CN115392349A - Fault diagnosis method and device for cutting part of heading machine and heading machine - Google Patents
Fault diagnosis method and device for cutting part of heading machine and heading machine Download PDFInfo
- Publication number
- CN115392349A CN115392349A CN202210919624.8A CN202210919624A CN115392349A CN 115392349 A CN115392349 A CN 115392349A CN 202210919624 A CN202210919624 A CN 202210919624A CN 115392349 A CN115392349 A CN 115392349A
- Authority
- CN
- China
- Prior art keywords
- neural network
- cutting part
- heading machine
- fault diagnosis
- algorithm
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Physiology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a fault diagnosis method and a fault diagnosis device for a cutting part of a heading machine and the heading machine, and relates to the technical field of fault diagnosis of heading machines, wherein the fault diagnosis method comprises the following steps: acquiring a data sample of the cutting part; constructing a characteristic sample set based on the data samples; constructing a BP neural network according to the characteristic sample set; optimizing the BP neural network by adopting a genetic algorithm to obtain a GA-BP neural network; combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers; and carrying out fault diagnosis on the cutting part by using a strong classifier model. The fault diagnosis method for the cutting part of the heading machine provided by the invention integrates the advantages of different sequences, achieves the purpose of 'double optimization' of the BP neural network, and greatly improves the prediction precision. The health state of the cutting part of the heading machine can be effectively monitored and diagnosed, and the diagnosis precision and accuracy are high.
Description
Technical Field
The invention relates to the technical field of fault diagnosis of a heading machine, in particular to a fault diagnosis method and device for a cutting part of the heading machine and the heading machine.
Background
The intelligent heading machine is used as core power equipment for underground coal mine excavation and plays a key role in safe and efficient production of coal mines. Due to the complex and severe working environment, the cutting part of the heading machine is easy to break down in the working process. And the coal mine has special requirements on a detection system and instrument equipment, so that the problems of large technical bottleneck, multiple interference factors, large difficulty in data and data acquisition and the like of the fully mechanized excavation face heading machine state parameter detection are brought. Therefore, finding a method suitable for diagnosing the fault of the cutting part of the excavator is particularly important.
In the related art, a fault diagnosis method for a cutting part of a heading machine based on a Back probabilistic neural Network (BP) is provided, and whether the cutting part of the heading machine breaks down or not is diagnosed. However, the model established by the diagnosis method is single in structure, the accuracy and precision of the model are not high, and the model is easy to fall into a local minimum value. Specifically, the local minimum problem is that after the BP neural network is trained for a certain number of times, although errors still exist, the speed of reducing the network errors is very slow, and even does not change any more, so that the diagnosis precision and accuracy of the cutting part of the heading machine are low.
Therefore, how to provide a fault diagnosis method for the cutting part of the heading machine, which can improve the diagnosis precision and accuracy, becomes a problem to be solved urgently at present.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art or the related art.
Therefore, an object of the present invention is to provide a method of diagnosing a failure of a cutting unit of a heading machine.
A second object of the present invention is to provide a failure diagnosis device for a cutting unit of a heading machine.
A third object of the present invention is to provide a failure diagnosis device for a cutting unit of a heading machine.
It is a fourth object of the invention to provide a readable storage medium.
A fifth object of the present invention is to provide a heading machine.
In order to achieve the above object, a technical solution of a first aspect of the present invention provides a fault diagnosis method for a cutting part of a heading machine, including: acquiring a data sample of the cutting part; constructing a characteristic sample set based on the data samples; constructing a BP neural network according to the characteristic sample set; optimizing a BP neural Network (Back propagation neural Network) by adopting a Genetic Algorithm (GA for short) to obtain a GA-BP neural Network; combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers; and carrying out fault diagnosis on the cutting part by using a strong classifier model.
According to the fault diagnosis method for the cutting part of the heading machine, provided by the invention, the BP neural network optimized by a genetic algorithm is used as a basic classifier, the GA-BP neural network and the iterative algorithm are effectively combined according to the enhancement capability of the iterative algorithm, the prediction sample output of the GA-BP neural network is repeatedly trained, the weight and the threshold of the BP neural network are optimized by utilizing the optimization capability of the genetic algorithm, and a strong classifier model consisting of a plurality of GA-BP neural network weak classifiers is obtained through combination of the iterative algorithm, so that the fault diagnosis of the cutting part of the heading machine by the strong classifier is realized. The strong classifier model integrates the characteristics of Genetic Algorithm (GA) global optimization and BP neural network local optimization, and the strong classifier of the iterative algorithm gives different weights to a plurality of prediction sequences of the weak classifier, integrates the advantages of the different sequences, achieves the aim of 'double optimization' of the BP neural network, and greatly improves the prediction precision. Compared with the fault diagnosis method of the cutting part of the heading machine based on the multi-sensor information BP neural network in the related technology, the fault diagnosis method of the cutting part of the heading machine can effectively monitor and diagnose the health state of the cutting part of the heading machine, and the diagnosis precision and accuracy are high.
In addition, the fault diagnosis method for the cutting part of the heading machine provided by the application can also have the following additional technical characteristics:
in the above technical solution, the step of obtaining the data sample of the cutting part specifically includes: obtaining an effective value of the vibration energy of the cutting part; acquiring the kurtosis of the impact degree of the cutting part; and acquiring the standard deviation of the cutting part representing the degree of deviation of the vibration signal from the mean value.
In the technical scheme, the step of obtaining the data sample of the cutting part specifically comprises the following steps: and selecting an input variable which has small correlation and is sensitive to faults, reflecting the effective value of the vibration energy of the cutting part, reflecting the degree of impact on the cutting part and representing the standard deviation of the vibration signal from the mean value, and using the input variable for diagnosing the faults of the cutting part of the heading machine. If the vibration energy is changed violently under the normal condition, the mechanical fault of the cutting part of the heading machine can be inferred. Vibration energy changes are used as a criterion to be introduced into fault diagnosis of the cutting part of the heading machine, vibration signals are stable under the normal working condition of the heading machine, and the vibration energy, the received impact and the vibration deviation from the mean value are greatly larger than those under the normal condition under the fault condition, so that whether the cutting part of the heading machine breaks down or not can be effectively diagnosed.
In the above technical solution, the step of constructing the feature sample set based on the data sample includes: and processing the data samples, and selecting a characteristic sample set by combining a factor analysis method.
In the technical scheme, the step of constructing the feature sample set based on the data samples comprises the following steps: and processing the data samples, and selecting a characteristic sample set by combining a factor analysis method. The data samples are processed, and a factor analysis method is combined to select a characteristic sample set, so that the fault of the cutting part of the heading machine can be effectively identified.
In the above technical solution, the iterative algorithm includes an Adaboost algorithm (an iterative algorithm), the GA-BP neural network is combined with the Adaboost algorithm to obtain a plurality of GA-BP neural network weak classifiers, and the step of forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers includes: searching a plurality of weak classifiers, training the GA-BP neural network by using training data and outputting prediction training data to obtain error sum of prediction classification results; calculating the weight of the prediction classification result according to the prediction error sum of the prediction classification result; adjusting the weight of the training data of the next round according to the weight of the prediction classification result; and carrying out weighted combination on the plurality of trained weak classifiers to obtain a strong classifier model.
In the technical scheme, the Adaboost algorithm is an integrated learning method, the performance of the learning algorithm can be obviously improved, the weight distribution of each learning sample is obtained, equal initial distribution weights are provided, and the weights of the samples are continuously adjusted in the training process. It trains multiple different weak classifiers or weak classifiers to the same training set, weights each sample through the performance in the training process, and transmits it downwards, so as to strengthen the training degree of the sample with large error in the next weak classifier or weak classifier training, and finally uses the linear combination method to assemble a strong classifier or strong classifier, to realize the great improvement of classification and prediction accuracy, and to make the modeling result more stable until reaching a certain predetermined small enough error rate. In the training process by using the Adaboost algorithm, each training sample is assigned with a weight, and the weight can be regarded as the probability that the training sample is selected into a training set by a certain classifier. If a sample point has been accurately classified, then in the next training set, the probability that it is selected is reduced; instead, the weight of this sample point is increased. In this way, the Adaboost algorithm can "focus" on samples where the information points are rich and difficult to distinguish.
In the above technical solution, the step of obtaining the GA-BP neural network by optimizing the BP neural network using a genetic algorithm comprises: and initializing the weight value and the threshold value of the BP neural network.
In the technical scheme, the step of optimizing the BP neural network by adopting a genetic algorithm to obtain the GA-BP neural network comprises the following steps: and initializing the weight and the threshold of the BP neural network so as to optimize the weight and the threshold of the BP neural network by utilizing the optimizing capability of a genetic algorithm.
In the above technical solution, combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and the step of forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers comprises: the raw data is pre-processed including data and quantization normalization.
In the technical scheme, the step of combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers comprises the following steps: the method comprises the steps of preprocessing original data including data and quantization normalization, selecting multiple groups of training data from a sample space, and initializing distribution weights of the training data to preprocess the original data.
The technical scheme of the second aspect of the invention provides a fault diagnosis device for a cutting part of a heading machine, which comprises the following components: the acquisition module is used for acquiring a data sample of the cutting part; the first construction module is used for constructing a characteristic sample set based on the data sample; the second construction module is used for constructing a BP neural network according to the characteristic sample set; the neural network optimization module is used for optimizing the BP neural network by adopting a genetic algorithm to obtain a GA-BP neural network; the strong classifier module is used for combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers; and the fault diagnosis module is used for diagnosing faults of the cutting part by utilizing the strong classifier model.
The fault diagnosis device for the cutting part of the heading machine comprises an acquisition module, a first construction module, a second construction module, a neural network optimization module, a strong classifier module and a fault diagnosis module. The acquisition module can acquire data samples of the cutting part. The first construction module can construct a characteristic sample set based on the data samples. The second construction module can construct the BP neural network according to the characteristic sample set. The neural network optimization module can optimize the BP neural network by adopting a genetic algorithm to obtain the GA-BP neural network. The strong classifier module can combine the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and a strong classifier model is formed according to the plurality of GA-BP neural network weak classifiers. The fault diagnosis module can perform fault diagnosis on the cutting part by using the strong classifier model. The strong classifier model integrates the characteristics of Genetic Algorithm (GA) global optimization and BP neural network local optimization, and the strong classifier of the iterative algorithm gives different weights to a plurality of prediction sequences of the weak classifier, integrates the advantages of the different sequences, achieves the aim of 'double optimization' of the BP neural network, and greatly improves the prediction precision. Compared with the fault diagnosis method for the cutting part of the heading machine based on the multi-sensor information BP neural network in the related art, the fault diagnosis method for the cutting part of the heading machine can effectively monitor and diagnose the health state of the cutting part of the heading machine, and is high in diagnosis precision and accuracy.
The technical scheme of the third aspect of the invention provides a fault diagnosis device for a cutting part of a heading machine, which comprises the following components: the fault diagnosis method comprises a memory and a processor, wherein the memory stores programs or instructions, and the programs or instructions are executed by the processor to realize the steps of the fault diagnosis method for the cutting part of the heading machine in any one technical scheme of the first aspect.
The fault diagnosis device for the cutting part of the heading machine provided by the invention comprises a memory and a processor, wherein the memory stores programs or instructions, and the programs or instructions are executed by the processor to realize the fault diagnosis method for the cutting part of the heading machine in any technical scheme of the first aspect. Therefore, the fault diagnosis device for the cutting part of the heading machine provided by the invention also has all the beneficial effects of the fault diagnosis method for the cutting part of the heading machine provided by any one of the technical schemes in the first aspect, and details are not repeated herein.
An aspect of a fourth aspect of the present invention provides a readable storage medium on which a program or instructions are stored, which when executed, implement the steps of the method of diagnosing a fault of a cutting section of a heading machine of the first aspect.
According to the readable storage medium provided by the invention, the method for diagnosing the fault of the cutting part of the heading machine is realized. Therefore, the readable storage medium provided by the invention also has all the beneficial effects of the fault diagnosis method for the cutting part of the heading machine provided by any technical scheme of the first aspect, and details are not repeated herein.
The technical scheme of the fifth aspect of the invention provides a heading machine, which comprises: the fault diagnosis device for the cutting part of the heading machine according to the second or third aspect, and/or the readable storage medium according to the fourth aspect.
The heading machine provided by the invention comprises the fault diagnosis device of the cutting part of the heading machine of the second aspect or the third aspect, and/or the readable storage medium of the fourth aspect. Therefore, the heading machine provided by the invention also has the whole beneficial effects of the fault diagnosis device of the cutting part of the heading machine in the second aspect or the third aspect, and/or the readable storage medium in the fourth aspect, which are not described herein again.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a flowchart of a fault diagnosis method of a cutting part of a heading machine according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a comprehensive structure of a GA-BP-Adaboost model according to an embodiment of the present invention;
FIG. 3 is a flow diagram of a GA-BP-Adaboost algorithm according to an embodiment of the invention;
FIG. 4 is a graph comparing prediction errors for standard BP and GA-BP neural networks according to an embodiment of the present invention;
FIG. 5 is a graph of predicted and actual values versus a standard BP and GA-BP neural networks according to an embodiment of the present invention;
FIG. 6 is a graph of the results of the GA-BP-Adaboost algorithm test, according to an embodiment of the present invention.
Detailed Description
So that the manner in which the above recited objects, features and advantages of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to the embodiments thereof which are illustrated in the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.
A method of diagnosing a fault of a cutting part of a heading machine according to some embodiments of the present invention will be described below with reference to fig. 1 to 6.
Example one
As shown in fig. 1, a first embodiment of a first aspect of the present invention provides a fault diagnosis method for a cutting part of a heading machine, including:
s101, acquiring data samples of the cutting part.
And S102, constructing a characteristic sample set based on the data samples.
And S103, constructing the BP neural network according to the characteristic sample set.
And S104, optimizing the BP neural network by adopting a genetic algorithm to obtain the GA-BP neural network.
And S105, combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers.
And S106, carrying out fault diagnosis on the cutting part by using the strong classifier model.
According to the fault diagnosis method for the cutting part of the heading machine, provided by the invention, the BP neural network optimized by the genetic algorithm is used as a basic classifier, the GA-BP neural network and the iterative algorithm are effectively combined according to the enhancement capability of the iterative algorithm, the prediction sample output of the GA-BP neural network is repeatedly trained, the weight and the threshold of the BP network are optimized by utilizing the optimization capability of the genetic algorithm, and a strong classifier model consisting of a plurality of GA-BP neural network weak classifiers is obtained through the combination of the iterative algorithm, so that the fault diagnosis of the cutting part of the heading machine by the strong classifier is realized. The strong classifier model integrates the characteristics of Genetic Algorithm (GA) global optimization and BP neural network local optimization, and the strong classifier of the iterative algorithm gives different weights to a plurality of prediction sequences of the weak classifier, integrates the advantages of the different sequences, achieves the aim of 'double optimization' of the BP neural network, and greatly improves the prediction precision. Compared with the fault diagnosis method of the cutting part of the heading machine based on the multi-sensor information BP neural network in the related technology, the fault diagnosis method of the cutting part of the heading machine can effectively monitor and diagnose the health state of the cutting part of the heading machine, and the diagnosis precision and accuracy are high.
In the above embodiment, the step of acquiring the data sample of the cutting part specifically includes: obtaining an effective value of the vibration energy of the cutting part; acquiring the degree of impact on the cutting part; and acquiring the standard deviation of the cutting part representing the degree of deviation of the vibration signal from the mean value.
In this embodiment, the step of acquiring the data sample of the cutting portion specifically includes: and selecting an input variable which has small relevance and is sensitive to faults, reflecting an effective value of the vibration energy of the cutting part, reflecting the kurtosis of the impact degree of the cutting part and representing the standard deviation of the vibration signal from the mean value, and using the kurtosis and the standard deviation to diagnose the faults of the cutting part of the heading machine. If the vibration energy is changed violently under the normal condition, the mechanical fault of the cutting part of the heading machine can be inferred. Vibration energy changes are used as criteria to be introduced into fault diagnosis of the cutting part of the heading machine, vibration signals are stable under the normal working condition of the heading machine, and the vibration energy, the received impact and the vibration deviation degree from the mean value under the fault condition are much larger than those under the normal condition, so that whether the cutting part of the heading machine breaks down or not can be effectively diagnosed.
In the above embodiment, the step of constructing the feature sample set based on the data samples previously comprises: and processing the data samples, and selecting a characteristic sample set by combining a factor analysis method.
In this embodiment, the step of constructing the feature sample set based on the data samples previously comprises: and processing the data samples, and selecting a characteristic sample set by combining a factor analysis method. By processing the data samples and selecting the characteristic sample set by combining a factor analysis method, the fault of the cutting part of the heading machine can be effectively identified.
In the above embodiment, the iterative algorithm includes an Adaboost algorithm (an iterative algorithm), the GA-BP neural network is combined with the Adaboost algorithm to obtain a plurality of GA-BP neural network weak classifiers, and the step of forming the strong classifier model according to the plurality of GA-BP neural network weak classifiers includes:
find weak classifier g t (T =1,2,3 \ 8230;, T): when the t-th weak classifier is trained, the GA-BP neural network is trained by using training data and the predicted training data is output, so that the error sum e of the predicted classification result g (t) is obtained t The specific relationship is as follows:
g(t)≠y
wherein: y is the expected classification result, g (t) is the predicted classification result, D i (i) As weight distribution of training data, e t Is a weak classifier with weight distribution D i The error in (2);
calculating the weight of the prediction classification result: prediction error sum e from prediction classification result g (t) t Calculating the weight a of the predicted classification result t The method comprises the following steps:
updating the weight: according to the weight a of the prediction classification result t And adjusting the weight of the training data of the next round as follows:
in the formula: b is t Is a normalization factor, in order to make the sum of the distribution weights 1,
namely, it is
y t Is a relevant classification condition;
strong classifier function: obtaining T weak classifier functions g after training T rounds t (x) (T =1,2,3 \ 8230;, T), and then weighted by T weak classifier functions, a strong classifier function H (x) is obtained:
in this embodiment, the Adaboost algorithm is an integrated learning method, which can significantly improve the performance of the learning algorithm, obtain the weight distribution of each learning sample, provide equal initial distribution weights, and continuously adjust the weights of the samples during the training process. The method trains a plurality of different weak classifiers or weak classifiers to the same training set, weights are given to each sample through the performance in the training process, and the weighted samples are transmitted downwards so as to strengthen the training degree of the samples with large errors in the training of the next weak classifier or weak classifier, and finally a strong classifier or a strong classifier is assembled by using a linear combination method, thereby realizing the great improvement of the classification and prediction accuracy and enabling the modeling result to be more stable until a certain preset error rate which is small enough is reached. In the training process by using the Adaboost algorithm, each training sample is assigned with a weight, and the weight can be regarded as the probability that the training sample is selected into a training set by a certain classifier. If a sample point has been accurately classified, then in the next training set, the probability that it is selected is reduced; instead, the weight of this sample point is increased. In this way, the Adaboost algorithm can "focus" on samples where the information points are rich and difficult to distinguish.
In the above embodiment, the step of optimizing the BP neural network by using a genetic algorithm to obtain the GA-BP neural network comprises: and initializing the weight value and the threshold value of the BP neural network.
In this embodiment, the step of optimizing the BP neural network by using a genetic algorithm to obtain the GA-BP neural network includes: and initializing the weight and the threshold of the BP neural network so as to optimize the weight and the threshold of the BP neural network by utilizing the optimizing capability of a genetic algorithm.
In the above embodiment, the step of combining the GA-BP neural network with the iterative algorithm to obtain the plurality of GA-BP neural network weak classifiers according to which the strong classifier model is composed comprises: preprocessing the original data including data and quantization normalization, selecting m groups of training data from a sample space, initializing the distribution weight of the test data, and distributing the weight D of the test data t (i)＝1/m。
In this embodiment, the step of combining the GA-BP neural network with the iterative algorithm to obtain the plurality of GA-BP neural network weak classifiers and forming the strong classifier model according to the plurality of GA-BP neural network weak classifiers includes: preprocessing the original data including data and quantization normalization, selecting m groups of training data from a sample space, initializing the distribution weight of the test data to preprocess the original data, and testing the distribution weight D of the test data t (i)＝1/m。
An embodiment of a second aspect of the present invention provides a fault diagnosis device for a cutting part of a heading machine, including: the acquisition module is used for acquiring a data sample of the cutting part; the first construction module is used for constructing a characteristic sample set based on the data sample; the second construction module is used for constructing a BP neural network according to the characteristic sample set; the neural network optimization module is used for optimizing the BP neural network by adopting a genetic algorithm to obtain a GA-BP neural network; the strong classifier module is used for combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers; and the fault diagnosis module is used for diagnosing faults of the cutting part by utilizing the strong classifier model.
The fault diagnosis device for the cutting part of the heading machine comprises an acquisition module, a first construction module, a second construction module, a neural network optimization module, a strong classifier module and a fault diagnosis module. The acquisition module can acquire data samples of the cutting part. The first construction module can construct a feature sample set based on the data samples. The second construction module can construct the BP neural network according to the characteristic sample set. The neural network optimization module can optimize the BP neural network by adopting a genetic algorithm to obtain the GA-BP neural network. The strong classifier module can combine the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and a strong classifier model is formed according to the plurality of GA-BP neural network weak classifiers. The fault diagnosis module can perform fault diagnosis on the cutting part by using the strong classifier model. The strong classifier model integrates the characteristics of Genetic Algorithm (GA) global optimization and BP neural network local optimization, and the strong classifier of the iterative algorithm gives different weights to a plurality of prediction sequences of the weak classifier, integrates the advantages of the different sequences, achieves the aim of 'double optimization' of the BP neural network, and greatly improves the prediction precision. Compared with the fault diagnosis method for the cutting part of the heading machine based on the multi-sensor information BP neural network in the related art, the fault diagnosis method for the cutting part of the heading machine can effectively monitor and diagnose the health state of the cutting part of the heading machine, and is high in diagnosis precision and accuracy.
An embodiment of the third aspect of the present invention provides a fault diagnosis device for a cutting part of a heading machine, including: a memory storing a program or instructions which, when executed by the processor, implement the steps of the method of fault diagnosis of a cutting section of a heading machine in any one of the embodiments of the first aspect.
The fault diagnosis device for the cutting part of the heading machine provided by the invention comprises a memory and a processor, wherein the memory stores programs or instructions, and the programs or instructions are executed by the processor to realize the steps of the fault diagnosis method for the cutting part of the heading machine in any embodiment of the first aspect. Therefore, the fault diagnosis device for the cutting part of the heading machine provided by the invention also has all the beneficial effects of the fault diagnosis method for the cutting part of the heading machine provided by any embodiment of the first aspect, and details are not repeated herein.
Embodiments of a fourth aspect of the invention provide a readable storage medium having stored thereon a program or instructions which, when executed, carry out the steps of the method of diagnosing a fault of a cutting section of a heading machine of an embodiment of the first aspect.
According to the present invention, there is provided a readable storage medium as it is a step of implementing the method of diagnosing a failure of a cutting section of a heading machine according to any one of the embodiments of the first aspect. Therefore, the readable storage medium provided by the invention also has all the beneficial effects of the fault diagnosis method for the cutting part of the heading machine provided by any embodiment of the first aspect, and the details are not repeated herein.
An embodiment of a fifth aspect of the present invention provides a heading machine, including: the fault diagnosis device of the cutting part of the heading machine of the second aspect embodiment or the third aspect embodiment, and/or the readable storage medium of the fourth aspect embodiment.
According to the invention, a heading machine is provided, which comprises the fault diagnosis device of the cutting part of the heading machine of the second aspect embodiment or the third aspect embodiment, and/or the readable storage medium of the fourth aspect embodiment. Therefore, the heading machine provided by the present invention further has the whole beneficial effects of the fault diagnosis device of the cutting part of the heading machine of the second aspect embodiment or the third aspect embodiment, and/or the readable storage medium of the fourth aspect embodiment, which are not described herein again.
The BP neural network, genetic Algorithm (GA), GA-BP neural network, and Adaboost Algorithm are explained in further detail below:
the BP neural network is a multilayer forward network based on a back propagation algorithm, is trained according to error reverse propagation, and has the structural form of single input, multiple hiding and single output. The processing units of each layer only receive the signals output by the processing units of the previous layer, process the received signals and input the processed signals to the next layer, and the processing units of the same layer are mutually independent. The BP algorithm has good nonlinear mapping capability, fault tolerance and generalization capability, and has the defect of easy falling into local minimum. Specifically, the local minimum problem is that after the BP network is trained for a certain number of times, although the error still exists, the network error is reduced at a slow speed and even does not change any more. The BP algorithm is essentially an algorithm which takes the error square sum of a BP neural network as an objective function and solves the objective function according to a gradient descent method so as to reach the minimum value. The BP network is restricted in application due to the structural characteristics of the BP network, so that the defect is solved by optimizing the weight and the threshold of the BP network through a genetic algorithm.
Genetic Algorithm (GA), which is a global optimization probability search technology based on genetic inheritance and natural evolution principle. The method has the advantages of global convergence, calculation parallelism and the like, is wide in applicability and needs little prior knowledge. The algorithm calculates the fitness of each sample according to the selected fitness function, encodes an object to be optimized, and screens individuals through fitness function calculation and intersection and variation in genetic operation, so that individuals in a new population are better, population samples are continuously evolved, the fitness value of the individuals is gradually improved until set control conditions are met, and finally, a global optimal solution or a suboptimal solution is obtained.
The GA-BP neural network optimizes the BP neural network by adopting a genetic algorithm, codes the weight and the threshold of the BP neural network to generate an initial population, uses a fitting error cost function as a fitness function, selects the optimal chromosome as the weight and the threshold of the BP neural network through genetic operation, then gives the threshold and the weight vector to the BP network, and obtains an approximate optimal value by using local search capability. The GA-BP genetic neural network combines the local optimization of the BP neural network with the global optimization characteristic of the genetic algorithm, ensures that the BP neural network converges to the global optimization and can obtain a more accurate optimization result, and theoretically avoids the defects that the BP is easy to fall into a local minimum value, the network convergence is slow and the stability is poor.
The Adaboost algorithm is an ensemble learning method. The method can obviously improve the performance of a learning algorithm, obtain the weight distribution of each learning sample, provide equal initial distribution weights, and continuously adjust the weights of the samples in the training process. The method trains a plurality of different weak classifiers or weak classifiers to the same training set, weights are given to each sample through the performance in the training process, and the weighted samples are transmitted downwards so as to strengthen the training degree of the samples with large errors in the training of the next weak classifier or weak classifier, and finally a strong classifier or a strong classifier is assembled by using a linear combination method, thereby realizing the great improvement of the classification and prediction accuracy and enabling the modeling result to be more stable until a certain preset error rate which is small enough is reached. In the training process using the Adaboost algorithm, each training sample is assigned a weight, and the weight can be regarded as the probability that the training sample is selected into the training set by a certain classifier. If a sample point has been accurately classified, then in the next training set, the probability that it is selected is reduced; instead, the weight of this sample point is increased. In this way, the Adaboost algorithm can "focus" on samples where the information points are rich and difficult to distinguish. Although the Adaboost algorithm can improve the diagnostic effect of the weak classifier, in more practical studies, the effect achieved by using the Adaboost algorithm alone is not ideal.
As shown in fig. 2, in order to effectively identify the fault of the cutting part of the heading machine as much as possible, firstly, a fault signal of the cutting part of the heading machine is processed, a factor analysis method is combined to select a characteristic sample set, a BP neural network optimized by a genetic algorithm is used as a basic classifier, a GA-BP neural network and an Adaboost algorithm are effectively combined according to the enhancement capability of the Adaboost algorithm, GA-BP neural network prediction samples are repeatedly trained to be output, the weight and the threshold of the BP neural network are optimized by utilizing the optimization capability of the genetic algorithm, and a GA-BP-Adaboost strong classifier consisting of a plurality of GA-BP neural network weak classifiers is obtained by combining the Adaboost algorithm, so that the diagnosis of the fault of the cutting part of the heading machine by the strong classifier is realized.
Design of a GA-BP-Adaboost prediction model:
the Adaboost method is a process of combining the outputs of multiple weak classifiers to produce an efficient system, and is also the most representative algorithm in the Boosting family. The main idea is as follows: first give weakLearning algorithm and sample space (x) i ，y i ) And selecting m groups of training data from the sample space, wherein the weight of each group of data is 1/m. Performing iterative operation for T times by using a weak learning algorithm, updating the weight distribution of the training data according to the prediction result after each iterative operation, and giving a larger weight to the training samples with low prediction precision, so that the training samples are more concerned in the next iterative operation, thereby obtaining a prediction function sequence f 1 ，f 2 ，…，f T And giving each prediction function f i And giving a weight to the function with better prediction effect, wherein the corresponding weight is larger. After T iterations, the final strong classifier function T is weighted by the weak classifier function. The GA-BP-Adaboost prediction model is that a GA-BP neural network is used as a weak classifier, GA-BP neural network prediction samples are repeatedly trained to be output, and a strong classifier consisting of a plurality of GA-BP neural network weak classifiers is obtained through an Adaboost algorithm.
The steps of the GA-BP-Adaboost strong prediction model are as follows:
the method comprises the following steps: data preprocessing and network initialization. The raw data is pre-processed including data and quantization normalization. Selecting m groups of training data from a sample space, and initializing distribution weight D of test data t (i) And =1/m. The number of nodes of the hidden layer is set to be 2q +1 by adopting Kolmogorov (Kelmogorov) theorem, wherein q is the number of nodes of the input layer.
Step two: the genetic algorithm optimizes the BP neural network. And (3) taking the threshold and the weight of the BP neural network as chromosomes of a genetic algorithm, training the BP neural network by using training data by using a fitness function, and taking the sum of the prediction errors as an individual fitness value.
Step three: find weak classifier g t (T =1,2,3 \ 8230;, T). When the t-th weak classifier is trained, the GA-BP genetic neural network is trained by using training data and the predicted training data is output, so that the error sum e of the predicted classification result g (t) is obtained t ：
g(t)≠y
Wherein: y is the expected classification result; g (t) is the predicted classification result, D i (i) For weight distribution of training data, e t Is a weak classifier with weight distribution D i The error of (2).
Step four: and calculating the weight of the prediction classification result. Prediction error sum e from prediction classification result g (t) t Calculating the weight a of the predicted classification result t 。
Step five: the weights are updated. The weight of the training data of the next round is according to the weight a of the prediction classification result t Adjusting;
in the formula: b is t Is a normalization factor in order to make the sum of the distribution weights 1. Namely, it is
Step six: a strong classifier function. Obtaining T weak prediction functions g after training T rounds t (x) (T =1,2,3 \ 8230;, T), and then weighted by T weak classifier functions, a strong classifier function H (x) is obtained:
data selection and pretreatment:
the vibration acceleration sensor is installed at a plurality of measuring points of the horizontal rotating mechanism of the cutting part of the intelligent heading machine, the data acquisition box is installed on the machine body of the heading machine, and vibration signals are transmitted back to the ground dispatching room in real time. The vibration signal time domain characteristic value of the vibration acceleration sensor can be divided into dimensional indexes and dimensionless indexes, and the dimensional indexes comprise a maximum value, a peak-to-peak value, a mean value, a variance and an effective value; the dimensionless indexes include kurtosis, skewness, form factor and peak factor. The GA-BP-Adaboost algorithm needs to select input variables which are small in correlation and sensitive to faults, so that an effective value of vibration energy of the cutting part is reflected, kurtosis representing the impact degree of the cutting part and a standard deviation representing the deviation degree of vibration signals from a mean value can be used for fault diagnosis of the cutting part of the heading machine.
If the vibration energy is changed violently under the normal condition, the mechanical fault of the cutting part of the heading machine can be inferred. Vibration energy changes are used as criteria to be introduced into fault diagnosis of a cutting part of the heading machine, vibration signals are stable under the normal working condition of the heading machine, and the vibration energy, the received impact and the degree of deviation of vibration from a mean value under the fault condition are much larger than those under the normal condition. Therefore, in order to verify the effectiveness and the precision of the algorithm model, the effective value, the variance and the kurtosis of a vibration acceleration sensor signal arranged on the heading machine body, the temperature of a cutting motor and the three-phase current value (A) are taken as examples for carrying out data analysis.
The method comprises the steps of performing dimensionality reduction on original data by using principal component analysis, extracting a principal component factor, and normalizing original vibration training data of a cutting part of the heading machine by using a Mapminmax function in Matlab (matrix laboratory) for facilitating convergence of a GA-BP (genetic algorithm-back propagation) neural network due to the fact that the data subjected to principal component analysis processing have large difference. And selecting a cutting part vibration acceleration characteristic value and a cutting motor three-phase current value as samples when the heading machine is in no-load, and training and testing by adopting a GA-BP-Adaboost algorithm.
Description of the experiments:
in order to verify the prediction accuracy and precision of the GA-BP-Adaboost algorithm, analysis is performed through two schemes: the method adopts the first 1900 groups of data collected by a data collecting box of the heading machine as a sample for training and the last 100 groups as a test sample. For convenience of comparison, each scheme adopts a BP neural network standard model, and a Sigmoid function is adopted between the hidden layer and the input layer. The BP network structure parameters of each scheme are specifically as follows: the number of nodes of an input layer is 2, the number of nodes of an output layer is 1, the number of nodes of an implicit layer is 10, the momentum factor and the learning rate are 0.01, the minimum performance gradient is 0.000001, the maximum failure frequency is 6, the display frequency is 0.025, and the maximum cycle frequency is set to be 20. The genetic algorithm population range is set to 10, the mutation probability and the cross probability are respectively 0.2 and 0.8, and the maximum iteration times are both 30. The prediction accuracy of the model is evaluated by adopting the average absolute value, the mean square error, the root mean square error and the average percentage error, and the method specifically comprises the following steps:
1) Mean Absolute Error (MAE):
2) Mean Square Error (MSE):
3) Root Mean Square Error (RMSE):
4) Mean Absolute Percentage Error (MAPE):
in the formula: y is t In order to be the actual value of the measurement,
for the prediction value, n is the total number of samples, and t is the test number. And inputting the original data into each model for training to obtain a predicted output value.
As can be seen from analysis, as shown in FIG. 4, each index of the standard BP neural network prediction model is larger than that of the scheme II, which shows that the GA-BP neural network model has a great improvement in the fault prediction accuracy of the cutting part of the heading machine compared with the standard BP neural network, as shown in FIG. 5, the predicted values and the actual values of the current mean values of the cutting parts of the models are compared, so that the first scheme and the second scheme have basically consistent trends and have small differences between the predicted values in the front section and the middle section of the prediction section of the cutting part of the heading machine, and the two schemes have deviations in the rear section. And for the prediction precision of each model, the second scheme is smaller than the first scheme in terms of mean square error index performance, which shows that the GA-BP has better classification performance and high accuracy identification. And in the aspect of the root mean square error and the average absolute error index, the precision of the scheme two is slightly lower, but the precision of the scheme two is almost the same in the aspect of the average absolute percentage error, and the standard BP neural network is slightly higher than that of the GA-BP genetic neural network model, so that the defects that the processing running time of the GA-BP neural network prediction model is long, the precision is improved limitedly, the time complexity is changed more quickly and the like are overcome.
The GA-BP-Adaboost algorithm is constructed and tested:
the GA-BP and BP neural networks are compared to study the algorithm, and then the diagnosis effect of the GA-BP-Adaboost algorithm is studied. For the prediction accuracy of each model, the indexes of mean square error, mean absolute error and mean absolute percentage error of the GA-BP-Adaboost strong prediction model are minimum in 3 models, and are respectively 0.0033, 0.04804 and 3.31%. This shows that the average absolute percentage error based on the BP neural network is very unstable, and the prediction effect of the strong classifier based on the GA-BP-Adaboost algorithm is more stable and relatively small. Compared with the GA-BP model optimized by the genetic algorithm, the relative value of the mean absolute percentage error of the two is reduced from 4.24 percent to 3.31 percent, and the fluctuation of the GA-BP and the BP is obvious in the repeated diagnosis process. While the volatility of the BP and GABP optimized by Adaboost is smaller. In contrast, GA-BP-Adaboost has better repeated diagnosis effect and weaker fluctuation.
By analyzing and obviously improving the prediction results of the GA-BP and standard BP neural networks and adopting a heading machine cutting part fault prediction model established by a GA-BP-Adaboost algorithm, the global prediction precision can be improved, and the performance is stable. The results are contrastively analyzed as shown in fig. 6, the generalization capability of the model is enhanced by the aid of a strong classifier formed by weighting and combining a GA-BP base predictor through an Adaboost algorithm, the Adaboost strong classifier adjusts the weight among a plurality of groups of weak classifiers according to prediction errors, different prediction results of a BP neural network after random selection, crossing and mutation optimization of a genetic algorithm can be integrated, obviously, a model prediction value based on the GA-BP-Adaboost algorithm is consistent with the actual detection result of a cutting part of the heading machine on the whole variation trend, and the difference value between the prediction value and the actual detection result is smaller than that of the GA-BP and BP network models, so that the goal of 'best-out selection' of the Adaboost strong classifier is realized, and finally the average absolute percentage error of all prediction samples is controlled within 4%, which shows that the GA-BP-Adaboost model adopted in the application can meet the requirements of errors and application through training and testing. The model prediction accuracy is improved to the maximum extent, and the superiority and feasibility of the BP-Adaboost strong prediction model optimized based on the genetic algorithm in the fault diagnosis of the cutting part of the heading machine are verified.
The training process of the GA-BP-Adaboost algorithm is as follows, as shown in FIG. 3:
s1, firstly, determining a training and testing sample, and constructing a BP neural network of a proper model. The data sample is obtained by installing the data sample in a data acquisition box of the mining monitoring device of the cutting part of the heading machine.
And S2, initializing the weight and the threshold of the BP neural network, and training the BP neural network until the training is finished.
And S3, outputting the detection result of the BP neural network to a GA algorithm as an initial condition for calculating the fitness value.
And S4, initializing the population, taking the initialization result and the BP neural network test result as input conditions, calculating a fitness value, and then performing operations such as selection, intersection, mutation and the like until the termination condition is met.
And S5, initializing the weight and the iteration frequency of the training sample, training the GA-BP neural network by using the training sample and the GA-BP output result, calculating the error sum and the sequence weight of the GA-BP neural network training sample, updating the weight of the training sample in the next round, judging whether the error is lower than a threshold value or the iteration frequency is met, and finishing the training if the condition is met.
And S6, repeating the steps S1-S4, and realizing the training process of the GA-BP-Adaboost algorithm.
In the description of the present specification, the terms "connect", "mount", "fix", and the like are to be understood in a broad sense, for example, "connect" may be a fixed connection, a detachable connection, or an integral connection; may be directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
In the description of the present specification, it is to be understood that the terms "upper", "lower", and the like indicate orientations or positional relationships based on orientations or positional relationships shown in the drawings, which are merely for convenience of describing the present invention and simplifying the description, but do not indicate or imply that the referred device or unit must have a specific direction, be configured and operated in a specific orientation, and thus, should not be construed as limiting the present invention.
In the description of the present specification, the description of "one embodiment," "some embodiments," "specific embodiments," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A fault diagnosis method for a cutting part of a heading machine is characterized by comprising the following steps:
acquiring a data sample of the cutting part;
constructing a feature sample set based on the data samples;
constructing a BP neural network according to the characteristic sample set;
optimizing the BP neural network by adopting a genetic algorithm to obtain a GA-BP neural network;
combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers;
and utilizing the strong classifier model to carry out fault diagnosis on the cutting part.
2. The method according to claim 1, wherein the step of acquiring the data samples of the cutting unit specifically includes:
obtaining an effective value of the vibration energy of the cutting part;
obtaining the kurtosis of the impact degree of the cutting part;
and acquiring the standard deviation of the cutting part representing the degree of deviation of the vibration signal from the mean value.
3. The method of diagnosing a fault in a cutting section of a heading machine according to claim 1, wherein the step of constructing a feature sample set based on the data samples is preceded by:
and processing the data samples, and selecting the characteristic sample set by combining a factor analysis method.
4. The method for diagnosing the fault of the cutting part of the heading machine according to claim 1, wherein the iterative algorithm comprises an Adaboost algorithm, the GA-BP neural network is combined with the Adaboost algorithm to obtain a plurality of GA-BP neural network weak classifiers, and the step of forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers comprises the following steps:
searching a plurality of weak classifiers, training the GA-BP neural network by using training data and predicting the output of the training data to obtain the error sum of the prediction classification result;
calculating the weight of the prediction classification result according to the prediction error sum of the prediction classification result;
adjusting the weight of the training data of the next round according to the weight of the prediction classification result;
and weighting and combining the weak classifiers after the training to obtain a strong classifier model.
5. The method of claim 1, wherein the step of optimizing the BP neural network using a genetic algorithm to obtain a GA-BP neural network comprises:
and initializing the weight value and the threshold value of the BP neural network.
6. The method according to any one of claims 1 to 5, wherein the step of combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers and composing a strong classifier model according to the plurality of GA-BP neural network weak classifiers comprises:
the raw data is pre-processed including data and quantization normalization.
7. A failure diagnosis device of a cutting part of a heading machine is characterized by comprising:
the acquisition module is used for acquiring a data sample of the cutting part;
a first construction module for constructing a feature sample set based on the data samples;
the second construction module is used for constructing a BP neural network according to the characteristic sample set;
the neural network optimization module is used for optimizing the BP neural network by adopting a genetic algorithm to obtain a GA-BP neural network;
the strong classifier module is used for combining the GA-BP neural network with an iterative algorithm to obtain a plurality of GA-BP neural network weak classifiers, and forming a strong classifier model according to the plurality of GA-BP neural network weak classifiers;
and the fault diagnosis module is used for diagnosing the fault of the cutting part by utilizing the strong classifier model.
8. A failure diagnosis device of a cutting part of a heading machine is characterized by comprising:
a memory and a processor, the memory storing a program or instructions which when executed by the processor implement the steps of the method of diagnosing a fault of a cutting section of a heading machine as claimed in any one of claims 1 to 6.
9. A readable storage medium, characterized in that a program or instructions are stored thereon which, when executed, carry out the steps of the method of diagnosing a fault of a cutting section of a heading machine according to any one of claims 1 to 6.
10. A heading machine, comprising:
the failure diagnosis device of the cutting section of the heading machine according to claim 7 or 8; and/or
The readable storage medium of claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210919624.8A CN115392349A (en) | 2022-08-02 | 2022-08-02 | Fault diagnosis method and device for cutting part of heading machine and heading machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210919624.8A CN115392349A (en) | 2022-08-02 | 2022-08-02 | Fault diagnosis method and device for cutting part of heading machine and heading machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115392349A true CN115392349A (en) | 2022-11-25 |
Family
ID=84119527
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210919624.8A Pending CN115392349A (en) | 2022-08-02 | 2022-08-02 | Fault diagnosis method and device for cutting part of heading machine and heading machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115392349A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115628930A (en) * | 2022-12-16 | 2023-01-20 | 太原理工大学 | Method for predicting underground cutting working condition of heading machine based on RBF neural network |
| CN117541847A (en) * | 2023-10-16 | 2024-02-09 | 陕西小保当矿业有限公司 | Detection method and system for tunneling environment |
| WO2024124672A1 (en) * | 2022-12-13 | 2024-06-20 | 中铁工程装备集团有限公司 | Fault diagnosis method for tunnel boring device and intelligent tunnel boring device |
-
2022
- 2022-08-02 CN CN202210919624.8A patent/CN115392349A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024124672A1 (en) * | 2022-12-13 | 2024-06-20 | 中铁工程装备集团有限公司 | Fault diagnosis method for tunnel boring device and intelligent tunnel boring device |
| CN115628930A (en) * | 2022-12-16 | 2023-01-20 | 太原理工大学 | Method for predicting underground cutting working condition of heading machine based on RBF neural network |
| CN117541847A (en) * | 2023-10-16 | 2024-02-09 | 陕西小保当矿业有限公司 | Detection method and system for tunneling environment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115392349A (en) | Fault diagnosis method and device for cutting part of heading machine and heading machine | |
| CN112784881B (en) | Network abnormal flow detection method, model and system | |
| CN107992976B (en) | Hot topic early development trend prediction system and prediction method | |
| KR101967978B1 (en) | Apparatus for predicting net penetration rate of shield tunnel boring machine and method thereof | |
| CN109670714B (en) | Ship gas turbine comprehensive state evaluation method based on membership degree analysis | |
| CN114676742A (en) | Power grid abnormal electricity utilization detection method based on attention mechanism and residual error network | |
| CN103390154A (en) | Face recognition method based on extraction of multiple evolution features | |
| CN116341901B (en) | Integrated evaluation method for landslide surface domain-monomer hazard early warning | |
| CN115983087A (en) | Method for detecting time sequence data abnormity by combining attention mechanism and LSTM and terminal | |
| CN115618732A (en) | Nuclear reactor digital twin key parameter autonomous optimization data inversion method | |
| CN114819777A (en) | Enterprise sales business analysis and management system based on digital twin technology | |
| Liu et al. | Residual useful life prognosis of equipment based on modified hidden semi-Markov model with a co-evolutional optimization method | |
| CN111582634B (en) | Multi-factor safety grading method and system for underground large-space construction | |
| CN116029220B (en) | Voltage transformer operation error assessment method, system, equipment and medium | |
| CN112926016A (en) | Multivariable time series change point detection method | |
| CN117271979A (en) | Deep learning-based equatorial Indian ocean surface ocean current velocity prediction method | |
| CN117477544A (en) | LSTM ultra-short-term photovoltaic power prediction method and system integrating time mode features | |
| CN117171702A (en) | Multi-mode power grid fault detection method and system based on deep learning | |
| CN117272202A (en) | Dam deformation abnormal value identification method and system | |
| CN116188266A (en) | Spatial interpolation method for geological investigation based on fuzzy theory and ensemble learning | |
| CN115618743A (en) | State evaluation method and state evaluation system of sighting telescope system | |
| CN112184037B (en) | Multi-modal process fault detection method based on weighted SVDD | |
| CN114118249A (en) | Structure damage diagnosis method based on optimized stacked self-encoder and multi-signal fusion | |
| CN114880954A (en) | Landslide sensitivity evaluation method based on machine learning | |
| CN114266013A (en) | Deep learning virtual perception network-based transmission system vibration decoupling method |
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 |