A textile machine working fault state monitoring system and fault data analysis method

Abstract

The present application relates to the field of data analysis technology, and specifically to a textile machine working fault state monitoring system and fault data analysis method, the method comprising: obtaining data of each operating parameter of the textile machine at each moment in the real-time monitoring cycle, and data of each operating parameter at each moment in each historical fault cycle when each fault occurs, and the fault type corresponding to each historical fault cycle; determining the abnormal score and noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle; determining the fault characterization strength and fault mode sensitivity; determining the local window length; obtaining the smoothed data of each operating parameter at all moments in the real-time monitoring cycle, combining with a neural network model, obtaining the probability corresponding to each fault type, and diagnosing the fault of the textile machine. When the present application smoothes the data of the textile machine, it can retain more potential fault information that may be contained in the data, and improve the recognition effect of multiple faults of the textile machine.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis, and in particular to a textile machine working fault status monitoring system and a fault data analysis method. Background Art

[0002] The textile industry is a pillar industry in the traditional industry. The textile machine is the core equipment of the textile industry. It is a tool for processing raw materials such as thread, silk, and hemp into silk thread and then weaving it into cloth. With the continuous improvement of scientific level, the textile industry has been able to realize automated textile production, but the intelligent process of data collection, status detection, fault diagnosis, life prediction, etc. of textile machine equipment is generally low. Among them, real-time monitoring of the working status of the textile machine and timely troubleshooting can help reduce the failure rate and downtime of the textile machine and improve the production efficiency of the textile industry.

[0003] Due to the complex environment of the textile workshop, there is strong electromagnetic interference caused by electrical circuits and noise interference caused by equipment operation, so it is necessary to clean the acquired textile machine data. The Locally Weighted Scatterplot Smoothing (Lowess) algorithm performs weighted fitting through the local features of the data points. It is suitable for processing data sets with noise and nonlinear patterns. It can effectively correct the monitoring data of textile machines and greatly improve the accuracy of textile machine fault diagnosis. When the existing technology uses the Lowess algorithm to smooth the working monitoring data of textile machines, it is necessary to pre-set a fixed local window length, which is easy to cause overfitting of the working monitoring data of the textile machine and loss of fault feature information, resulting in a low recognition rate of faults in the working state of the textile machine, affecting the production efficiency of the textile machine. Summary of the invention

[0004] In order to solve the above technical problems, a textile machine working fault status monitoring system and a fault data analysis method are provided to solve the existing problems.

[0005] The solution to the technical problem of the present application is to provide a textile machine working fault status monitoring system and a fault data analysis method, comprising the following steps:

[0006] In a first aspect, an embodiment of the present application provides a method for analyzing working fault data of a textile machine, the method comprising the following steps:

[0007] Obtain the data of each operating parameter of the textile machine at each moment in the real-time monitoring cycle, and the data of each operating parameter at each moment in each historical fault cycle when each fault occurs, as well as the fault type corresponding to each historical fault cycle;

[0008] Determine the abnormal score of each operating parameter at each moment in the real-time monitoring cycle according to the mutation characteristics and change trends of each operating parameter at different moments in the real-time monitoring cycle, as well as the abnormal changes; analyze the deviation of the abnormal score to determine the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle;

[0009] Cluster all historical fault cycles of each operating parameter according to the discreteness of the data at all times in each historical fault cycle of each operating parameter; analyze the number of all elements in each cluster corresponding to each operating parameter and the density of the element distribution to determine the fault characterization strength of each cluster corresponding to each operating parameter; determine the fault mode sensitivity of each operating parameter based on the fault characterization strength;

[0010] Based on the noise interference deviation and the fault mode sensitivity, the local window length of the filtering algorithm for each operating parameter at each moment in the real-time monitoring period is determined; based on the local window length, a filtering algorithm is used to obtain the smoothed data of each operating parameter at all moments in the real-time monitoring period, and combined with a neural network model, the probability corresponding to each fault type is obtained to diagnose the fault of the textile machine.

[0011] Preferably, each of the operating parameters includes yarn tension, yarn speed, horizontal vibration data, vertical vibration data, spindle speed, voltage, current, humidity, and temperature.

[0012] Preferably, determining the abnormal score of each operating parameter at each time comprises:

[0013] Perform mutation point detection on the data of each operating parameter at all times in the real-time monitoring cycle to obtain the mutation probability of each operating parameter at each time in the real-time monitoring cycle;

[0014] The data of each operating parameter at each time in the real-time monitoring cycle and multiple times in its neighborhood are combined into a local monitoring sequence;

[0015] Using Sen slope estimation algorithm to obtain the trend slope of the local monitoring sequence;

[0016] The mutation probability and the trend slope are combined into a two-dimensional array, and anomaly detection is performed on the two-dimensional array of each operating parameter at all times in the real-time monitoring cycle to obtain an anomaly score of each operating parameter at each time in the real-time monitoring cycle.

[0017] Preferably, the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle is determined by the following calculation formula: ,in, For the The operating parameters are the first The noise interference deviation at the moment, For the The operating parameters are the first The abnormal score at the time, For the The average of the abnormal scores of the operating parameters at all times in the real-time monitoring period, is the preset initial interference coefficient.

[0018] Preferably, clustering all historical fault cycles of each operating parameter includes:

[0019] The mean and standard deviation of the data of each operating parameter at all times in each historical fault cycle are formed into a two-dimensional array;

[0020] The two-dimensional array of each operating parameter in all historical fault cycles is clustered to obtain multiple clusters.

[0021] Preferably, the determining of the fault characterization strength of each cluster corresponding to each operating parameter includes:

[0022] Calculate the mean of the distances between all elements in each cluster and its cluster center corresponding to each operating parameter, and record it as the first mean;

[0023] The ratio of the number of all elements in each cluster corresponding to each operating parameter to the first mean value is used as the fault characterization strength of each cluster corresponding to each operating parameter.

[0024] Preferably, the failure mode sensitivity of each operating parameter is determined by a calculation formula: ,in, For the The failure mode sensitivity of the operating parameters, For the The number of all clusters corresponding to the running parameters of the item, For the The operation parameters corresponding to The fault characterization strength of the clusters, is the normalization function.

[0025] Preferably, the local window length of the filtering algorithm for each operating parameter at each moment in the real-time monitoring cycle is determined by the following calculation formula: ,in, For the The operating parameters are the first The local window length of the moment filtering algorithm, To preset the initial window length, For the The operating parameters are the first The noise interference deviation at the moment, For the Failure mode sensitivity of an operating parameter.

[0026] Preferably, the diagnosing the fault of the textile machine includes: if the probability is greater than a preset threshold, the textile machine has a fault; otherwise, the textile machine does not have a fault.

[0027] In the second aspect, an embodiment of the present application also provides a textile machine working fault status monitoring system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of any one of the above-mentioned textile machine working fault data analysis methods are implemented.

[0028] This application has at least the following beneficial effects:

[0029] The present application determines the abnormal score of each operating parameter at each moment in the real-time monitoring cycle according to the mutation characteristics and change trends of each operating parameter at different moments in the real-time monitoring cycle, as well as the abnormal changes; analyzes the offset of the abnormal score to determine the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle, which has the beneficial effect of considering the local change trend and abnormal change situation in the real-time monitoring cycle, analyzing the instability of the change of the operating parameters of the textile machine, so as to reflect the degree to which each operating parameter of the textile machine is affected by the noise of the textile workshop; clustering all historical fault cycles of each operating parameter according to the discrete situation of the data of all moments of each operating parameter in each historical fault cycle; analyzing the number of all elements in each clustering cluster corresponding to each operating parameter and the density of the element distribution, and determining the fault characterization strength of each clustering cluster corresponding to each operating parameter; based on the fault characterization strength, determining the fault mode sensitivity of each operating parameter, which has the beneficial effect of analyzing the occurrence of historical faults. During the period, each operating parameter characterizes each fault type to reflect the sensitivity of each operating parameter to the fault type; based on the noise interference deviation and the fault mode sensitivity, the local window length of the filtering algorithm for each operating parameter at each moment in the real-time monitoring period is determined, and the filtering algorithm is used to obtain the smoothed data of each operating parameter at all moments in the real-time monitoring period. The beneficial effect is that it can avoid the situation where highly sensitive data changes and fluctuates greatly when a textile machine fault occurs, and a larger local window length is easily adopted. By selecting a suitable local window length, not only can the data be smoothed and denoised, but more potential fault information that may be contained in the highly sensitive textile machine data can also be retained; combined with the neural network model, the probability of each fault type is obtained, and the fault of the textile machine is diagnosed. The beneficial effect is that it can improve the recognition effect of multiple faults of the textile machine, effectively reduce the subsequent production risks caused by insufficient fault diagnosis of the textile machine, timely repair the textile machine fault, and improve the production efficiency of the textile machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] A textile machine working fault data analysis method of the present application is further described in detail below in conjunction with the accompanying drawings.

[0031] Figure 1 A flowchart of a method for analyzing working fault data of a textile machine provided in an embodiment of the present application;

[0032] Figure 2 A flowchart of the steps of a method for obtaining an abnormal score of each operating parameter at each moment in a real-time monitoring cycle provided in an embodiment of the present application;

[0033] Figure 3 A schematic diagram of clustering of textile machine operating parameters provided in an embodiment of the present application;

[0034] Figure 4 A flowchart of the steps of a method for obtaining the fault characterization strength of each cluster corresponding to each operating parameter provided in an embodiment of the present application;

[0035] Figure 5 and Figure 6 Schematic diagram of the smoothing effect of the Lowess algorithm before and after the improvement provided in the embodiment of the present application. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the following is a further detailed description of a textile machine working fault state monitoring system and a fault data analysis method proposed in the present application in conjunction with the accompanying drawings and implementation examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0037] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0038] See also Figure 1 , which shows a flowchart of a method for analyzing working fault data of a textile machine provided by an embodiment of the present application, the method comprising the following steps:

[0039] Step 1, obtaining the data of each operating parameter of the textile machine at each moment in the real-time monitoring cycle, the data of each operating parameter at each moment in each historical fault cycle when each fault occurs, and the fault type corresponding to each historical fault cycle.

[0040] The Supervisory Control And Data Acquisition (SCADA) system collects data of each operating parameter of the textile machine from various sensors and measuring devices of the textile machine in real time, and stores all the data of each operating parameter collected, wherein the collection time interval is t and the collection cycle is T, and normalizes all the data collected for each operating parameter to obtain the data of each operating parameter at each moment in the real-time monitoring cycle;

[0041] Secondly, the data of each operating parameter at the time of N failures of the textile machine are collected from the SCADA system, where the collection time interval is t and the collection cycle is T, and the data of each operating parameter collected are normalized. Therefore, the data of each operating parameter at each moment when the textile machine has N failures are recorded as the data of each operating parameter at each moment in each historical failure cycle; secondly, the textile machine has N failures, and each failure corresponds to a failure type. Therefore, the failure type corresponding to each historical failure cycle is obtained.

[0042] Preferably, in this embodiment, an optical mouse sensor is used to measure the yarn tension data and yarn speed data of the textile machine; eddy current sensors are installed in the horizontal and vertical directions of the spindle bearing of the textile machine, respectively, to measure the horizontal vibration data and vertical vibration data of the spindle bearing of the textile machine; an incremental encoder is installed on the driving motor of the main shaft of the textile machine to measure the main shaft speed data of the textile machine; an electric energy meter is installed on the working circuit of the textile machine to measure the voltage data and current data of the textile machine when it is working; a temperature and humidity sensor is installed inside the textile machine to measure the humidity data and temperature data of the textile machine; thereby, the data of each operating parameter at each moment in the real-time monitoring cycle and the data of each operating parameter at each moment in each historical fault cycle are obtained, wherein each operating parameter includes yarn tension, yarn speed, horizontal vibration data, vertical vibration data, main shaft speed, voltage, current, humidity, and temperature.

[0043] It should be noted that the fault types that occur each time may include yarn guide abnormality, yarn breakage, bearing overheating, gear meshing, bearing wear, motor winding failure, motor overload and other fault types.

[0044] Preferably, in this embodiment, the collection time interval is 0.1s and the collection cycle is 5min. As other implementation modes, the implementer can set them according to the actual situation. Z-score standardization is used for normalization processing, wherein Z-score standardization is a well-known technology and will not be described in detail here. As other implementation modes, the implementer can adopt other methods of the prior art, such as the maximum and minimum value normalization method, etc. This embodiment does not impose any special restrictions on this. The data of each operating parameter at each moment in each historical fault cycle when the textile machine has 100 faults is collected. The implementer can set it according to the actual situation.

[0045] At this point, the data of each operating parameter of the textile machine at each moment in the real-time monitoring cycle, the data of each operating parameter at each moment in each historical fault cycle when each fault occurs, and the fault type corresponding to each historical fault cycle are obtained.

[0046] Step 2: Determine the abnormal score of each operating parameter at each moment in the real-time monitoring cycle according to the mutation characteristics and change trends of each operating parameter at different moments in the real-time monitoring cycle, as well as the abnormal changes; analyze the deviation of the abnormal score to determine the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle.

[0047] Since the production process of textile machines is often highly standardized and automated, and the same instructions are always executed repeatedly, the operating parameters of textile machines have the ability to characterize textile machine failures. In theory, each operating parameter should have strong stability and repeatability within a collection cycle. However, in the actual working process of textile machines, due to the complex environment of the textile workshop site, strong electromagnetic interference caused by electrical circuits, and noise interference caused by equipment operation, there are actually a large number of prominent points caused by noise in the operating parameters of textile machines, which destroy the stability and repeatability of the operating parameters of textile machines.

[0048] Based on the above analysis, the mutation characteristics and change trends of each operating parameter in the real-time monitoring period are analyzed to reflect the stability of the textile machine during operation, specifically:

[0049] The mutation point detection algorithm is used to perform mutation point detection on the data of each operating parameter at all times in the real-time monitoring cycle to obtain the mutation probability of each operating parameter at each time in the real-time monitoring cycle;

[0050] Preferably, in this embodiment, a Bayesian mutation detection algorithm is used to obtain the mutation probability of each operating parameter at each moment in the real-time monitoring cycle, wherein the Bayesian mutation detection algorithm is a well-known technology and will not be described in detail here.

[0051] The data of each operating parameter at each time in the real-time monitoring cycle and multiple times in its neighborhood are combined into a local monitoring sequence;

[0052] Preferably, in this embodiment, data of each moment and 10 moments in its neighborhood are selected as other implementation methods, and implementers can set them according to actual conditions.

[0053] Using the Sen slope estimation algorithm, the trend slope of the local monitoring sequence of each operating parameter at each moment in the real-time monitoring cycle is obtained;

[0054] It should be noted that the Sen slope estimation algorithm is a well-known technology and will not be described in detail here. Through the Sen slope estimation algorithm, the influence of outliers can be eliminated and the local trend of the operating parameter data of the textile machine can be more accurately reflected.

[0055] The mutation probability and the trend slope are combined into a two-dimensional array, and an anomaly detection algorithm is used to perform anomaly detection on the two-dimensional array of each operating parameter at all times in the real-time monitoring cycle to obtain an anomaly score of each operating parameter at each time in the real-time monitoring cycle;

[0056] Preferably, in this embodiment, the LOF anomaly detection algorithm is used for anomaly detection, wherein the neighborhood distance of the LOF anomaly detection algorithm is obtained by a cross-validation method, and an anomaly score is obtained. The LOF anomaly detection algorithm is a well-known technology and will not be described in detail here.

[0057] It should be noted that the lower the abnormality score is, the more serious the data distortion of the corresponding operating parameters is, the more thoroughly the stability and repeatability are destroyed, and the more likely it is to be significantly affected by the noise in the textile workshop.

[0058] Further, the step flow chart of the method for obtaining the abnormal score of each operating parameter at each moment in the real-time monitoring cycle provided in the embodiment of the present application is as follows: Figure 2 shown.

[0059] Further, based on the abnormal score, the noise interference deviation is determined to reflect the degree to which each operating parameter is affected by the noise of the textile workshop at each moment in the real-time monitoring cycle, specifically:

[0060] The calculation formula for the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle is: ,in, For the The operating parameters are the first The noise interference deviation at the moment, For the The operating parameters are the first The abnormal score at the time, For the The average of the abnormal scores of the operating parameters at all times in the real-time monitoring period, is the preset initial interference coefficient.

[0061] Preferably, in this embodiment, the initial interference coefficient is preset The value is 0.15. It is related to the robustness of the textile machine control system. As other implementation methods, the implementer can set it according to the actual situation.

[0062] It should be noted that The larger it is, the more serious the data distortion of the corresponding operating parameters is, and the more likely it is to be significantly affected by the noise in the textile workshop. The larger the noise interference deviation is, the greater the deviation between the measured data and the actual value of each operating parameter at each moment is.

[0063] At this point, the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle is obtained.

[0064] Step 3: Cluster all historical fault cycles of each operating parameter according to the discreteness of the data at all times in each historical fault cycle of each operating parameter; analyze the number of all elements in each cluster corresponding to each operating parameter and the density of the element distribution to determine the fault characterization strength of each cluster corresponding to each operating parameter; based on the fault characterization strength, determine the fault mode sensitivity of each operating parameter.

[0065] Based on the above analysis, the larger the noise interference deviation is, the larger the local window size should be used when smoothing and denoising the data through the Locally Weighted Scatterplot Smoothing (Lowess) algorithm to smooth out the noise effect in the data. In the actual operation of textile machinery, textile machinery is usually complex in structure, involving a variety of moving parts and control systems, and highly sensitive data has the ability to characterize a variety of textile machine faults. Highly sensitive data can quickly identify the occurrence of faults in textile machines. Although it contains potential fault information, it is easy to be misjudged as noise because the data changes rapidly. At this time, if a larger local window size is used, a large amount of real fault detail information will be lost, which will greatly affect the recognition accuracy of subsequent textile machine faults, making it difficult to perform maintenance in time, and shortening the service life of the textile machine.

[0066] Different distribution characteristics of highly sensitive data may contain different potential fault information. The operating parameters of textile machines in historical fault cycles contain a large amount of detailed information on textile machine faults, which can effectively characterize the fault mode of textile machines.

[0067] Therefore, the clustering algorithm is used to analyze the distribution characteristics of each operating parameter in all historical fault cycles to reflect the fault types of textile machines, specifically:

[0068] Calculate the mean and standard deviation of each operating parameter at all times in each historical fault cycle;

[0069] The mean and the standard deviation of each operating parameter in each historical fault cycle are combined into a two-dimensional data group; a clustering analysis algorithm is used to cluster the two-dimensional data group of each operating parameter in all historical fault cycles to obtain a plurality of cluster clusters;

[0070] Preferably, in this embodiment, AP (Affinity Propagation) clustering algorithm is used for clustering, wherein the damping coefficient in the AP clustering algorithm is set to 0.5. The AP clustering algorithm is a well-known technology and will not be described in detail here. As other implementation methods, implementers can adopt other methods of the prior art, such as hierarchical clustering algorithms, etc. This embodiment does not impose any special restrictions on this.

[0071] Furthermore, the clustering diagram of the textile machine operating parameters provided in the embodiment of the present application is as follows: Figure 3 As shown, the horizontal axis is the mean value and the vertical axis is the standard deviation.

[0072] It should be noted that each cluster represents a fault type, that is, it represents a textile machine failure mode.

[0073] Furthermore, based on each cluster of each operating parameter, the fault characterization strength is determined to reflect the characterization strength of each operating parameter for each fault type, specifically:

[0074] Calculate the mean of the distances between all elements in each cluster and its cluster center corresponding to each operating parameter, and record it as the first mean;

[0075] The ratio of the number of all elements in each cluster corresponding to each operating parameter to the first mean value is used as the fault characterization strength of each cluster corresponding to each operating parameter;

[0076] In this embodiment, the calculation method of the fault characterization strength of each cluster corresponding to each operating parameter is as follows: ,in, For the The operation parameters corresponding to The fault characterization strength of the clusters, For the The operation parameters corresponding to The number of all elements in a cluster, For the The operation parameters corresponding to The mean of the distances between all elements in a cluster and its cluster center is the first mean.

[0077] Further, the flowchart of the steps of the method for obtaining the fault characterization strength of each cluster corresponding to each operating parameter provided in the embodiment of the present application is as follows: Figure 4 shown.

[0078] It should be noted that when The operation parameters corresponding to The higher the density of the elements in a cluster, the higher the first mean The smaller the The operating parameters contain The more reliable the fault details of the textile machine fault mode are, the better. The more elements there are in a cluster, the The larger the The more the operating parameters can characterize the The more textile machine failure modes, the more The operating parameters are identified Failure modes of textile machines and the resulting fault characterization strength The larger the The operation parameters of The higher the characterization intensity of the failure mode of a textile machine.

[0079] Furthermore, the stronger the data sensitivity of all operating parameters is, the more capable it is of characterizing various textile machine fault modes, that is, the more capable it is of characterizing various fault types that occur in textile machines, thereby having a higher fault characterization strength for each cluster. Therefore, based on the fault characterization strength, the fault mode sensitivity is determined, specifically:

[0080] Calculating the cumulative sum of the fault characterization strengths of all clusters corresponding to each operating parameter, and taking the normalized result of the product of the number of all clusters corresponding to each operating parameter and the cumulative sum as the fault mode sensitivity of each operating parameter;

[0081] In this embodiment, the calculation formula of the failure mode sensitivity of each operating parameter is: ,in, For the The failure mode sensitivity of the operating parameters, For the The number of all clusters corresponding to the running parameters of the item, For the The operation parameters corresponding to The fault characterization strength of the clusters, is a normalization function. In this embodiment, a sigmoid function is used for normalization. The sigmoid function is a well-known technology and will not be described in detail here. As other implementation methods, the implementer can adopt other methods of the prior art, such as the tanh function, etc. This embodiment does not impose any special restrictions on this.

[0082] It should be noted that the fault characterization intensity The larger the The operating parameters contain more The fault details of the textile machine fault mode, that is, when the textile machine The more the failure mode of the textile machine is, the more The operation parameters are identified. The larger the The operating parameters are applicable to various textile machine failure modes, namely The stronger the sensitivity of the operating parameters, the greater the sensitivity of the resulting failure mode, which means that The operating parameters are more sensitive to the occurrence of textile machine failures.

[0083] At this point, the failure mode sensitivity of each operating parameter is obtained.

[0084] Step 4: Based on the noise interference deviation and the fault mode sensitivity, determine the local window length of the filtering algorithm for each operating parameter at each moment in the real-time monitoring period; based on the local window length, use the filtering algorithm to obtain the smoothed data of each operating parameter at all moments in the real-time monitoring period, combine with the neural network model, obtain the probability corresponding to each fault type, and diagnose the fault of the textile machine.

[0085] Further, based on the noise interference deviation and the fault mode sensitivity, the local window length of the Lowess algorithm is determined to smooth and denoise the data, specifically:

[0086] The calculation formula for the local window length of each operating parameter at each moment in the real-time monitoring cycle is: ,in, For the The operating parameters are the first The local window length at time, To preset the initial window length, For the The operating parameters are in the real-time monitoring cycle. The noise interference deviation at the moment, For the Failure mode sensitivity of an operating parameter.

[0087] Preferably, in this embodiment, the initial window length is preset The value is ,in, is the number of all moments in the acquisition cycle. is 3000, so the initial window length is preset The value is 150. As other implementation methods, the implementer can set it according to the actual situation.

[0088] It should be noted that the noise interference deviation The smaller the value, the closer the data of the corresponding operating parameter at the corresponding time is to the actual value, and the less affected by noise interference. At this time, there is no need to use a larger window length for data smoothing and noise elimination. If the fault mode sensitivity The larger the The operating parameters may contain more detailed information about the fault of the textile machine. In order to prevent the loss of detailed information about the fault, the filtering algorithm is used to filter the first When smoothing and denoising the data of the operating parameters, the local window length should be small. The smaller.

[0089] Based on the local window length of each operating parameter at each moment, the Lowess algorithm is used to smooth and denoise the data of each operating parameter at each moment in the real-time monitoring cycle, and obtain the smoothed data of each operating parameter at all moments in the real-time monitoring cycle;

[0090] Preferably, in this embodiment, the Lowess algorithm is used to smooth the data, wherein the Bisquare kernel function is used as the weight function of the Lowess algorithm; secondly, the Lowess algorithm is a well-known technology and will not be described in detail here. As other implementation methods, the implementer can adopt other methods of the prior art, such as Savitzky-Golay filter, etc., and this embodiment does not impose any special restrictions on this.

[0091] Furthermore, the schematic diagram of the smoothing effect of the Lowess algorithm before and after the improvement provided in this embodiment is as follows: Figure 5 and Figure 6 As shown, the horizontal axis is time and the vertical axis is yarn speed. Figure 5 This is a schematic diagram of the smoothing effect of the Lowess algorithm before improvement. A fixed local window length is used to smooth the data. Figure 6 This is a schematic diagram of the smoothing effect of the improved Lowess algorithm. Smoothing is performed using the local window length calculated in real time. Figure 5 , Figure 6 By using the method provided in this embodiment, the smoothed data is more consistent with the changes in the original data, and the smoothing effect is relatively better.

[0092] The operating parameters at all times in all historical fault cycles are taken as data sets, and the fault type corresponding to each fault cycle is taken as the data set label;

[0093] In this embodiment, 60% of the data in the data set is used as a training set of the neural network model, 20% of the data is used as a test set of the neural network model, and 20% of the data is used as a validation set of the neural network model, so as to train the neural network model and obtain a trained neural network model;

[0094] Preferably, in this embodiment, a recurrent neural network model is used for model training, wherein a stochastic gradient descent algorithm (SGD) is used as the optimizer of the RNN recurrent neural network, and cross entropy is used as the loss function of the RNN recurrent neural network. The recurrent neural network model is a well-known technology and will not be described in detail here. As other implementation methods, the implementer can adopt other methods of the prior art, such as a convolutional neural network model, etc. This embodiment does not impose any special restrictions on this.

[0095] The smoothed data of each operating parameter at all times during the real-time monitoring period is used as the input of the trained neural network model to obtain the probability of each fault type;

[0096] If the probability is greater than the preset threshold, the textile machine has a fault, and the fault type corresponding to the probability greater than the preset threshold is repaired; otherwise, the textile machine does not have a fault, and the fault diagnosis of the textile machine is completed to improve the recognition effect of multiple faults that may occur in the textile machine.

[0097] Preferably, in this embodiment, the preset threshold value is 0.3. As other implementation modes, the implementer can set it according to the actual situation.

[0098] Based on the same inventive concept as the above method, an embodiment of the present application also provides a textile machine working fault status monitoring system, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the steps of any one of the above-mentioned textile machine working fault data analysis methods are implemented.

[0099] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0100] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0101] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the present application. It should be pointed out that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application without departing from the content of the technical solution of the present application, shall fall within the protection scope of the technical solution of the present application.

Claims

1. A method for analyzing textile machine working fault data, characterized in that: The method comprises the following steps: Obtain the data of each operating parameter of the textile machine at each moment in the real-time monitoring cycle, and the data of each operating parameter at each moment in each historical fault cycle when each fault occurs, as well as the fault type corresponding to each historical fault cycle; Determine the abnormal score of each operating parameter at each moment in the real-time monitoring cycle according to the mutation characteristics and change trends of each operating parameter at different moments in the real-time monitoring cycle, as well as the abnormal changes; analyze the deviation of the abnormal score to determine the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle; Cluster all historical fault cycles of each operating parameter according to the discreteness of the data at all times in each historical fault cycle of each operating parameter; analyze the number of all elements in each cluster corresponding to each operating parameter and the density of the element distribution to determine the fault characterization strength of each cluster corresponding to each operating parameter; determine the fault mode sensitivity of each operating parameter based on the fault characterization strength; Based on the noise interference deviation and the fault mode sensitivity, the local window length of the filtering algorithm of each operating parameter at each moment in the real-time monitoring cycle is determined; based on the local window length, the filtering algorithm is used to obtain the smoothed data of each operating parameter at all moments in the real-time monitoring cycle, and the probability corresponding to each fault type is obtained in combination with the neural network model to diagnose the fault of the textile machine; Each of the operating parameters includes yarn tension, yarn speed, horizontal vibration data, vertical vibration data, spindle speed, voltage, current, humidity, and temperature; Determining the abnormal score of each operating parameter at each time includes: Perform mutation point detection on the data of each operating parameter at all times in the real-time monitoring cycle to obtain the mutation probability of each operating parameter at each time in the real-time monitoring cycle; The data of each operating parameter at each time in the real-time monitoring cycle and multiple times in its neighborhood are combined into a local monitoring sequence; Using Sen slope estimation algorithm to obtain the trend slope of the local monitoring sequence; The mutation probability and the trend slope are combined into a two-dimensional array, and anomaly detection is performed on the two-dimensional array at all times in the real-time monitoring period for each operating parameter to obtain an anomaly score for each operating parameter at each time in the real-time monitoring period; The calculation formula for determining the noise interference deviation of each operating parameter at each moment in the real-time monitoring cycle is: ,in, For the The operating parameters are the first The noise interference deviation at the moment, For the The operating parameters are the first The abnormal score at the time, For the The average of the abnormal scores of the operating parameters at all times in the real-time monitoring period, is the preset initial interference coefficient; The local window length of the filtering algorithm for each operating parameter at each moment in the real-time monitoring cycle is determined by the following calculation formula: ,in, For the The operating parameters are the first The local window length of the moment filtering algorithm, To preset the initial window length, For the The operating parameters are the first The noise interference deviation at the moment, For the Failure mode sensitivity of an operating parameter.

2. A textile machine working fault data analysis method as claimed in claim 1, characterized in that: The clustering of all historical fault cycles of each operating parameter includes: The mean and standard deviation of the data of each operating parameter at all times in each historical fault cycle are formed into a two-dimensional array; The two-dimensional array of each operating parameter in all historical fault cycles is clustered to obtain multiple clusters.

3. A textile machine working fault data analysis method as claimed in claim 1, characterized in that: The determining of the fault characterization strength of each cluster corresponding to each operating parameter includes: Calculate the mean of the distances between all elements in each cluster and its cluster center corresponding to each operating parameter, and record it as the first mean; The ratio of the number of all elements in each cluster corresponding to each operating parameter to the first mean value is used as the fault characterization strength of each cluster corresponding to each operating parameter.

4. A textile machine working fault data analysis method as claimed in claim 1, characterized in that: The failure mode sensitivity of each operating parameter is determined by the following calculation formula: ,in, For the Failure mode sensitivity of the operating parameters, For the The number of all clusters corresponding to the running parameters of the item, For the The operation parameters corresponding to The fault characterization strength of the clusters, is the normalization function.

5. A textile machine working fault data analysis method as claimed in claim 1, characterized in that: The diagnosing the fault of the textile machine includes: if the probability is greater than a preset threshold, the textile machine has a fault; otherwise, the textile machine does not have a fault.

6. A textile machine working fault status monitoring system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of a textile machine working fault data analysis method as described in any one of claims 1-5 are implemented.

Images