Search

CN-121996932-A - Health state assessment method for all subsystems of rapier loom

CN121996932ACN 121996932 ACN121996932 ACN 121996932ACN-121996932-A

Abstract

The invention relates to the technical field of state evaluation, in particular to a health state evaluation method of subsystems of a rapier loom, which comprises the steps of selecting characteristic parameters, extracting based on wavelet transformation characteristics, building a health state model based on GAN enhanced abnormal data, carrying out attribute reduction and rule extraction on evaluation indexes by utilizing a rough set theory, building an initial confidence rule base of each subsystem, carrying out fusion processing on activated rules by a evidence reasoning algorithm, outputting confidence degrees of each subsystem under each health state level, carrying out data distribution hierarchical sampling processing on missing data, carrying out secondary fusion on all reasoning conclusions by utilizing an ER algorithm in combination with a plurality of reasoning results to obtain a final reasoning result, taking a reasoning error as an objective function, carrying out interpretation constraint WOA algorithm on the parameter optimization of the confidence rule base, and updating the optimized rule base by using an incremental rough set, and keeping the recent data by using a sliding window to enable the model to continuously learn new data. The evaluation precision and reliability are remarkably improved.

Inventors

  • XIAO YANJUN
  • WANG FUZHEN

Assignees

  • 河北工业大学

Dates

Publication Date
20260508
Application Date
20260204

Claims (10)

  1. 1. The health state evaluation method of each subsystem of the rapier loom is characterized in that the method optimizes the parameters of the confidence rule base and the dynamic update mechanism of the confidence rule base based on an interpretability constraint whale algorithm, and comprises the following steps: The method comprises the steps of firstly, exploring fault mechanisms of all subsystems of a rapier loom, selecting characteristic parameters, extracting characteristics based on wavelet transformation, and enhancing abnormal data based on GAN so as to construct a health state model; secondly, performing attribute reduction and rule extraction on the evaluation index by utilizing a rough set theory, and constructing an initial confidence rule base of each subsystem of the rapier loom; Thirdly, fusing the activated rules by utilizing an evidence reasoning algorithm, and outputting the confidence coefficient of each subsystem under each health state level; step four, carrying out hierarchical sampling processing on missing data based on data distribution, and carrying out secondary fusion on all reasoning conclusions by utilizing an ER algorithm in combination with a plurality of reasoning results to obtain a final reasoning result; Optimizing parameters of a belief rule base by using a WOA algorithm with an interpretability constraint by taking the reasoning error as an objective function; and step six, updating the optimized rule base by using the incremental rough set, and reserving recent data by using a sliding window to enable the model to continuously learn the new data.
  2. 2. The method for evaluating health status of each subsystem of a rapier loom according to claim 1, wherein in the step one Each subsystem of the rapier loom comprises a warp let-off subsystem, a weaving subsystem and a reeling subsystem, wherein the characteristic parameters comprise a temperature signal, a vibration signal and a tension signal, and the specific steps of the first step are as follows: Preprocessing the characteristic parameters; Carrying out signal decomposition and extraction on the characteristic parameters by utilizing wavelet transformation, and carrying out multi-scale time-frequency characteristic and characteristic selection, so as to enhance fault sensitive information in the signals; And generating abnormal data by using the GAN, and balancing the distribution of the training set.
  3. 3. The method for evaluating the health status of each subsystem of a rapier loom according to claim 2, characterized in that the wavelet transformation comprises the following specific steps: Signal decomposition: performing multi-scale wavelet packet decomposition on a vibration signal of a rapier loom, selecting a Daubechies 4 wavelet basis function, wherein the decomposition layer number is 5 layers, and 16 sub-frequency bands are obtained, and the decomposition formula is as follows: ; Wherein, the As a parameter of the dimensions of the device, In order to be able to carry out the parameters of the translation, Is a wavelet basis function; feature extraction: and (3) calculating energy entropy, namely calculating the energy entropy for each layer of wavelet coefficient, and representing the complexity of the signal: , ; multi-scale feature fusion: Fusing the energy entropy of each layer with the original time domain feature to form a comprehensive feature vector which is used as the input of a confidence rule base; Feature selection: ReliefF algorithm weight update: ; Wherein NH is the same kind nearest neighbor, NM is the different kind nearest neighbor, and k is the sampling frequency.
  4. 4. The method for evaluating the health status of each subsystem of a rapier loom according to claim 2, wherein the step of generating abnormal data by using GAN and balancing the distribution of the training set is as follows: The generator adopts a full-connection layer and deconvolution structure, inputs random noise, and outputs characteristic vectors of synthesized vibration, temperature and tension signals; a discriminator, which adopts a convolutional neural network, inputs true/synthetic feature vectors and outputs discrimination probability; A loss function, namely using WASSERSTEIN GAN loss to improve training stability; ; data balancing, namely generating abnormal samples through GAN, and adjusting the ratio to be 3:1.
  5. 5. The method for evaluating the health status of each subsystem of a rapier loom according to claim 1, wherein the specific steps of extracting the confidence rule by using the relative reduction of the rough set in the second step are as follows: Dividing the data into equivalence classes according to the attribute reference value; Calculating the confidence coefficient of the decision rule with the same condition attribute under different decision attributes, and extracting the rule with higher confidence coefficient based on the rough set theory; and (3) reasoning by using the extracted confidence rules, and optimizing the number and the distance of the initially determined attribute reference values according to the reasoning result.
  6. 6. The method for evaluating the health status of each subsystem of a rapier loom according to claim 1, wherein the specific steps of the third step are as follows: After the model receives the input information, matching the input parameters with rule front pieces of each confidence rule; and combining and weighting the matched confidence rules by using a evidence reasoning algorithm to comprehensively consider the influence of each rule on the output result, thereby obtaining the output confidence result.
  7. 7. The method for evaluating the health status of each subsystem of a rapier loom according to claim 1, wherein the step four of processing missing data comprises the following specific steps: carrying out statistical analysis on the historical data to determine the data distribution characteristics of the missing attribute; performing hierarchical sampling, dividing data distribution into a plurality of layers or intervals, and extracting samples according to the distribution condition of data in each interval and the proportion; And combining the reconstructed attribute value with other known attribute information, and comprehensively evaluating through a confidence rule base and a evidence reasoning algorithm to finally obtain the health state of the subsystem.
  8. 8. The method for evaluating the health status of each subsystem of a rapier loom according to claim 1, wherein the specific steps of the fifth step are as follows: initialization and expert knowledge fusion: Initializing whale population, and defining search space and maximum iteration times; Expert knowledge guides spreading points, takes expert experience as a center, generates initial whale positions in the neighborhood of the expert knowledge, ensures that an optimization starting point accords with an actual system mechanism, and is a first step The individual whales may be expressed as: ; Wherein: confidence level of expert knowledge; To return one Is a random matrix of (a); fitness calculation and constraint design: Calculating the fitness value of each whale individual, and taking the mean square error as an objective function; Interpretability constraint: Whale behavior simulation and parameter optimization: surrounding prey, namely updating whale positions according to the current optimal solution, and reducing the searching range; Spiral predation, namely simulating whale spiral approaching prey behaviors, and locally and finely adjusting parameters; Randomly searching, namely randomly selecting individual positions when the exploration coefficient exceeds a threshold value, and enhancing the global searching capability; Dynamically updating and verifying: continuously verifying the rationality of the rule base in the iterative process, and resetting parameters if the constraint condition is violated; and outputting a final optimized confidence rule base, and ensuring that the final optimized confidence rule base has high precision and interpretability.
  9. 9. The method for evaluating the health status of each subsystem of a rapier loom of claim 8, wherein the interpretable constraint comprises: Constraint 1, the optimized regular confidence distribution is required to be consistent with the actual health state of each system of the rapier loom; ; under the kth rule, the interpretation constraint of the confidence distribution is marked as Ek, which is determined by the analysis of each practical system and is in a non-fixed form, wherein K is the number of confidence rules; Constraint 2, the adjustment range of the rule confidence coefficient is required to be within a reasonable interval preset by an expert; ; Wherein: The kth confidence level for the nth rule; 、 The minimum and maximum values of confidence given in connection with each expert are respectively given.
  10. 10. The method for evaluating the health status of each subsystem of a rapier loom according to claim 1, wherein in the sixth step, the incremental rough set update comprises the following specific steps: when new data arrives, only the confidence and weight of the affected rule are updated, and the formula is: ; the sliding window mechanism is characterized by retaining the data of the last 7 days, forgetting the old data according to exponential decay and decay factor The dynamic adjustment rule weight is: 。

Description

Health state assessment method for all subsystems of rapier loom Technical Field The invention relates to the technical field of state evaluation, in particular to a health state evaluation method for all subsystems of a rapier loom. Background Along with the development of the diversity of new generation information technology, the textile industry is also developed towards the direction of intellectualization and high-end, along with the improvement of industrial automation level, the requirements on the reliability and stability of mechanical equipment are higher and higher, the rapier loom is used as key equipment, and the health state evaluation is crucial to the guarantee of the production efficiency and the product quality. Traditional health state assessment methods rely on expert experience and periodic checks, lacking in real-time and accuracy. Nowadays, the health status assessment of a mechanical device is a way to perform "predictive maintenance" on the device, taking necessary maintenance measures by grasping the real-time running status of the device. Meanwhile, the existing health state assessment method has the following defects: the redundant information problem is that when the health state evaluation method (such as a support vector machine, a neural network and the like) based on data driving is used for processing the health state evaluation of the rapier loom, a large amount of redundant information is often introduced, so that the model is high in complexity and low in calculation efficiency. The problem of scarce abnormal data is that the abnormal working state is very small in the actual operation process of the rapier loom, so that the effective data is lost, and the problem of model bias caused by data imbalance can occur. The problem of data loss is that in actual operation, the monitoring data of the rapier loom is often lost due to sensor faults or human misoperation, and the existing method is difficult to effectively process the problem of data loss, so that an evaluation result is inaccurate. The reasoning error problem is that the traditional method (such as fuzzy reasoning, analytic hierarchy process and the like) relies on expert experience, so that the reasoning error is large, and the method is difficult to adapt to complex and changeable running states of the rapier loom. The equipment state change problem is that after the equipment runs for a long time, the mechanical abrasion and the working condition change, and the static evaluation model can be gradually invalid. Based on the above problems, the prior China patent with publication number CN118411154B discloses a power distribution equipment safety state evaluation method and system, integrates historical data and machine learning technology, builds a fault prediction model, and improves the reliability and maintenance efficiency of power supply. However, the method depends on a large amount of monitoring data, the rapier loom is rare in abnormal data during working, the data is strongly coupled and nonlinear, so that effective data is lost, a small-scale abnormal state is difficult to accurately identify, the model lacks explanation of an intrinsic change mechanism of the system, and an evaluation result is inaccurate. Chinese patent publication No. CN118607065A discloses a building structure deformation prediction system based on a fuzzy logic control algorithm, which performs probability prediction on the impending structural deformation based on building structure deformation data in combination with fuzzy logic controller output and machine learning technology. But fuzzy reasoning relies on expert experience and priori knowledge, accuracy is limited, and models focus on static features, ignoring equipment dynamic changes. Therefore, how to reduce the influence of uncertainty factors and loss of valid data on the evaluation result, and to improve the robustness, accuracy and precision of health status evaluation are problems that a person skilled in the art needs to solve at present. Therefore, the application provides a health state evaluation method for each subsystem of the rapier loom. Disclosure of Invention In order to make up the defects of the prior art and solve the technical problems in the background art, the invention provides a health state evaluation method of each subsystem of a rapier loom. The invention is realized by the following technical scheme: the method for evaluating the health state of each subsystem of the rapier loom optimizes the parameters of the confidence rule base and the dynamic update mechanism of the confidence rule base based on a whale algorithm with the interpretation constraint, and comprises the following steps: The method comprises the steps of firstly, exploring fault mechanisms of all subsystems of a rapier loom, selecting characteristic parameters, extracting characteristics based on wavelet transformation, and enhancing abnormal data based on GAN so as to construc