CN118013375A - Knitting machine fault detection method based on gsinSOA-ELM model - Google Patents
Knitting machine fault detection method based on gsinSOA-ELM model Download PDFInfo
- Publication number
- CN118013375A CN118013375A CN202410410954.3A CN202410410954A CN118013375A CN 118013375 A CN118013375 A CN 118013375A CN 202410410954 A CN202410410954 A CN 202410410954A CN 118013375 A CN118013375 A CN 118013375A
- Authority
- CN
- China
- Prior art keywords
- knitting machine
- gsinsoa
- optimal
- model
- vibration
- 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.)
- Granted
Links
- 238000009940 knitting Methods 0.000 title claims abstract description 86
- 238000001514 detection method Methods 0.000 title claims abstract description 58
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 46
- 238000005457 optimization Methods 0.000 claims abstract description 37
- 238000013507 mapping Methods 0.000 claims abstract description 28
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 19
- 230000000694 effects Effects 0.000 claims abstract description 10
- 210000002569 neuron Anatomy 0.000 claims abstract description 10
- 238000001914 filtration Methods 0.000 claims description 44
- 238000000034 method Methods 0.000 claims description 35
- 239000002245 particle Substances 0.000 claims description 33
- 230000006870 function Effects 0.000 claims description 31
- 238000004364 calculation method Methods 0.000 claims description 23
- 230000008569 process Effects 0.000 claims description 22
- 230000005540 biological transmission Effects 0.000 claims description 21
- 241000272168 Laridae Species 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 12
- 230000009471 action Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 238000012216 screening Methods 0.000 claims description 7
- 230000006399 behavior Effects 0.000 claims description 6
- 230000002431 foraging effect Effects 0.000 claims description 6
- 230000005012 migration Effects 0.000 claims description 5
- 238000013508 migration Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 230000005484 gravity Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000004088 simulation Methods 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 abstract description 14
- 239000000306 component Substances 0.000 description 48
- 230000008901 benefit Effects 0.000 description 12
- 230000000739 chaotic effect Effects 0.000 description 10
- 230000000295 complement effect Effects 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000000737 periodic effect Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000000712 assembly Effects 0.000 description 1
- 238000000429 assembly Methods 0.000 description 1
- 239000003638 chemical reducing agent Substances 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000009189 diving Effects 0.000 description 1
- 230000019637 foraging behavior Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005461 lubrication Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The application relates to the technical field of fault detection, in particular to a knitting machine fault detection method based on gsinSOA-ELM model, which comprises the following steps: s1, acquiring a vibration data set of the knitting machine, which is acquired by a vibration sensor; s2, performing EMD (empirical mode decomposition), wiener Kalman denoising and data reconstruction on the vibration data set acquired in the S1 to acquire reconstructed data; s3, performing time domain feature extraction, frequency domain feature extraction and permutation entropy feature extraction on the reconstructed data obtained in the step S2, and rearranging the features into a data set; s4, establishing gsinSOA-ELM model, and optimizing the connection weight w and the neuron threshold b parameter in the model by using a seagull optimization algorithm based on gsin mapping to obtain an optimal gsinSOA-ELM model; s5, collecting real-time vibration data of the knitting machine, performing fault detection by using the optimal gsinSOA-ELM model obtained in the S4, and outputting a fault detection result of the knitting machine. The application can improve the recognition effect and the fault detection rate of the fault diagnosis of the knitting machine, and has high real-time performance and better stability of the fault diagnosis.
Description
Technical Field
The invention relates to the technical field of fault detection, in particular to a knitting machine fault detection method based on gsinSOA-ELM model.
Background
Knitting machines are very prone to failure during long-term use due to the load operation of the needle bed, needle cylinder, mechanical components and drive train (e.g. processes such as draw back, yarn laying, stitch bonding, etc. are accompanied by load operation of the mechanical components and drive train). Wherein: the needle bed and the needle cylinder are used as core components of the knitting machine and are responsible for knitting operation, and the needle bed and the needle cylinder are easy to wear, deform or break due to long-time work and stress conditions, so that the knitting machine cannot work normally. And the transmission system of the knitting machine comprises: and the mechanical structure is composed of a motor, a speed reducer, a belt, a chain and the like and is used for driving the needle bed and the needle cylinder to move. The transmission system is easy to cause faults due to abrasion, looseness or poor lubrication of parts and the like, and the normal operation of the knitting machine is affected. Any failure of the knitting machine drive system can lead to unnecessary shutdowns and affect the yield of the knitted product, and therefore, health monitoring and failure detection of the knitting machine are of great importance to the yield of the knitted product.
Currently, a vibration sensor is added into a knitting machine, the vibration condition of a transmission system of the knitting machine is monitored in real time through the vibration sensor, and operation state information (such as operation stability, abnormal sound, vibration amplitude and the like) of the knitting machine is obtained through analysis of a vibration signal. The operation data and the state of the knitting machine are analyzed through fault detection to determine whether the knitting machine is operating normally or not and in case of faults, the position and the type of the faults. Fault detection techniques are critical to improving operational reliability of knitting machines, reducing maintenance costs, and reducing downtime. If faults occur during the operation of the knitting machine, at least one system characteristic or variable changes slightly, and the trained fault detection model can capture the slight differences of working conditions, so that timely diagnosis can be performed. Knitting machines are composed of a large number of components, and signals are transmitted between the components, and are coupled and modulated with each other, so that the signals face various difficulties in the feature extraction process. In the process of data cleaning and fault feature extraction of knitting machine data acquired by the vibration sensor, a filtering function is used for filtering, and possible non-abnormal values outside a threshold value are directly removed, so that follow-up feature extraction is incomplete, an optimal fault detection model cannot be called out in the process of fault detection model training, and accuracy and stability of knitting machine fault detection can be affected.
Disclosure of Invention
The invention aims to solve the technical problems that: in order to solve the technical problems that the optimal fault detection model cannot be called out due to incomplete filtering in the existing fault detection of the knitting machine and the accuracy and stability of the fault detection of the knitting machine are affected, the invention provides a fault detection method of the knitting machine based on gsinSOA-ELM model, and the accuracy and stability of the fault detection of the knitting machine can be improved by optimizing the fault detection method.
The technical scheme adopted for solving the technical problems is as follows: a knitting machine fault detection method based on gsinSOA-ELM model comprises the following steps:
s1, acquiring a vibration data set of the knitting machine, which is acquired by a vibration sensor;
s2, performing EMD (empirical mode decomposition), wiener Kalman denoising and data reconstruction on the vibration data set obtained in the step S1 to obtain reconstructed data;
S3, performing time domain feature extraction, frequency domain feature extraction and permutation entropy feature extraction on the reconstruction data obtained in the step S2, and rearranging the features into a data set;
S4, establishing gsinSOA-ELM model, and optimizing the connection weight w and the neuron threshold b parameter in the model by using a seagull optimization algorithm based on gsin mapping to obtain an optimal gsinSOA-ELM model;
s5, collecting real-time vibration data of the knitting machine, performing fault detection by using the optimal gsinSOA-ELM model obtained in the step S4, and outputting a fault detection result of the knitting machine;
in step S4, the optimization process of the connection weight w and the neuron threshold b in the model by the gull optimization algorithm based on gsin mapping includes the following steps:
s4-1, initializing a seagull population through gsin mapping, and setting a model error rate as an fitness function;
the equation for initializing the gull population by gsin mapping is:
Wherein: m, l represent constants, i represent the current iteration number, X best represent the currently known optimal solution, item max represent the set maximum iteration number, X n is the current state, X n+1 is the next state, |lx best-Xn | represent the difference between the optimal position of the particle generated at the current time and the particle at the current time in the optimizing process, A square operator representing the ratio of the current iteration times to the iteration times;
S4-2, calculating the fitness function value of each sea gull individual;
S4-3, comparing fitness function values of individual seagulls, and searching a current optimal position;
S4-4, calculating an optimal fitness value of the prey, and updating the individual attack position;
s4-5, judging whether iteration times are reached, if not, repeating the steps S4-1 to S4-4 by taking the current optimal position as a search area; if yes, executing the step S4-6;
S4-6, obtaining the optimal connection weight w and the neuron threshold b, and further obtaining the optimal gsinSOA-ELM model.
Therefore, through EMD decomposition and wiener Kalman denoising of the vibration data of the knitting machine, the rapid filtering of noise in the full frequency range of the vibration signal can be realized, meanwhile, the diagnosis efficiency of a fault detection model can be accelerated, through introducing the arrangement entropy feature, the accurate extraction of the mechanism feature of the knitting machine can be realized, the position of particles in a batch-to-batch self-adaptive optimization initialization algorithm is carried out on the initial total group parameters of the sea gull in the SOA through gsin mapping, the space global dynamic search of the fault detection model parameters is realized, the situation that the fault detection model falls into the local optimal solution can be effectively reduced, the parameters in the algorithm are optimized, the optimizing speed in each cycle is accelerated, and the accuracy and the stability of the fault detection of the knitting machine can be improved.
The method has the advantages that the loop iteration is carried out on the gsinSOA-ELM model through a mew optimization algorithm based on gsin mapping, the search range is continuously reduced in the iteration process, and meanwhile random points of the search are increased in the reduced range, so that the global search capability of parameter optimization is improved, a better solution can be found in a solution space, and the complex optimization problem can be solved. Meanwhile, the algorithm is further randomized on the initial value of the seagull at gsinSOA, so that stronger global searching capability is achieved, and the algorithm has the advantages of strong self-adaption and high convergence speed.
By taking |lx best-Xn | as the difference between the optimal position of the particle generated at the current moment and the particle at the current moment in the process of algorithm optimization, the position and the direction of the initial seagull particle are more changeable when the difference is placed in the optimizing space and can be regarded as vector parameters pointing to the direction of the optimal solution. Meanwhile, in order for the algorithm to continue to generate particle small-range optimization near the optimal solution of the searched space along with iteration, the distance of the initial position for generating particles needs to be controllable by combining the square operator of the ratio of the current iteration times to the iteration timesRandom search probability is increased so as to realize nonlinear self-adaptive dynamic adjustment of initial particle positions along with the increase of iteration times when parameter optimization is performed in an algorithm, thereby improving fault diagnosis precision.
Further, in step S4, the global in-scope search is achieved for the gull optimization algorithm based on gsin mapping, including: position updating, action direction updating and avoiding the mutual collision of individuals;
the variable A is introduced to calculate the position C Z of the sea gull individual, and the calculation formula is as follows:
；
the variable B is introduced to be responsible for balancing global optimization and local optimization, and the expression of the action direction of the model is as follows:
；
Introducing a linear increment factor The calculation formula of B is:
;
In the simulation space, the individual seagull particles move according to the optimal direction, and after continuous iteration, the expression of the migration to the optimal position D Z（t）,DZ (t) is as follows:
；
The seagull foraging captures the prey foraging by spiral dive, and the behavior is simulated into a three-dimensional space, so that the following motion trail formula is obtained:
；
the local search formula for the seagull individuals is as follows:
；
Wherein: f c is the control factor of the total sea-gull group, t is the current iteration number, max iter is the maximum iteration number, P Z (t) is the current sea-gull position, M Z (t) is the calculated optimal individual action direction, P Z (t) is the optimal individual sea-gull position, r d is the random number in [0,1], r represents the spiral radius, and is controlled by constant u and constant v, Is a random number of (a) in the memory.
Further, in step S4, establishing gsinSOA-ELM models includes: and extracting the characteristics of the same property from the vibration data reconstructed by the knitting machine under different working conditions to form characteristic vectors, dividing the characteristic vectors into training test sets, putting the training sets into different classifiers, and evaluating the characteristics and the optimal classifier according to classification results obtained by the classifiers.
Further, in step S4, in the gsinSOA-ELM model, the calculation formula of the output matrix H (x) of the activation function g (w ixi+bi) is:
；
The actual output f L (x) of the model is defined, and the calculation formula of the actual output f L (x) is as follows:
；
to ensure that the model output error is minimal, solving equation e (x) yields:
；
Subtracting the output f L (x) of the network from the sample label T to obtain a minimum norm solution as an objective function e (x), obtaining H (x) β=t, solving the equation to equal order, searching the minimum two-dimensional solution of the linear equation H (x) β=t, and deriving the obtained optimal solution:
；
Wherein: w represents the connection weight (weight) of the input layer to the hidden layer, b represents the bias of the hidden layer, Representing an activation function (activation function)/>Using the sigmoid function, x represents the input data (input),/>G (w ixi+bi) is an activation function, β is the weight of the output layer, and H is the hidden layer output matrix H (x) for the optimal parameters of the output layer.
Further, the step S2 includes the steps of:
S2-1, performing EMD (empirical mode decomposition) on the vibration data set obtained in the step S1, comparing the decomposed components with pearson, signal to noise ratio and root mean square difference of original vibration data to represent the noise content of the vibration data, and performing post-wiener Kalman denoising treatment;
S2-2, fitting a maximum value and a minimum value of vibration data x (t) through a spline interpolation method, obtaining an upper envelope line and a lower envelope line through cubic spline fitting, and calculating to obtain a mean value m 1 (t) of the envelope lines;
S2-3, judging whether the first-order modal component h 1 (t) meets the condition of the modal component, if yes, the first-order modal component is h 1 (t), and if not, repeating the step S2-2 for k times until Satisfying the condition/>As a first-order modal component and denoted as c 1 (t), after the decomposition condition is satisfied, the vibration signal x (t) is decomposed into a plurality of modal components c j (t) and a residual signal r n (t); wherein, h 1（t）=x（t）-m1（t）;h1 (t) meets the judging condition of the modal component as follows: the number of maximum points of the vibration signal x (t) differs from the number of zero crossing points by not more than 1; the mean value of the upper envelope curve and the lower envelope curve of vibration data x (t) is constant at 0;
S2-4, denoising the vibration signal component with high noise by utilizing wiener Kalman denoising: comparing each modal component with various coefficients of the vibration signal x (t), and screening the first-order modal components according to a screening principle of a correlation coefficient r i and a correlation coefficient r i;
S2-5, if the correlation coefficient r i is smaller than rho, denoising processing is needed for the modal component, and if the correlation coefficient r i is larger than rho, the relevance between the modal component and the vibration signal x (t) is high, and characteristic information of the vibration signal x (t) is reserved, so that the modal component is reserved;
The calculation formula of the correlation coefficient r i is as follows:
；
The calculation formula of ρ is:
；
x i is the component of the mode shape, Is the mean value of the modal components, y is the vibration signal,/>The maximum correlation coefficient r i is marked as max Correlation coefficient for the mean value of the vibration signal and K is the number of layers of modal decomposition;
S2-6, adding the mode component after the wiener Kalman denoising, the residual mode component and the residual error at the same time, and reconstructing a pure reconstruction signal.
Therefore, EMD is rapidly decomposed, modal components can be decomposed in a very short time, and the real-time requirement of the follow-up fault detection of the model is met; because the knitting machine transmission system works at a high speed, a vibration sensor installed on the knitting machine transmission system collects true signals and simultaneously contains noise signals and knitting machine working characteristic information, a redundant periodic stable signal part in the knitting machine vibration signals can have a good filtering effect through wiener filtering, but is not suitable for a nonlinear system, accurate noise information statistics is needed, kalman filtering is suitable for a dynamic system, the condition of dynamic change of the system can be effectively processed, the Kalman filtering is self-adaptive and strong, filtering parameters can be dynamically adjusted according to the actual condition of the system, advantages of the two are complementary, the advantages are complementary through a combined mode, the wiener filtering is mainly used for denoising and filtering of the signals, the authenticity of the signals is estimated through minimizing mean square errors, the Kalman filtering is used for state estimation and system dynamic modeling, and optimal state estimation is provided through combining a system dynamic model and observation data.
Further, in step 2-3, the calculation formula of the low frequency residual signal r 1 (t) is:
；
The calculation formula of the low frequency residual signal rj (t) is:
；
the vibration signal x (t) decomposition expression is:
。
further, the step S2-4 comprises the following steps:
S2-4-1, denoising and preprocessing the screened modal components through wiener filtering to improve signal quality and observability;
s2-4-2, establishing a dynamic model of a vibration signal of a transmission system of the knitting machine, comprising the following steps: state equations and observation equations to describe the evolution and observation processes of the drive train;
S2-2-3, estimating and predicting the state of the transmission system through Kalman filtering, and combining a system dynamic model and observation data to obtain optimal state estimation;
s2-4-4, processing the output of the Kalman filtering through the wiener filtering again to improve estimation accuracy and filtering effect.
Therefore, the wiener filtering and the Kalman filtering are combined, better effects can be obtained in the vibration data processing and system control of the knitting machine, the method is particularly suitable for denoising of the complex transmission system of the knitting machine and the cooperative work among parts, and the complex situations of denoising of the vibration data and real-time state estimation of the transmission system can be considered at the same time.
Further, in step S3, the time domain features include: maximum, minimum, peak-to-peak, mean, average amplitude, square root amplitude, variance, standard deviation, root mean square value, kurtosis, skewness, waveform factor, peak factor, pulse factor, margin factor, and clearance factor;
the frequency domain features include: center of gravity frequency, mean square frequency, frequency variance, and root mean square frequency;
the expression of the permutation entropy H PE (i) of the ith modal component after EMD decomposition is:
；
wherein: p j represents the probability of occurrence of the j-th symbol sequence of the modal component after phase space reconstruction, and N represents the length of the time sequence, which is used to represent the number of data points involved in the permutation. Therefore, the permutation entropy can reflect the dynamic characteristics and nonlinear characteristics of signal wave bands in the aspect of dynamic mutation detection, can amplify and extract the dynamic characteristics hidden in the vibration signals of the knitting machine well, and achieves better training of the follow-up fault diagnosis model on the characteristics of the knitting machine.
Compared with the prior art, the invention has the beneficial effects that:
The method has the advantages that EMD decomposition and wiener Kalman denoising are carried out on the vibration data of the knitting machine, so that the rapid filtering of noise in the full frequency range of the vibration signal can be realized, meanwhile, the diagnosis efficiency of a fault detection model can be accelerated, the accurate extraction of the mechanism characteristics of the knitting machine can be realized by introducing the arrangement entropy characteristics, the positions of particles in a batch-to-batch self-adaptive optimization initialization algorithm are carried out on the initial total group parameters of the sea gull in the SOA through gsin mapping, the space global dynamic search on the parameters of the fault detection model is realized, the condition that the fault detection model falls into the local optimal solution can be effectively reduced, the parameter B value in the algorithm is optimized, the optimizing speed in each cycle is accelerated, and the accuracy and the stability of the fault detection of the knitting machine can be improved.
Drawings
The invention will be further described with reference to the drawings and examples.
FIG. 1 is a schematic flow chart of a knitting machine fault detection method based on gsinSOA-ELM model of the present invention;
FIG. 2 is a training schematic diagram of an ELM model;
FIG. 3 is a schematic flow chart of gsinSOA model of the present invention.
Detailed Description
The invention will now be described in further detail with reference to the accompanying drawings. The drawings are simplified schematic representations which merely illustrate the basic structure of the invention and therefore show only the structures which are relevant to the invention.
In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the device or element being referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, features defining "first", "second" may include one or more such features, either explicitly or implicitly. In the description of the present invention, unless otherwise indicated, the meaning of "a plurality" is two or more.
In the description of the present invention, it should be noted that, unless explicitly specified and limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be either fixedly connected, detachably connected, or integrally connected, for example; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
As shown in fig. 1 to 3, which are preferred embodiments of the present invention, the method for detecting a knitting machine fault based on gsinSOA-ELM model of the present embodiment includes the steps of:
s1, acquiring a data set of vibration data x (t) of the knitting machine, which is acquired by a vibration sensor;
s2, performing EMD (empirical mode decomposition), wiener Kalman denoising and data reconstruction on the vibration data set obtained in the step S1 to obtain reconstructed data;
S3, performing time domain feature extraction, frequency domain feature extraction and permutation entropy feature extraction on the reconstruction data obtained in the step S2, and rearranging the features into a data set;
S4, establishing gsinSOA-ELM model, and optimizing the connection weight w and the neuron threshold b parameter in the model by using a seagull optimization algorithm based on gsin mapping to obtain an optimal gsinSOA-ELM model;
S5, collecting real-time vibration data x (t) of the knitting machine, performing fault detection by using the optimal gsinSOA-ELM model obtained in the step S4, and outputting a fault detection result of the knitting machine, wherein the real-time vibration data x (t) is a time sequence signal.
Therefore, through EMD decomposition and wiener Kalman denoising of the vibration data of the knitting machine, the rapid filtering of noise in the full frequency range of the vibration signal can be realized, meanwhile, the diagnosis efficiency of a fault detection model can be accelerated, through introducing the arrangement entropy feature, the accurate extraction of the mechanism feature of the knitting machine can be realized, the position of particles in a batch-to-batch self-adaptive optimization initialization algorithm is carried out on the initial total group parameters of the sea gull in the SOA through gsin mapping, the space global dynamic search of the fault detection model parameters is realized, the situation that the fault detection model falls into the local optimal solution can be effectively reduced, the parameters in the algorithm are optimized, the optimizing speed in each cycle is accelerated, and the accuracy and the stability of the fault detection of the knitting machine can be improved.
In this embodiment, the step S2 includes the steps of:
S2-1, performing EMD (empirical mode decomposition) on the vibration data set obtained in the step S1, comparing the decomposed components with pearson, signal to noise ratio and root mean square difference of original vibration data to represent the noise content of the vibration data, and performing post-wiener Kalman denoising treatment;
S2-2, fitting a maximum value and a minimum value of vibration data x (t) through a spline interpolation method, obtaining an upper envelope line and a lower envelope line through cubic spline fitting, and calculating to obtain a mean value m 1 (t) of the envelope lines;
S2-3, judging whether the first-order modal component h 1 (t) meets the condition of the modal component, wherein the first-order modal component is used for reflecting the local oscillation characteristic of the signal. Each modal component is localized in time and frequency, which enables them to better capture transient and non-stationary characteristics in the signal, reflecting non-linear and non-stationary characteristics in complex non-linear and non-stationary signals. Wherein the first order modal components are: h 1（t）=x（t）-m1 (t);
Calculating h 1 (t) according to the above formula, and recording the current first-order modal component if h 1 (t) meets the following conditions simultaneously ：
1. The number of maximum points of the vibration signal x (t) differs from the number of zero crossing points by not more than 1;
2. The mean value of the upper envelope curve and the lower envelope curve of vibration data x (t) is constant at 0;
if the two conditions are not satisfied, repeating the step S2-2 until the kth iteration If the condition is satisfiedFor the first order modal component, denoted as c 1 (t), after the decomposition condition is met, the vibration data x (t) is decomposed into a number of modal components c j (t) and a residual signal r n (t), where j represents the j-th modal signal after the last iteration:
In step 2-3, the calculation formula of the low frequency residual signal r 1 (t) is:
；
The calculation formula of the residual signal in the iterative process, the low frequency residual signal r j (t), is:
；
the vibration signal x (t) decomposition expression is:
。
S2-4, denoising the vibration signal component with high noise by utilizing wiener Kalman denoising: comparing each modal component with various coefficients of the vibration signal x (t), and screening the modal components through a correlation coefficient r i and a correlation coefficient r i screening principle;
The calculation formula of the correlation coefficient r i is as follows:
；
Wherein x i is the modal component, Is the mean value of the modal components, y is the vibration signal,/>K is the average value of the vibration signals, and K is the number of layers of modal decomposition.
Therefore, EMD is rapidly decomposed, modal components can be decomposed in a very short time, and the real-time requirement of the follow-up fault detection of the model is met; because the knitting machine transmission system works at a high speed, a vibration sensor installed on the knitting machine transmission system collects true signals and simultaneously contains noise signals and knitting machine working characteristic information, a redundant periodic stable signal part in the knitting machine vibration signals can have a good filtering effect through wiener filtering, but is not suitable for a nonlinear system, accurate noise information statistics is needed, kalman filtering is suitable for a dynamic system, the condition of dynamic change of the system can be effectively processed, the Kalman filtering is self-adaptive and strong, filtering parameters can be dynamically adjusted according to the actual condition of the system, advantages of the two are complementary, the advantages are complementary through a combined mode, the wiener filtering is mainly used for denoising and filtering of the signals, the authenticity of the signals is estimated through minimizing mean square errors, the Kalman filtering is used for state estimation and system dynamic modeling, and optimal state estimation is provided through combining a system dynamic model and observation data.
Specifically, the wiener kalman denoising process includes:
S2-4-1, denoising and preprocessing the screened modal components through wiener filtering to improve signal quality and observability;
s2-4-2, establishing a dynamic model of a vibration signal of a transmission system of the knitting machine, comprising the following steps: state equations and observation equations to describe the evolution and observation processes of the transmission system:
the state vector of the transmission system is The observation vector is/>The time step is k. The state equation can be expressed as:
Wherein: is a state transition matrix describing the evolution of the state of the system from time k-1 to time k; /(I) Is an input control matrix describing the effect of external control on state; /(I)Is a control input vector,/>Is process noise that represents a random change in system state.
The observation equation can be expressed as:
Wherein: is an observation matrix describing how the state vector maps to the observation space; /(I) Is observation noise, representing random errors of the observed data; in the equation,/>Is Gaussian white noise set to a zero-mean matrix Q k, and the zero-mean matrix Q k is used in Kalman filtering to describe the process noise/>Characteristic parameters of/>Is Gaussian white noise set as covariance matrix R k, covariance matrix R k is used in Kalman filtering to describe observed noise/>Wherein the values of Q k and R k are gaussian distributed random variables that are independently co-distributed and are uncorrelated at different times or observation points. This enables the kalman filter to efficiently handle noise and provide an accurate estimate of the state of the system.
S2-2-3, estimating and predicting the state of the transmission system through Kalman filtering, and combining a system dynamic model and observation data to obtain optimal state estimation;
S2-4-4, processing the output of the Kalman filtering through the wiener filtering again to improve estimation accuracy and filtering effect. Therefore, the wiener filtering and the Kalman filtering are combined, better effects can be obtained in the vibration data processing and system control of the knitting machine, the method is particularly suitable for denoising of complex transmission systems and cooperative work among parts of the knitting machine, and the complex conditions of denoising of the vibration data and real-time state estimation of the transmission systems can be considered at the same time.
S2-5, correlation coefficient r i screening principle: if the correlation coefficient r i is smaller than ρ, the modal component needs to be subjected to denoising processing, if the correlation coefficient r i is larger than ρ, the relevance between the modal component and vibration data x (t) is high, characteristic information of the vibration data x (t) is reserved, the modal component is reserved, and the largest correlation coefficient r i is recorded as max Correlation coefficient ;
；
S2-6, adding the mode component after the wiener Kalman denoising, the residual mode component and the residual error at the same time, and reconstructing a pure reconstruction signal;
In this embodiment, in step S3, the time domain features include: maximum, minimum, peak-to-peak, mean, average amplitude, square root amplitude, variance, standard deviation, root mean square value, kurtosis, skewness, waveform factor, peak factor, pulse factor, margin factor, and clearance factor;
the frequency domain features include: center of gravity frequency, mean square frequency, frequency variance, and root mean square frequency;
the expression of the permutation entropy H PE (i) of the ith modal component after EMD decomposition is:
；
wherein: p j represents the probability of occurrence of the j-th symbol sequence of the modal component after phase space reconstruction, and N represents the length of the time sequence, which is used to represent the number of data points involved in the permutation. Therefore, the permutation entropy can reflect the dynamic characteristics and nonlinear characteristics of signal wave bands in the aspect of dynamic mutation detection, can amplify and extract the dynamic characteristics hidden in the vibration signals of the knitting machine well, and achieves better training of the follow-up fault diagnosis model on the characteristics of the knitting machine.
And (3) rearranging the characteristics into a data set, and performing model training and testing in the step (S4) according to the ratio of 7:3 as a fault diagnosis model training set and a test set.
In this embodiment, step S4 specifically includes the following steps:
s4-0, constructing gsinSOA-ELM model of EMD.
S4-0-1, in a gsinSOA-ELM model, extracting the characteristics of the same property from the vibration data reconstructed by the knitting machine under different working conditions to form characteristic vectors, dividing the characteristic vectors into training test sets, putting the training sets into different classifiers, and evaluating the characteristics and the optimal classifier according to classification results obtained by the classifiers.
In gsinSOA-ELM model, the activation function g (w ixi+bi) contains L hidden layer nodes
The calculation formula of the output matrix H (x) of (a) is as follows:
；
The actual output f L (x) of the model is defined, and the calculation formula of the actual output f L (x) is as follows:
；
to ensure that the model output error is minimal, solving equation e (x) yields:
Subtracting the output f L (x) of the network from the sample label T to obtain a minimum norm solution as an objective function e (x), obtaining H (x) β=t, solving the equation to equal order, searching the minimum two-dimensional solution of the linear equation H (x) β=t, and deriving the obtained optimal solution:
；
Wherein: w represents the connection weight (weight) of the input layer to the hidden layer, b represents the bias of the hidden layer, Representing an activation function (activation function)/>Using the sigmoid function, x represents the input data (input), β is the weight of the output layer, g (w ixi+bi) is the activation function,/>H is the hidden layer output matrix H (x) for the optimal parameters of the output layer. The Extreme Learning Machine (ELM) model has rapid training speed and good generalization capability, and is a classification model applicable to large-scale data sets.
At present, due to the fact that a plurality of transmission assemblies exist in fault diagnosis of a knitting machine, a traditional Extreme Learning Machine (ELM) model cannot achieve high-precision identification under the influence of multidimensional dynamic characteristics on training data, no feasible measure is available at present to seek an optimal parameter value, and a large amount of time is consumed by a parameter one-by-one debugging method and errors are prone to being caused. And performing optimization treatment on two parameters of the ELM by adopting gsinSOA algorithm, and then establishing an optimized ELM model to improve diagnosis precision.
The existing sine chaotic mapping is a chaotic system, and a chaotic sequence can be generated through a sine function. The sine function is a periodic nonlinear function, and when used in a chaotic system, can generate a complex, random sequence, and the sine chaotic map is conventionally formed as follows:
Wherein x n is the current state, x n+1 is the next state, a and h are constants, and are commonly used for regulating the behavior of the chaotic system, and the sequence generated by sine chaotic mapping shows randomness to a certain extent and has good pseudo-random property. However, because the sine chaotic mapping is realized based on the sine function, the sine chaotic mapping has periodicity, and if initialized particles are not sufficiently dispersed in space, the problem that the diagnostic model falls into a local optimal solution still occurs.
In order to solve the problem of jumping out of a local optimal solution in the optimizing process and improve a better effect of a random initial population, gsin mapping is provided to realize the initialization of a gull population:
Wherein: m, l represent constants, i represent the current iteration number, X best represent the currently known optimal solution, item max represent the set maximum iteration number, X n is the current state, X n+1 is the next state, |lx best-Xn | represent the difference between the optimal position of the particle generated at the current time and the particle at the current time in the optimizing process, A square operator representing the ratio of the current iteration number to the iteration number.
In gsin mapping, |lx best-Xn | is used as the difference between the optimal position of the particle generated at the current moment and the particle at the current moment in the optimizing process of the algorithm, and the difference is placed in the optimizing space and can be regarded as a vector parameter pointing to the direction of the optimal solution, so that the position and the direction of the initial gull particle are more changeable. Meanwhile, in order for the algorithm to continue to generate particle small-range optimization near the optimal solution of the searched space along with iteration, the distance of the initial position for generating particles needs to be controllable by combining the square operator of the ratio of the current iteration times to the iteration timesRandom search probability is increased so as to realize nonlinear self-adaptive dynamic adjustment of initial particle positions along with the increase of iteration times when parameter optimization is performed in an algorithm.
Therefore, in the fault detection of the knitting machine, the transmission system is numerous, the traditional Extreme Learning Machine (ELM) model cannot achieve high-precision identification under the influence of multidimensional dynamic characteristics on training data, no feasible measure is available at present to seek the optimal parameter value, and a large amount of time is easy to be consumed by a parameter one-by-one debugging method and errors are easy to cause. And performing optimization treatment on two parameters of the ELM by adopting gsinSOA algorithm, and then establishing an optimized ELM model to improve diagnosis precision.
The total particle number initialized by the seagull population is unchanged, the total particle number is divided into item max, and once in each cycle of the algorithm, one seagull particle is released for regional optimization. The value of i at the initial stage of the algorithm is small, the refreshing positions of the seagull particles of the same batch can be mapped to search and optimize in a large range in space, and as the circulation times are increased, the initial population particles release a certain amount of seagull particles near the optimal solution, the iteration times i are continuously increased, and the position distance of the released particles is closer, so that fine optimization is performed in a small range, nonlinear dynamic adjustment of the optimizing process is realized, and the self-adaptive optimizing of the algorithm at different moments is achieved. The mapping improved based on sine chaotic mapping is named gsin mapping, and population initialization in a seagull algorithm is realized through gsin mapping, so that the method has the advantages of self-adaption, wide search range and randomness for population initialization at different moments. Therefore, the self-adaptive wide optimizing of the algorithm in the whole space is achieved, and meanwhile, the method has the advantage of avoiding the algorithm from falling into a local optimal solution.
The migration behavior simulation algorithm achieves global-range searching, wherein the migration behavior simulation algorithm comprises three parts, namely updating of positions, updating of action directions and avoiding of collision of individuals.
The variable A is introduced to calculate the individual position C Z, and the calculation formula is:
；
Wherein f c is a control factor of the seagull total group, and is generally set to 2; t is the current iteration number; max iter is the maximum number of iterations; p z (t) is the current gull position.
The variable B is introduced to be responsible for balancing global optimization and local optimization, and the expression of the action direction of the model is as follows:
；
Wherein M Z (t) is the calculated optimal individual action direction, For optimal position of sea gull individual, r d is random number in [0,1], for optimizing parameter B, linear increment factor/>, is introducedThe calculation formula of B is:
;
After the linear increment factor is introduced, the linear increment factor becomes larger as i increases, so that the dynamic linear acceleration of the algorithm can be realized, and the algorithm search is accelerated.
In the simulation space, the individual seagull particles move according to the optimal direction, and after continuous iteration, the expression of the migration to the optimal position D Z（t）,DZ (t) is as follows:
S4-0-2, and optimizing in a small area while completing global searching. Therefore, the local search process for making the seagull foraging behavior a sense of inspiration is as follows.
The seagull foraging is performed by spiral type diving and capturing hunting object foraging, the behavior is simulated into a three-dimensional space, and the following motion trail formula can be obtained:
；/>
the local search formula for the seagull individuals is as follows:
；
Wherein: f c is the control factor of the total sea-gull group, t is the current iteration number, max iter is the maximum iteration number, P Z (t) is the current sea-gull position, M Z (t) is the calculated optimal individual action direction, P Z (t) is the optimal individual sea-gull position, r d is the random number in [0,1], r represents the spiral radius, and is controlled by constant u and constant v, Is a random number of (a) in the memory.
The SOA algorithm can improve global searching capability of parameter optimization, can find a better solution in a solution space, and is beneficial to solving the complex optimization problem. Meanwhile, the algorithm is further randomized on the initial value of the seagull at gsinSOA, so that stronger global searching capability is achieved, and the algorithm has the advantages of strong self-adaption and high convergence speed.
Based on gsinSOA-ELM model in S4-0, the process of optimizing connection weight w and hidden layer neuron threshold b in optimizing extreme learning machine to obtain the best classification model comprises:
S4-1, initializing a certain number of SOA populations through gsin mapping, and taking the model error rate as an fitness function;
s4-2, calculating individual adaptability function values of seagulls;
s4-3, comparing individual fitness values, and searching an optimal position;
S4-4, calculating an optimal fitness value of the prey, and updating the individual attack position;
S4-5, presetting the iteration times of the model in the step S4-0, judging whether the iteration number is reached in the step, and if not, repeating the steps S4-1 to S4-4 by taking the current optimal position as a search area; if yes, executing the step S4-7;
s4-6, repeating the steps S4-1 to S4-5 for a plurality of times until the iteration condition is met;
S4-7, obtaining an optimal connection weight w and a neuron threshold b, and further obtaining an optimal gsinSOA-ELM model. In the iterative search process of the gsinSOA-ELM model, the search range is reduced in each iteration, and further random point search is carried out in the reduced search range, so that the randomness of the algorithm is enhanced, the SOA algorithm can improve the global search capability of parameter optimization, and can find out a better solution in a solution space, thereby being beneficial to solving the complex optimization problem. Meanwhile, the algorithm is further randomized on the initial value of the seagull at gsinSOA, so that stronger global searching capability is achieved, and the algorithm has the advantages of strong self-adaption and high convergence speed.
In summary, according to the invention, EMD decomposition and wiener Kalman denoising are performed on the vibration data of the knitting machine, so that the rapid filtering of noise in the full frequency range of the vibration signal can be realized, meanwhile, the diagnosis efficiency of the fault detection model can be accelerated, the accurate extraction of the mechanism characteristics of the knitting machine can be realized by introducing the arrangement entropy characteristics, the position of particles in the batch-wise self-adaptive optimization initialization algorithm is performed on the initial total group parameters of seagulls in the SOA through gsin mapping, the spatial global dynamic search of the fault detection model parameters is realized, the optimization is performed in a small range of the area while the global search is completed, the initial value of seagulls is further randomized by the gsinSOA algorithm, the stronger global search capability is realized, the algorithm has the advantages of self-adaption strong and rapid convergence speed, the situation that the fault detection model falls into the local optimal solution can be effectively reduced, the parameter B value in the algorithm is optimized, the optimizing speed in each cycle is accelerated, and the fault detection accuracy and stability of the knitting machine can be improved.
The above-described preferred embodiments according to the present invention are intended to suggest that, from the above description, various changes and modifications can be made by the worker in question without departing from the technical spirit of the present invention. The technical scope of the present invention is not limited to the description, but must be determined as the scope of the claims.
Claims (8)
1. A knitting machine fault detection method based on gsinSOA-ELM model is characterized by comprising the following steps:
s1, acquiring a vibration data set of the knitting machine, which is acquired by a vibration sensor;
s2, performing EMD (empirical mode decomposition), wiener Kalman denoising and data reconstruction on the vibration data set obtained in the step S1 to obtain reconstructed data;
S3, performing time domain feature extraction, frequency domain feature extraction and permutation entropy feature extraction on the reconstruction data obtained in the step S2, and rearranging the features into a data set;
S4, establishing gsinSOA-ELM model, and optimizing the connection weight w and the neuron threshold b parameter in the model by using a seagull optimization algorithm based on gsin mapping to obtain an optimal gsinSOA-ELM model;
s5, collecting real-time vibration data of the knitting machine, performing fault detection by using the optimal gsinSOA-ELM model obtained in the step S4, and outputting a fault detection result of the knitting machine;
in step S4, the optimization process of the connection weight w and the neuron threshold b in the model by the gull optimization algorithm based on gsin mapping includes the following steps:
s4-1, initializing a seagull population through gsin mapping, setting a model error rate as an fitness function, and initializing a formula of the seagull population through gsin mapping, wherein the formula is as follows:
Wherein: m, l represent constants, i represent the current iteration number, X best represent the currently known optimal solution, item max represent the set maximum iteration number, X n is the current state, X n+1 is the next state, |lx best-Xn | represent the difference between the optimal position of the particle generated at the current time and the particle at the current time in the optimizing process, A square operator representing the ratio of the current iteration times to the iteration times;
S4-2, calculating the fitness function value of each sea gull individual;
S4-3, comparing fitness function values of individual seagulls, and searching a current optimal position;
S4-4, calculating an optimal fitness value of the prey, and updating the individual attack position;
s4-5, judging whether iteration times are reached, if not, repeating the steps S4-1 to S4-4 by taking the current optimal position as a search area; if yes, executing the step S4-6;
S4-6, obtaining the optimal connection weight w and the neuron threshold b, and further obtaining the optimal gsinSOA-ELM model.
2. The knitting machine fault detection method based on gsinSOA-ELM model according to claim 1, characterized in that in step S4, the global in-scope search is reached for the gull optimization algorithm based on gsin mapping, comprising: position updating, action direction updating and avoiding the mutual collision of individuals;
the variable A is introduced to calculate the position C Z of the sea gull individual, and the calculation formula is as follows:
；
the variable B is introduced to be responsible for balancing global optimization and local optimization, and the expression of the action direction of the model is as follows:
；
Introducing a linear increment factor The calculation formula of B is:
;
In the simulation space, the individual seagull particles move according to the optimal direction, and after continuous iteration, the expression of the migration to the optimal position D Z（t）,DZ (t) is as follows:
；
The seagull foraging captures the prey foraging by spiral dive, and the behavior is simulated into a three-dimensional space, so that the following motion trail formula is obtained:
；
the local search formula for the seagull individuals is as follows:
；
Wherein: f c is the control factor of the total sea-gull group, t is the current iteration number, max iter is the maximum iteration number, P Z (t) is the current sea-gull position, M Z (t) is the calculated optimal individual action direction, P Z (t) is the optimal individual sea-gull position, r d is the random number in [0,1], r represents the spiral radius, and is controlled by constant u and constant v, Is a random number of (a) in the memory.
3. The knitting machine failure detection method based on gsinSOA-ELM model according to claim 1, characterized in that in step S4, establishing gsinSOA-ELM model includes: and extracting the characteristics of the same property from the vibration data reconstructed by the knitting machine under different working conditions to form characteristic vectors, dividing the characteristic vectors into training test sets, putting the training sets into different classifiers, and evaluating the characteristics and the optimal classifier according to classification results obtained by the classifiers.
4. The knitting machine failure detection method based on gsinSOA-ELM model according to claim 1, characterized in that in step S4, in the gsinSOA-ELM model, the calculation formula of the activation function g (w ixi+bi) output matrix H (x) is:
；
The actual output f L (x) of the model is defined, and the calculation formula of the actual output f L (x) is as follows:
；
to ensure that the model output error is minimal, solving equation e (x) yields:
；
Subtracting the output f L (x) of the network from the sample label T to obtain a minimum norm solution as an objective function e (x), obtaining H (x) β=t, solving the equation to equal order, searching the minimum two-dimensional solution of the linear equation H (x) β=t, and deriving the obtained optimal solution:
；
Wherein: w represents the connection weight (weight) of the input layer to the hidden layer, b represents the bias of the hidden layer, Representing an activation function (activation function)/>Using the sigmoid function, x represents the input data (input),/>G (w ixi+bi) is an activation function, β is the weight of the output layer, and H is the hidden layer output matrix H (x) for the optimal parameters of the output layer.
5. The knitting machine failure detection method based on gsinSOA-ELM model according to claim 1, characterized in that the step S2 includes the steps of:
S2-1, performing EMD (empirical mode decomposition) on the vibration data set obtained in the step S1, comparing the decomposed components with pearson, signal to noise ratio and root mean square difference of original vibration data to represent the noise content of the vibration data, and performing post-wiener Kalman denoising treatment;
S2-2, fitting a maximum value and a minimum value of vibration data x (t) through a spline interpolation method, obtaining an upper envelope line and a lower envelope line through cubic spline fitting, and calculating to obtain a mean value m 1 (t) of the envelope lines;
S2-3, judging whether the first-order modal component h 1 (t) meets the condition of the modal component, if yes, the first-order modal component is h 1 (t), and if not, repeating the step S2-2 for k times until Satisfying the condition/>As a first-order modal component and denoted as c 1 (t), after the decomposition condition is satisfied, the vibration signal x (t) is decomposed into a plurality of modal components c j (t) and a residual signal r n (t); wherein, h 1（t）=x（t）-m1（t）;h1 (t) meets the judging condition of the modal component as follows: the number of maximum points of the vibration signal x (t) differs from the number of zero crossing points by not more than 1; the mean value of the upper envelope curve and the lower envelope curve of vibration data x (t) is constant at 0;
S2-4, denoising the vibration signal component with high noise by utilizing wiener Kalman denoising: comparing each modal component with various coefficients of the vibration signal x (t), and screening the first-order modal components according to a screening principle of a correlation coefficient r i and a correlation coefficient r i;
S2-5, if the correlation coefficient r i is smaller than rho, denoising processing is needed for the modal component, and if the correlation coefficient r i is larger than rho, the relevance between the modal component and the vibration signal x (t) is high, and characteristic information of the vibration signal x (t) is reserved, so that the modal component is reserved;
The calculation formula of the correlation coefficient r i is as follows:
；
The calculation formula of ρ is:
；
x i is the component of the mode shape, Is the mean value of the modal components, y is the vibration signal,/>The maximum correlation coefficient r i is marked as max Correlation coefficient for the mean value of the vibration signal and K is the number of layers of modal decomposition;
S2-6, adding the mode component after the wiener Kalman denoising, the residual mode component and the residual error at the same time, and reconstructing a pure reconstruction signal.
6. The knitting machine failure detection method based on gsinSOA-ELM model according to claim 5, characterized in that in step 2-3, the calculation formula of the low frequency residual signal r 1 (t) is:
；
The calculation formula of the low frequency residual signal rj (t) is:
；
the vibration signal x (t) decomposition expression is:
。
7. The method for detecting a knitting machine failure based on gsinSOA-ELM model according to claim 5, characterized in that S2-4 includes the steps of:
S2-4-1, denoising and preprocessing the screened modal components through wiener filtering to improve signal quality and observability;
s2-4-2, establishing a dynamic model of a vibration signal of a transmission system of the knitting machine, comprising the following steps: state equations and observation equations to describe the evolution and observation processes of the drive train;
S2-2-3, estimating and predicting the state of the transmission system through Kalman filtering, and combining a system dynamic model and observation data to obtain optimal state estimation;
s2-4-4, processing the output of the Kalman filtering through the wiener filtering again to improve estimation accuracy and filtering effect.
8. The knitting machine fault detection method based on gsinSOA-ELM model according to claim 1, characterized in that in step S3, the time domain features include: maximum, minimum, peak-to-peak, mean, average amplitude, square root amplitude, variance, standard deviation, root mean square value, kurtosis, skewness, waveform factor, peak factor, pulse factor, margin factor, and clearance factor;
the frequency domain features include: center of gravity frequency, mean square frequency, frequency variance, and root mean square frequency;
the expression of the permutation entropy H PE (i) of the ith modal component after EMD decomposition is:
；
Wherein: p j represents the probability of occurrence of the j-th symbol sequence of the modal component after phase space reconstruction, and N represents the length of the time sequence, which is used to represent the number of data points involved in the permutation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410410954.3A CN118013375B (en) | 2024-04-08 | 2024-04-08 | Knitting machine fault detection method based on gsinSOA-ELM model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410410954.3A CN118013375B (en) | 2024-04-08 | 2024-04-08 | Knitting machine fault detection method based on gsinSOA-ELM model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118013375A true CN118013375A (en) | 2024-05-10 |
| CN118013375B CN118013375B (en) | 2024-06-04 |
Family
ID=90947290
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410410954.3A Active CN118013375B (en) | 2024-04-08 | 2024-04-08 | Knitting machine fault detection method based on gsinSOA-ELM model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118013375B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114689322A (en) * | 2022-04-24 | 2022-07-01 | 沈阳航空航天大学 | Medium bearing fault diagnosis method based on grey wolf algorithm optimization ELM |
| US20230273093A1 (en) * | 2022-02-28 | 2023-08-31 | Zhejiang University | Method and device for fault diagnosis of wind turbine pitch bearing based on neural network |
| CN117471252A (en) * | 2023-10-31 | 2024-01-30 | 沈阳工业大学 | DC fault arc diagnosis method based on ICEEMDAN-KPCA feature extraction and SOA-KELM |
-
2024
- 2024-04-08 CN CN202410410954.3A patent/CN118013375B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20230273093A1 (en) * | 2022-02-28 | 2023-08-31 | Zhejiang University | Method and device for fault diagnosis of wind turbine pitch bearing based on neural network |
| CN114689322A (en) * | 2022-04-24 | 2022-07-01 | 沈阳航空航天大学 | Medium bearing fault diagnosis method based on grey wolf algorithm optimization ELM |
| CN117471252A (en) * | 2023-10-31 | 2024-01-30 | 沈阳工业大学 | DC fault arc diagnosis method based on ICEEMDAN-KPCA feature extraction and SOA-KELM |
Non-Patent Citations (2)
| Title |
|---|
| 孟博等: "基于VMD和优化SSA-ELM的齿轮箱故障诊断", 郑州大学学报(理学版), 13 March 2023 (2023-03-13) * |
| 王世豪: "冲击噪声下基于演化神经网络的调制识别方法研究", 中国优秀硕士学位论文全文数据库 (信息科技辑), 15 January 2024 (2024-01-15) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118013375B (en) | 2024-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhu et al. | A review of the application of deep learning in intelligent fault diagnosis of rotating machinery | |
| Jiao et al. | Deep coupled dense convolutional network with complementary data for intelligent fault diagnosis | |
| KR102145407B1 (en) | Method and control device for controlling technical systems | |
| Wu et al. | A weighted deep domain adaptation method for industrial fault prognostics according to prior distribution of complex working conditions | |
| Roverso | Plant diagnostics by transient classification: The aladdin approach | |
| CN112033463A (en) | Nuclear power equipment state evaluation and prediction integrated method and system | |
| Pu et al. | A one-class generative adversarial detection framework for multifunctional fault diagnoses | |
| EP4275149A1 (en) | Generation of realistic data for training of artificial neural networks | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN115809596A (en) | Digital twin fault diagnosis method and device | |
| Ma et al. | A collaborative central domain adaptation approach with multi-order graph embedding for bearing fault diagnosis under few-shot samples | |
| Zhao et al. | A novel semi-supervised graph-guided approach for intelligent health state diagnosis of a 3-PRR planar parallel manipulator | |
| CN117272022B (en) | Detection method of MEMS oscillator | |
| Zhang et al. | Graph neural network-based bearing fault diagnosis using Granger causality test | |
| CN118013375B (en) | Knitting machine fault detection method based on gsinSOA-ELM model | |
| Vo et al. | Harnessing attention mechanisms in a comprehensive deep learning approach for induction motor fault diagnosis using raw electrical signals | |
| CN114266013B (en) | Transmission system vibration decoupling method based on deep learning virtual perception network | |
| Luo et al. | A novel method for remaining useful life prediction of roller bearings involving the discrepancy and similarity of degradation trajectories | |
| Berenji et al. | Wavelet neural networks for fault diagnosis and prognosis | |
| Dhekane et al. | Enhanced annotation framework for activity recognition through change point detection | |
| CN114841196A (en) | Mechanical equipment intelligent fault detection method and system based on supervised learning | |
| Makansi et al. | Feature Generation and Evaluation for Data-Based Condition Monitoring of a Hydraulic Press | |
| Bodo et al. | Feature ranking under industrial constraints in continuous monitoring applications based on machine learning techniques | |
| Tang et al. | Research on prognosis for engines by LSTM deep learning method | |
| Chen et al. | A Multivariate Time Series Anomaly Detection Method Based on Generative Model |
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 | ||
| GR01 | Patent grant |