilizing a joint mutual information minimization method based on joint mutual information so as to construct a target feature subset, wherein the joint mutual information represents the dependence and the association degree between the features;
Constructing a target multi-core extreme learning machine model based on different types of kernel functions, and processing the target feature subset by using the target multi-core extreme learning machine model to predict the performance change trend of the cement-soil mixing wall so as to obtain a prediction result, wherein the performance of the cement-soil mixing wall comprises one or more of strength, stability, durability and structure;
Performing a dimension reduction and optimization process based on the joint mutual information by using a joint mutual information minimization method for a plurality of candidate features to construct a target feature subset, including:
Determining a selected feature subset from a plurality of candidate features;
calculating mutual information quantity between the plurality of candidate features and the tag feature based on joint probability distribution of the plurality of candidate features and the tag feature and edge probability distribution of the plurality of candidate features and the tag feature;
Calculating the conditional mutual information quantity between the selected feature subset and the tag feature based on the joint probability distribution of the selected feature subset and the tag feature and the edge probability distribution of the selected feature subset and the tag feature under the preset condition;
Taking the difference value between the mutual information quantity between the plurality of candidate features and the tag features and the conditional mutual information quantity between the selected feature subset and the tag features as the joint mutual information quantity among the selected feature subset, the plurality of candidate features and the tag features, and carrying out optimization solving with the aim of minimizing the joint mutual information quantity so as to carry out dimension reduction and optimization processing on the plurality of candidate features;
iteratively executing the steps of determining a selected feature subset from a plurality of candidate features and thereafter until the number of target feature subsets reaches a preset feature number threshold or meets a preset performance requirement to obtain target feature subsets;
the formula for calculating the mutual information quantity between the plurality of candidate features and the tag feature is as follows:
；
Wherein I (X k; C) represents the mutual information quantity between a plurality of candidate features and the tag features, X k represents the candidate features, C represents the tag features, X k is the value of a single candidate feature, C i is the value of a single tag feature, and p () represents a probability distribution function;
The formula for calculating the conditional mutual information quantity between the selected feature subset and the tag features is as follows:
；
Wherein I (X j; c|Y) represents the amount of conditional mutual information between the selected feature subset and the tag features, X j represents the selected features in the selected feature subset, Y is a given condition, X j represents the value of a single selected feature in the selected feature subset, and Y represents the value of the given condition;
the formula for calculating the amount of joint mutual information between the selected feature subset, the plurality of candidate features, and the tag feature is as follows:
；
Wherein I (X k;Xj; C) represents the amount of joint mutual information between the selected feature subset, the plurality of candidate features, and the tag feature;
the formula for optimization solution with the aim of minimizing the amount of the joint mutual information is as follows:
；
Wherein, Representing an optimization objective that minimizes the amount of joint mutual information, F represents a set of multiple candidate features, and S represents a selected subset of features.
2. The method of claim 1, wherein providing one or more sensors, acquiring data acquired by the one or more sensors based on wireless transmission, and processing the acquired data to obtain the monitoring data, comprises:
temperature sensors are respectively arranged at the top, the middle and the bottom of the stirring wall in a layered arrangement mode, so that temperatures of different depths of the stirring wall are collected to serve as monitoring data, and the temperature sensors are symmetrically arranged at the inner side and the outer side of the stirring wall, so that temperature differences between the inner side and the outer side of the stirring wall are collected to serve as monitoring data;
a plurality of humidity sensors are respectively arranged at the top, the middle and the bottom in the stirring wall in a layered arrangement mode so as to collect humidity of different depths of the stirring wall, and weighted average is carried out on data collected by the plurality of humidity sensors to be used as monitoring data;
Strain sensors are respectively arranged at the middle part and the bottom of the stirring wall in an embedded arrangement mode, strain values in the vertical direction and the horizontal direction of the stirring wall are collected to be used as monitoring data, 4 strain sensors are respectively arranged at the center and the edge of a monitoring section to form a strain monitoring array, and the strain values are collected to be used as monitoring data;
And filtering the displacement value acquired by the displacement sensor and the inclination value acquired by the inclination sensor respectively by adopting a self-adaptive Kalman filtering algorithm, wherein filtering parameters are dynamically adjusted to eliminate noise and abnormal interference, and the filtered displacement value and the filtered inclination value are used as monitoring data.
3. The method of claim 1, wherein the adaptive decomposition comprises a variational modal decomposition VMD, adaptively decomposing the monitored data and extracting features of different time-frequency scales to obtain a plurality of candidate features, comprising:
carrying out optimization solution on the monitoring data by utilizing a variational modal decomposition VMD model to obtain K intrinsic mode function IMF components, wherein each intrinsic mode function IMF component represents a local oscillation mode of the corresponding monitoring data, the variational modal decomposition VMD model comprises a first objective function and constraint conditions, and the formula of the first objective function of the variational modal decomposition VMD model is as follows:
；
Wherein u k is the kth intrinsic mode function IMF component, w k is the center frequency corresponding to the kth intrinsic mode function IMF component, K is the identity of the intrinsic mode function IMF component, K is the total number of the intrinsic mode function IMF components, t is time, u k (t) is the kth intrinsic mode function IMF component at time t, delta (·) is the dirac function, Representing a first derivative of time t, j being a preset parameter;
The formula of constraint conditions of the variational modal decomposition VMD model is as follows:
；
Wherein z (t) is the monitored data at time t;
Reconstructing the intrinsic mode function IMF components into m-dimensional space vectors for each intrinsic mode function IMF component, calculating the probability of pattern matching based on the distance between the space vectors and a preset similarity threshold, and calculating the sample entropy of each intrinsic mode function IMF component based on the probability ratio of the pattern matching to obtain a plurality of candidate features, wherein the formula is as follows:
；
Wherein SampEn k represents the sample entropy corresponding to the kth intrinsic mode function IMF component, m represents the spatial dimension, r represents the preset similarity threshold, and B m (r) represents the probability of pattern matching in the spatial dimension m and the preset similarity threshold r.
4. The method of claim 1, wherein constructing the target multi-core extreme learning machine model based on different types of kernel functions comprises:
And combining based on a Gaussian kernel function and a linear kernel function to obtain a composite kernel function, wherein the Gaussian kernel function has the following formula:
；
Wherein K RBF (·) represents a Gaussian kernel function, n represents a sample of a target feature subset, n' represents a sample of a non-target feature subset, σ is a width parameter of the Gaussian kernel;
the formula of the linear kernel function is as follows:
；
Wherein K linear (·) represents a linear kernel function;
taking the composite kernel function as the basis of an initial multi-core extreme learning machine model, and randomly initializing parameters from an input layer to a hidden layer of the initial multi-core extreme learning machine model to construct the initial multi-core extreme learning machine model, wherein the parameters from the input layer to the hidden layer comprise weights and bias values;
Training an initial multi-core extreme learning machine model by using a first training sample, optimizing parameters of a composite kernel function and the node number of a hidden layer based on a second objective function, replacing the composite kernel function and the node number of the hidden layer in the initial multi-core extreme learning machine model by the optimized composite kernel function and the node number of the optimized hidden layer, and determining an intermediate multi-core extreme learning machine model;
And training the intermediate multi-core extreme learning machine model by using the second training sample, and optimizing parameters from an input layer to a hidden layer of the intermediate multi-core extreme learning machine model based on a third objective function to obtain the target multi-core extreme learning machine model.
5. The method of claim 4, wherein the second objective function is formulated as follows:
；
Wherein L (·) represents a loss function, H represents the number of nodes of the hidden layer, Representing the true value in the d-th cross validation,/>Representing the predicted value output by the initial multi-core extreme learning machine model in the D-th cross validation, D represents the total times of the cross validation,
The formula of the third objective function is as follows:
；
Wherein HE is the hidden layer output matrix, β is the parameters of the input layer to the hidden layer, N is the second training sample, λ is the preset parameters;
The hidden layer output matrix is determined by calculating the similarity degree between each second training sample and the hidden layer node by using a composite kernel function, and the formula is as follows:
；
wherein HE represents a hidden layer output matrix, and alpha is a preset parameter;
the parameters from the input layer to the hidden layer are calculated by using a least square method based on the hidden layer output matrix, and the formula is as follows:
；
Wherein, For the mole-Penrose pseudo-inverse of the hidden layer output matrix HE, Y N is a sample prediction result of a second training sample output by the middle multi-core extreme learning machine model;
processing the target feature subset by using the target multi-core extreme learning machine model to predict the performance change trend of the cement-soil mixing wall to obtain a prediction result, wherein the method comprises the following steps of:
and processing the target feature subset by using a composite kernel function of the target multi-core extreme learning machine model to obtain a hidden layer output matrix of the target feature subset, and processing the hidden layer output matrix of the target feature subset by using parameters from an input layer to a hidden layer of the target multi-core extreme learning machine model to obtain a prediction result.
6. An intelligent monitoring system for an equal-thickness cement-soil mixing wall in a severe cold environment, for implementing the method of any one of the preceding claims 1-5, comprising:
The first unit is used for setting one or more sensors, respectively acquiring data acquired by the one or more sensors based on a wireless transmission mode, and processing the acquired data to obtain monitoring data, wherein the sensors comprise one or more of a temperature sensor, a humidity sensor, a strain sensor, a displacement sensor and an inclination sensor, and the acquired data comprise one or more of temperatures of different depths of a stirring wall, temperature differences inside and outside the stirring wall, humidity of different depths of the stirring wall, a strain value, a displacement value and an inclination value;
The second unit is used for carrying out self-adaptive decomposition on the monitoring data and extracting features with different time frequency scales to obtain a plurality of candidate features, carrying out dimension reduction and optimization processing on the plurality of candidate features by utilizing a joint mutual information minimization method based on joint mutual information so as to construct a target feature subset, wherein the joint mutual information represents the dependence and the association degree between the features;
And the third unit is used for constructing a target multi-core extreme learning machine model based on different types of kernel functions, processing the target feature subset by using the target multi-core extreme learning machine model so as to predict the performance change trend of the cement mixing wall and obtain a prediction result, wherein the performance of the cement mixing wall comprises one or more of strength, stability, durability and structure.
7. An electronic device, comprising:
A processor;
a memory for storing processor-executable instructions;
Wherein the processor is configured to invoke the instructions stored in the memory to perform the method of any of claims 1 to 5.
8. A computer readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410425088.5A CN118010103B (en) | 2024-04-10 | 2024-04-10 | Intelligent monitoring method and system for equal-thickness cement soil stirring wall in severe cold environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410425088.5A CN118010103B (en) | 2024-04-10 | 2024-04-10 | Intelligent monitoring method and system for equal-thickness cement soil stirring wall in severe cold environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118010103A CN118010103A (en) | 2024-05-10 |
| CN118010103B true CN118010103B (en) | 2024-06-14 |
Family
ID=90948946
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410425088.5A Active CN118010103B (en) | 2024-04-10 | 2024-04-10 | Intelligent monitoring method and system for equal-thickness cement soil stirring wall in severe cold environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118010103B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112085619A (en) * | 2020-08-10 | 2020-12-15 | 国网上海市电力公司 | Feature selection method for power distribution network data optimization |
| CN117005471A (en) * | 2023-06-25 | 2023-11-07 | 中冶成都勘察研究总院有限公司 | Pit deformation early warning and monitoring method based on multi-parameter variables |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114139634A (en) * | 2021-12-03 | 2022-03-04 | 吉林大学 | Multi-label feature selection method based on paired label weights |
| CN114692692B (en) * | 2022-04-02 | 2023-05-12 | 河海大学 | Snowfall recognition method based on microwave attenuation signal fusion kernel extreme learning machine |
-
2024
- 2024-04-10 CN CN202410425088.5A patent/CN118010103B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112085619A (en) * | 2020-08-10 | 2020-12-15 | 国网上海市电力公司 | Feature selection method for power distribution network data optimization |
| CN117005471A (en) * | 2023-06-25 | 2023-11-07 | 中冶成都勘察研究总院有限公司 | Pit deformation early warning and monitoring method based on multi-parameter variables |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118010103A (en) | 2024-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Raissi et al. | Statistical process optimization through multi-response surface methodology | |
| CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
| CN111967486A (en) | Complex equipment fault diagnosis method based on multi-sensor fusion | |
| CN115169479A (en) | Remote monitoring method, system and storage medium for sewage treatment process | |
| CN108445752A (en) | A kind of random weight Artificial neural network ensemble modeling method of adaptively selected depth characteristic | |
| CN108399434B (en) | Analysis and prediction method of high-dimensional time series data based on feature extraction | |
| CN110689183B (en) | Cluster photovoltaic power probability prediction method, system, medium and electronic device | |
| CN113901977A (en) | Deep learning-based power consumer electricity stealing identification method and system | |
| Guo et al. | A hybrid prognosis scheme for rolling bearings based on a novel health indicator and nonlinear Wiener process | |
| CN115495991A (en) | Rainfall interval prediction method based on time convolution network | |
| CN115271186B (en) | Reservoir water level prediction and early warning method based on delay factor and PSO RNN Attention model | |
| Wang et al. | Remaining useful life prediction of bearings based on convolution attention mechanism and temporal convolution network | |
| CN114091349A (en) | Multi-source field self-adaption based rolling bearing service life prediction method | |
| CN116738868A (en) | Rolling bearing residual life prediction method | |
| CN115510748A (en) | Landslide displacement prediction method based on variational modal decomposition and CNN-GRU | |
| CN116522124A (en) | Dissolved oxygen content prediction method and system based on influence of environmental factors | |
| CN117289668B (en) | Distributed speed reducer network cooperative control method, device, equipment and storage medium | |
| CN118010103B (en) | Intelligent monitoring method and system for equal-thickness cement soil stirring wall in severe cold environment | |
| Dang et al. | seq2graph: Discovering dynamic non-linear dependencies from multivariate time series | |
| CN117195114A (en) | Chemical production line identification method and system | |
| CN116595465A (en) | High-dimensional sparse data outlier detection method and system based on self-encoder and data enhancement | |
| CN116610973A (en) | Sensor fault monitoring and failure information reconstruction method and system | |
| CN113688773B (en) | Storage tank dome displacement data restoration method and device based on deep learning | |
| CN112069621B (en) | Method for predicting residual service life of rolling bearing based on linear reliability index | |
| CN113151842B (en) | Method and device for determining conversion efficiency of wind-solar complementary water electrolysis hydrogen production |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |