Rapier loom fault diagnosis system
By combining the fault tree model optimized by the whale algorithm with the probabilistic neural network, the problem of insufficient global perspective in the fault diagnosis system of rapier looms is solved, realizing real-time, complete and fast fault identification and location, and improving the accuracy and efficiency of diagnosis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEBEI UNIV OF TECH
- Filing Date
- 2026-02-04
- Publication Date
- 2026-05-12
AI Technical Summary
Existing rapier loom fault diagnosis systems lack a global perspective, resulting in low diagnostic efficiency and failing to meet the need for rapid and accurate diagnosis in complex control environments.
The whale algorithm is used to optimize the parameters of the fault tree model. Combined with a probabilistic neural network, a fault symptom dataset is constructed and normalized to achieve real-time fault identification and location.
It enables real-time, complete, and rapid fault identification and location for rapier looms, improving the accuracy and efficiency of diagnosis.
Smart Images
Figure CN122020481A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault diagnosis technology, and in particular to a fault diagnosis system for rapier looms. Background Technology
[0002] With the continuous improvement of the intelligence level of rapier looms, the equipment control systems are becoming increasingly complex. To adapt to the user needs of the new era of rapier loom development, ensuring the safety and reliability of rapier loom operation has become a current research hotspot. Due to the complex mechanical structure and weaving process of rapier looms, the probability of failure is relatively high. Once a failure occurs and causes downtime, it can result in economic losses at best, and endanger the lives of production workers at worst. Therefore, effective fault diagnosis is an important measure to ensure the normal operation of rapier looms, and a loom fault diagnosis system is of paramount importance to the reliable operation of the loom.
[0003] Existing research largely focuses on diagnosing localized faults in rapier looms, lacking a comprehensive diagnostic system for the entire equipment. This localized approach leads to insufficient understanding of the overall health of the equipment, hindering efficient overall fault early warning and maintenance. Due to the lack of a global perspective, the diagnostic process often requires checking multiple potential fault points one by one, a time-consuming and inefficient method. Furthermore, existing fault diagnosis systems for highly automated control systems are insufficient in terms of intelligence, failing to meet the demands for rapid and accurate diagnosis in complex control environments.
[0004] Therefore, this application proposes a fault diagnosis system for rapier looms. Summary of the Invention
[0005] In order to overcome the shortcomings of the prior art and solve the technical problems existing in the background art, the present invention proposes a rapier loom fault diagnosis system.
[0006] This invention is achieved through the following technical solution: A fault diagnosis system for rapier looms, the system integrating fault tree analysis based on whale algorithm optimization and probabilistic neural network, and the fault diagnosis steps are as follows: Step 1: Establish a fault tree diagnostic model for the rapier loom: Based on the fault mechanism and process flow of rapier looms, the fault tree model is divided into fault tree models of the start-up inspection subsystem, the weaving subsystem, and the warp feed and take-up subsystem. Qualitative and quantitative diagnosis is performed through Boolean operations and bottom event probability importance analysis.
[0007] Step 2: Optimize the fault tree model parameters using the whale algorithm: By optimizing the rule weights, prerequisite attribute weights, and output confidence of fault symptoms using the whale algorithm, the accuracy of fault diagnosis and decision-making ability are improved. Step 3: Construct and normalize the sample dataset: The optimized fault tree model data and real-time monitoring data were integrated to generate a sample dataset containing 11 fault symptom features, and then normalized. Step 4: Probabilistic Neural Network Training and Fault Diagnosis The normalized dataset is input into a probabilistic neural network for training, which outputs the probability of fault symptoms, calculates the importance of fault causes, and generates a suggested fault location and troubleshooting sequence.
[0008] Preferably, in step one, the qualitative analysis identifies the weak links in the system through minimal cut sets, and the quantitative analysis calculates the criticality of the cause of the failure through the importance of the probability of the bottom event.
[0009] Preferably, in step two, the whale algorithm optimizes the weight parameters of each subsystem in the fault tree model by simulating the random search and spiral mechanism of humpback whale predation behavior. The whale algorithm adopts the following process: Initialize the whale population; Calculate the fitness of each whale; Update the optimal individual whale; Update whale location; If P < 0.5 is satisfied, then determine whether A ≥ 1 is satisfied; if satisfied, search for and capture prey, and record the optimal target value. If P < 0.5 is not satisfied, surround the prey and record the optimal target value; Determine if the optimal objective value meets the termination condition. If it does, output the optimal objective value and terminate. If the condition is not met, update the whale's position and repeat this process until the termination condition is met and the optimal target value is output.
[0010] Preferably, the normalization process in step three uses a linear transformation method to map the sample feature values to the (0,1) interval, thereby eliminating the influence of dimensional differences on the classification results.
[0011] Preferably, the structure of the probabilistic neural network in step four includes an input layer with 11 neurons, a pattern layer with 1430 neurons, a summation layer with 11 neurons, and an output layer with 11 neurons, wherein the output layer determines the fault occurrence state through a probability threshold.
[0012] Preferably, the number of neurons in the pattern layer is the same as the number of training samples, each category layer corresponds to a type of fault symptom, and 100% classification and recognition accuracy is achieved through the training set.
[0013] Preferably, the system identifies abnormal states of the rapier loom in real time, and the abnormal states include abnormal oil level in the oil tank, excessive spindle speed, servo motor speed deviation, and excessive temperature.
[0014] Preferably, the importance of the fault causes is ranked by the probability values output by the probabilistic neural network, providing maintenance personnel with fault troubleshooting priority suggestions. During the diagnosis process, abnormal data samples that exceed the normal value range are automatically removed, and fault causes with high probability importance are given priority.
[0015] Preferably, the sample dataset includes 160 sets of data, of which the training set has 1430 samples and the test set has 330 samples. The fault symptom classification and recognition accuracy of the test set is 99.39%.
[0016] The beneficial effects of this invention are: This invention is based on a diagnostic model optimized by the fusion of probabilistic neural networks and fault tree algorithms. On the basis of the fault tree diagnostic model for rapier looms, it introduces real-time new status monitoring data from the monitoring system to construct a sample dataset of rapier looms composed of a fault tree fault symptom dataset and a new status dataset.
[0017] The dataset is then converted into sample feature vectors, normalized, and imported into a probabilistic neural network algorithm model for training, calculating the probability of output fault symptoms. Based on this, the importance of the fault cause corresponding to the fault symptom is calculated, and fault diagnosis results and suggestions for troubleshooting order are output, thereby realizing real-time, complete, and rapid fault identification and location in the rapier loom fault diagnosis system. Attached Figure Description
[0018] Figure 1 This is a flowchart of the whale algorithm of the present invention; Figure 2 This is a schematic diagram of the fusion algorithm model of the present invention; Figure 3 This is a diagram of the probabilistic neural network structure for the training set of this invention; Figure 4 This is a training result diagram of the pattern classification and recognition training set of the present invention; Figure 5 This is a training result diagram of the pattern classification and recognition test set of the present invention. Detailed Implementation
[0019] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Experimental methods in the following embodiments, unless otherwise specified, are generally performed under conventional conditions or as recommended by the manufacturer.
[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of skill in the art. The reagents and raw materials used in this invention are readily available through conventional means, and unless otherwise specified, they shall be used in accordance with conventional methods in the art or as per the product instructions. Furthermore, any methods and materials similar to or equivalent to those described herein may be applied to the methods of this invention. The invention will now be further described with reference to the accompanying drawings and specific embodiments. The preferred embodiments and materials described herein are for illustrative purposes only.
[0021] A rapier loom fault diagnosis system, such as Figures 1-5 As shown: Step 1: Establish a fault tree diagnostic model for the rapier loom: Based on the failure mechanism and process flow of rapier looms, the fault tree model is divided into fault tree models of the start-up inspection subsystem, the weaving subsystem, and the warp feed and take-up subsystem. Qualitative and quantitative diagnosis is performed through Boolean operations and bottom event probability importance analysis. In step one, the qualitative analysis identifies the weak links in the system through minimal cut sets, while the quantitative analysis calculates the criticality of the cause of the failure through the importance of the probability of the bottom event.
[0022] A fault tree model for rapier looms was constructed by combining the fault mechanism and fault tree algorithm structure. According to the weaving process of rapier looms, the fault tree model of rapier looms was divided into three subsystem fault tree models, namely the fault tree model of startup inspection subsystem, the fault tree model of weaving subsystem, and the fault tree model of warp feed and take-up subsystem.
[0023] Then, the established fault tree model of the rapier loom system was verified by a fault diagnosis example. Boolean operations were used to perform a qualitative analysis of the fault tree diagnosis model of the rapier loom, and the minimum cut set of the fault tree diagnosis system was obtained. The weak links of the system were quantitatively analyzed by using the probability importance of the bottom event, and the weak links of the rapier loom system were identified. It was concluded that the fault tree diagnosis model of the rapier loom improved the efficiency of fault location and troubleshooting.
[0024] Step 2: Optimize the fault tree model parameters using the whale algorithm: By optimizing the rule weights, prerequisite attribute weights, and output confidence of fault symptoms using the whale algorithm, the accuracy of fault diagnosis and decision-making ability are improved. In step two, the whale algorithm optimizes the weight parameters of each subsystem in the fault tree model by simulating the random search and spiral mechanism of humpback whale predation behavior. The whale algorithm adopts the following process: Initialize the whale population; Calculate the fitness of each whale; Update the optimal individual whale; Update whale location; If P < 0.5 is satisfied, then determine whether A ≥ 1 is satisfied; if satisfied, search for and capture prey, and record the optimal target value. If P < 0.5 is not satisfied, surround the prey and record the optimal target value; Determine if the optimal objective value meets the termination condition. If it does, output the optimal objective value and terminate. If the condition is not met, update the whale's position, and repeat this process until the termination condition is met and the optimal target value is output. To further optimize the parameters of fault symptoms in each subsystem and improve the assessment accuracy, the whale algorithm will be used to optimize the fault diagnosis method for each subsystem of the rapier loom based on fault tree analysis. The optimized model will have better accuracy and reliability in fault diagnosis.
[0025] The core idea of the Whale Optimization Algorithm (WOA) is to mimic the unique predation behavior of humpback whales using random or optimal search, and to introduce a spiral mechanism to simulate the whale's attack behavior using bubble nets. This algorithm's advantages lie in its simple and clear mechanism, relatively few parameters, and excellent global search capability. Given these significant advantages, this application chose WOA to optimize the rule weights, premise attribute weights, and output confidence of loom fault symptoms, thereby further improving the accuracy of fault reasoning and the decision-making ability to determine fault symptoms. Figure 1 As shown.
[0026] Step 3: Construct and normalize the sample dataset: The optimized fault tree model data and real-time monitoring data were integrated to generate a sample dataset containing 11 fault symptom features, and then normalized. In step three, the normalization process uses a linear transformation method to map the sample feature values to the (0,1) interval, eliminating the influence of dimensional differences on the classification results.
[0027] Step 4: Probabilistic Neural Network Training and Fault Diagnosis The normalized dataset is input into a probabilistic neural network for training, which outputs the probability of fault symptoms, calculates the importance of fault causes, and generates a suggested fault location and troubleshooting sequence.
[0028] The system identifies abnormal states of the rapier loom in real time, including abnormal oil level in the oil tank, excessive spindle speed, servo motor speed deviation, and excessive temperature.
[0029] Preferably, the importance of the fault causes is ranked by the probability values output by the probabilistic neural network, providing maintenance personnel with fault troubleshooting priority suggestions. During the diagnosis process, abnormal data samples that exceed the normal value range are automatically removed, and fault causes with high probability importance are given priority.
[0030] Based on the fault tree diagnostic model of rapier loom optimized by the whale algorithm, real-time new status monitoring data from the monitoring system is introduced to construct a rapier loom sample dataset composed of a fault tree fault symptom dataset optimized by the whale algorithm and a new status dataset.
[0031] The dataset is then transformed into sample feature vectors, normalized, and imported into a probabilistic neural network algorithm model for training, calculating the probability of output fault symptoms. Based on this, the importance of the fault cause corresponding to the fault symptom is calculated, and fault diagnosis results and suggestions for troubleshooting order are output, thereby realizing real-time, complete, and rapid fault identification and location in the rapier loom fault diagnosis system. Figure 2 As shown.
[0032] Specifically: Eleven fault symptoms of rapier looms were selected from the fault symptom set as fault diagnosis sample symptom types for rapier looms, and grouped into h1~h11. There are 34 fault causes corresponding to the above fault symptoms. A list of some fault symptoms and fault causes of the selected rapier loom fault diagnosis system is shown in Table 1 and Table 2.
[0033] Table 1: Fault Symptoms of Some Rapier Loom Fault Diagnosis Systems
[0034] Table 2: Causes of Fault Diagnosis Systems for Some Rapier Looms
[0035] To establish an effective algorithm simulation example for verification, a sample dataset of 160 groups containing 1760 data points was built using fault symptom signals collected from the rapier loom control system. Each sample group contains 11 features that can reflect the equipment status of the rapier loom at any specific time. The normal value range of the fault symptom samples is shown in Table 3, and their values are expressed by the output signals of the sensors.
[0036] Table 3: Normal Value Range of Fault Symptoms
[0037] A sample dataset of some rapier looms is shown in Table 4.
[0038] Table 4: Sample Dataset of Partial Rapier Looms
[0039] After optimization using the Whale Algorithm and normalization, the influence of differences in feature values among samples on the classification results is effectively eliminated. The sample dataset contains 160 samples, each containing 11 elements. The dataset samples are arranged into a sample matrix with 160 rows and 11 columns. Normalization is then performed to obtain the normalized sample dataset for rapier looms. Partial normalized sample datasets are shown in Table 5. The value of each element in the normalized samples ranges from 0 to 1.
[0040] Table 5: Partially Normalized Rapier Loom Sample Dataset
[0041] Simulation verification was performed based on the diagnostic model optimized by the whale algorithm. After normalizing the sample dataset, the third step, training using a probabilistic neural network, was initiated.
[0042] Due to algorithm requirements, the rapier loom sample dataset needs to be divided into two parts: a training sample dataset and a test sample dataset. Therefore, 1430 samples were randomly selected from the sample dataset as training samples, and the remaining 330 samples were used as test samples.
[0043] The normalized sample matrix of the training set is a 130x11 matrix. The input layer of the probabilistic neural network for rapier loom fault diagnosis training set has the same number of fault symptoms as the number of input neurons, with 11 neurons. To ensure the probabilistic neural network can correctly locate fault symptoms, the training categories are classified according to the number of fault symptoms. That is, the pattern layer has 11 category layers, each with 130 neurons corresponding to the number of sample groups in the training set, for a total of 1430 neurons in the pattern layer, consistent with the number of sample training sets. The summation layer calculates the probability of each fault symptom, with 11 neurons. The output layer outputs the calculated probability of the fault symptom; if the probability is greater than 50%, it outputs 1, indicating that the fault symptom has occurred; otherwise, it outputs 0. In summary, the training structure of the training sample dataset and the probabilistic neural network structure for the rapier loom fault diagnosis training set are obtained, as follows: Figure 3 As shown.
[0044] The training results of pattern classification and recognition on the training sample dataset are as follows: Figure 3 As shown, each set of fault symptoms has 130 samples, and there are a total of 1430 datasets for 11 sets of fault symptoms.
[0045] from Figure 2As can be seen, the predicted categories of the 1430 training sets are completely consistent with the true categories, with zero training error, indicating that the training effect is quite good. This proves that the probabilistic neural network structure for rapier loom fault diagnosis training sets can achieve 100% accurate classification and recognition of fault symptoms. Therefore, based on this network structure, fault symptom classification and recognition can be performed on the sample test set to verify the network's classification and recognition effect. Furthermore, the network structure can be further improved and optimized based on the training results.
[0046] The normalized sample matrix of the test set is a 30x11 matrix. In the probabilistic neural network structure of the test set, the input layer has 11 neurons, the pattern layer has 330 neurons, the summation layer has 11 neurons, and the output layer has 11 neurons. The training effect diagram for pattern classification and recognition on the sample test set is shown below. Figure 4 As shown.
[0047] Training results for pattern classification and recognition on the test set (Figure) Figure 5 As can be seen, the correct classification and recognition rate of fault symptoms in the test set is 99.39%, which basically achieves accurate classification of fault symptoms, indicating that the algorithm's training results are quite good. Analyzing the erroneous samples in the training results of the test set, fault symptom 5 was classified as fault symptom 6, and fault symptom 7 was classified as fault symptom 8. By referring to Table 1, which lists some fault symptoms of the rapier loom fault diagnosis system, we know that fault symptom 5 is warp tension value, fault symptom 6 is weft tension value, fault symptom 7 is warp feed servo motor speed, and fault symptom 8 is take-up servo motor speed. The fault diagnosis signals of the two sets of fault symptoms are similar; therefore, the erroneous diagnosis result can be considered to be within the allowable error range. The 13th row of the fault symptom matrix is selected for diagnosis, resulting in a result vector matrix. This indicates that fault symptom 11 has occurred, meaning the main spindle encoder is malfunctioning. The fault diagnosis result is that the fault symptom is a main encoder malfunction. The corresponding causes are a loose main encoder coupling, incorrect main encoder wiring, and a fault in the main encoder's internal circuitry. Quantitative calculations determine the probability and importance of each cause, and the recommended troubleshooting order is: check the main encoder's internal circuitry, check the main encoder coupling, and check the main encoder wiring. The output fault diagnosis results are shown in Table 6.
[0048] Table 6: Fault Diagnosis Results
[0049] Two misclassified sample datasets were removed from the test set. Diagnostic analysis was performed on the remaining 328 test sample datasets. Among them, the data of 6 sample datasets were outside the normal value range, while the data of the remaining 322 sample datasets were all within the normal value range. The diagnostic algorithm can accurately identify the fault symptoms of the problematic samples in the 6 sample datasets. The vector matrices of the remaining 322 datasets are all composed of 0.
[0050] Any aspects not covered in this invention are applicable to existing technologies.
Claims
1. A fault diagnosis system for rapier looms, characterized in that, The system integrates fault tree analysis based on the whale algorithm and probabilistic neural network, and the fault diagnosis steps are as follows: Step 1: Establish a fault tree diagnostic model for the rapier loom: Based on the failure mechanism and process flow of rapier looms, the fault tree model is divided into fault tree models of the start-up inspection subsystem, the weaving subsystem, and the warp feed and take-up subsystem. Qualitative and quantitative diagnosis is performed through Boolean operations and bottom event probability importance analysis. Step 2: Optimize the fault tree model parameters using the whale algorithm: By optimizing the rule weights, prerequisite attribute weights, and output confidence of fault symptoms using the whale algorithm, the accuracy of fault diagnosis and decision-making ability are improved. Step 3: Construct and normalize the sample dataset: The optimized fault tree model data and real-time monitoring data were integrated to generate a sample dataset containing 11 fault symptom features, and then normalized. Step 4: Probabilistic Neural Network Training and Fault Diagnosis The normalized dataset is input into a probabilistic neural network for training, which outputs the probability of fault symptoms, calculates the importance of fault causes, and generates a suggested fault location and troubleshooting sequence.
2. The rapier loom fault diagnosis system according to claim 1, characterized in that, In step one, the qualitative analysis identifies the weak links in the system through minimal cut sets, while the quantitative analysis calculates the criticality of the cause of the failure through the importance of the probability of the bottom event.
3. The rapier loom fault diagnosis system according to claim 1, characterized in that, In step two, the whale algorithm optimizes the weight parameters of each subsystem in the fault tree model by simulating the random search and spiral mechanism of humpback whale predation behavior. The whale algorithm adopts the following process: Initialize the whale population; Calculate the fitness of each whale; Update the optimal individual whale; Update whale location; If P < 0.5 is satisfied, then determine whether A ≥ 1 is satisfied; if satisfied, search for and capture prey, and record the optimal target value. If P < 0.5 is not satisfied, surround the prey and record the optimal target value; Determine if the optimal objective value meets the termination condition. If it does, output the optimal objective value and terminate. If the condition is not met, update the whale's position and repeat this process until the termination condition is met and the optimal target value is output.
4. The rapier loom fault diagnosis system according to claim 1, characterized in that, In step three, the normalization process uses a linear transformation method to map the sample feature values to the (0,1) interval, eliminating the influence of dimensional differences on the classification results.
5. The rapier loom fault diagnosis system according to claim 1, characterized in that, In step four, the probabilistic neural network structure includes an input layer with 11 neurons, a pattern layer with 1430 neurons, a summation layer with 11 neurons, and an output layer with 11 neurons. The output layer determines the fault occurrence state through a probability threshold.
6. The rapier loom fault diagnosis system according to claim 5, characterized in that, The number of neurons in the pattern layer is the same as the number of training samples. Each category layer corresponds to a type of fault symptom, and 100% classification accuracy is achieved through the training set.
7. The rapier loom fault diagnosis system according to claim 1, characterized in that, The system identifies abnormal states of the rapier loom in real time, including abnormal oil level in the oil tank, excessive spindle speed, servo motor speed deviation, and excessive temperature.
8. A rapier loom fault diagnosis system according to claim 1, characterized in that, The importance of the fault causes is ranked by the probability values output by the probabilistic neural network, providing maintenance personnel with suggestions for fault troubleshooting priorities. During the diagnosis process, abnormal data samples that exceed the normal value range are automatically removed, and fault causes with high probability importance are given priority.
9. A rapier loom fault diagnosis system according to claim 1, characterized in that, The sample dataset includes 160 sets of data, of which the training set has 1430 samples and the test set has 330 samples. The accuracy rate of fault symptom classification and recognition on the test set is 99.39%.