Fault diagnosis and prediction system for motor rotor die-casting die

Abstract

The invention provides a fault diagnosis and prediction system for a motor rotor die-casting die, and relates to the technical field of adaptive control, and the system comprises a multi-source signal monitoring module which is used for obtaining a multi-source monitoring signal set; the fault diagnosis module is used for generating a fault signal set; the model training module is used for constructing a fault graph network of the die-casting die; the fault propagation point identification module is used for outputting a first type of propagation fault nodes and a second type of propagation fault nodes; the fault prediction module is used for acquiring a first type of fault prediction indexes and a second type of fault prediction indexes; and the fault prediction result output module is used for outputting a fault prediction result of the die-casting die. According to the invention, the technical problems that the fault diagnosis and prediction are inaccurate and the fault propagation path cannot be effectively identified due to the single data source can be solved, and the technical effects of accurately identifying the fault propagation point and improving the fault diagnosis accuracy and prediction precision through early warning of the fault are achieved.

Description

Technical Field

[0001] The present application relates to the technical field of fault diagnosis, and in particular to a fault diagnosis and prediction system for a motor rotor die-casting mold. Background Art

[0002] With the rapid development of modern manufacturing industry and increasingly fierce market competition, motor rotor die-casting molds are key equipment in the motor manufacturing process. The stability and reliability of their performance are crucial to ensuring product quality and improving production efficiency. However, in the actual production process, motor rotor die-casting molds are prone to various faults such as wear, deformation, cracks, etc. due to various factors such as long-term operation, material fatigue, improper operation, etc. These faults not only affect the service life of the mold, but may also pose a threat to production safety. Traditional motor rotor die-casting mold fault diagnosis and prediction methods have low fault diagnosis accuracy, limited prediction ability, and cannot effectively identify the propagation path of the fault due to the single data source and complex and changeable fault characteristics. This limitation not only increases the uncertainty and risk in the production process, but also affects production efficiency. Summary of the invention

[0003] The purpose of this application is to provide a fault diagnosis and prediction system for a motor rotor die-casting mold, so as to solve the technical problems of inaccurate fault diagnosis and prediction and inability to effectively identify the fault propagation path due to a single data source.

[0004] In view of the above problems, the present application provides a fault diagnosis and prediction system for a motor rotor die-casting mold.

[0005] The present application provides a fault diagnosis and prediction system for a motor rotor die-casting mold, the system comprising: a multi-source signal monitoring module, used to perform multi-source signal monitoring on the die-casting mold according to a multi-source data acquisition device, and obtain a multi-source monitoring signal set; a fault diagnosis module, used to perform fault diagnosis on the multi-source monitoring signal set, and generate a fault signal set; a model training module, used to train a graph neural network model, the graph neural network model constructs a fault graph network of the die-casting mold according to the fault signal set; a fault propagation point identification module, used to perform fault propagation point identification on the fault graph network, and output a first type of propagation fault node and a second type of propagation fault node, wherein the first type of propagation fault node is a structural fault propagated along a physical connection structure, and the second type of propagation fault node is a functional fault propagated by affecting the entire die-casting mold; a fault prediction module, used to perform fault prediction on the first type of propagation fault node and the second type of propagation fault node, and obtain a first type of fault prediction index and a second type of fault prediction index; a fault prediction result output module, used to output the fault prediction result of the die-casting mold according to the first type of fault prediction index and the second type of fault prediction index.

[0006] The technical solution provided in this application has at least the following technical effects or advantages: The above-mentioned fault diagnosis and prediction system for the motor rotor die-casting mold comprehensively monitors the die-casting mold through a multi-source data acquisition device, collects various signals including temperature, pressure, vibration, etc., and forms a multi-source monitoring signal set. Subsequently, these signal sets are fault diagnosed, the existing faults are identified, and these faults are converted into fault signal sets. After that, these fault signals are analyzed using a fault graph network. The network is constructed by training a graph neural network model, which can clearly show the propagation path of the fault in the mold. In the fault graph network, the first type of propagation fault nodes and the second type of propagation fault nodes can be identified. By predicting and analyzing these two types of fault nodes, corresponding fault prediction indicators can be obtained. Then, based on these prediction indicators, the fault prediction results of the die-casting mold are output. This result can help operators discover potential faults in advance and formulate corresponding maintenance plans, thereby ensuring the normal operation of the die-casting mold and the smooth progress of production.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0009] Figure 1 This is a schematic diagram of the structure of a fault diagnosis and prediction system for a motor rotor die-casting mold in this application; Figure 2 A schematic diagram of a process flow for fault prediction optimization of a fault diagnosis and prediction system for a motor rotor die-casting mold in the present application.

[0010] Description of reference numerals: Multi-source signal monitoring module 1, fault diagnosis module 2, model training module 3, fault propagation point identification module 4, fault prediction module 5, fault prediction result output module 6. DETAILED DESCRIPTION

[0011] The present application provides a fault diagnosis and prediction system for a motor rotor die-casting mold, thereby solving the technical problems of inaccurate fault diagnosis and prediction and inability to effectively identify the fault propagation path due to a single data source, thereby achieving the technical effect of improving the accuracy of fault diagnosis and prediction precision through early warning of faults and accurate identification of fault propagation points.

[0012] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0013] Example Please see attached Figure 1 The present application provides a fault diagnosis and prediction system for a motor rotor die-casting mold, and the system specifically includes the following modules: The multi-source signal monitoring module 1 is used to perform multi-source signal monitoring on the die-casting mold according to the multi-source data acquisition device to obtain a multi-source monitoring signal set.

[0014] In the embodiment of the present application, inside the multi-source signal monitoring module 1, the system terminal calls a multi-source data acquisition device to monitor the die-casting mold. The multi-source data acquisition device collects various data about the operation of the die-casting mold from different angles and dimensions. These data include multi-source signals such as acoustic emission data, operating parameters, fault records and maintenance logs. Among them, the acoustic emission data is a tiny sound signal generated by the mold under extreme conditions such as high pressure and high temperature captured by the acoustic emission sensor. These sound signals contain key information such as stress changes inside the mold, material fatigue or crack extension. The operating parameters include key parameters such as the working pressure, temperature, and speed of the die-casting machine. These parameters reflect the workload and state of the mold during actual operation. The fault record is an important document that records various faults that occur during the operation of the mold and their occurrence time, cause, and treatment method, and can understand the fault type and frequency of the mold. The maintenance log is an important document that records the maintenance activities of the mold, including maintenance time, maintenance content, maintenance personnel, etc., and can understand the maintenance history and maintenance effect of the mold. These signals together constitute a multi-source monitoring signal set, which provides a rich information basis for subsequent fault diagnosis and prediction.

[0015] The fault diagnosis module 2 is used to perform fault diagnosis on the multi-source monitoring signal set to generate a fault signal set.

[0016] In one embodiment, in the fault diagnosis module 2, after collecting the multi-source monitoring signal set of the die-casting mold, the system terminal uses the healthy signal samples inside the multi-source fault diagnosis channel to analyze the obtained multi-source monitoring signal set to identify possible faults or abnormal conditions in the die-casting mold. By comparing the multi-source monitoring signal set with the healthy signal samples, it is possible to capture signals that deviate from the normal mode or threshold, which are regarded as potential fault indications. Once these abnormal signals are identified, the system terminal will classify and organize them into a fault signal set. The fault signal set is a direct output of the fault diagnosis process, which contains specific information about the die-casting mold fault, such as the fault type, fault location, fault degree, etc. This information is crucial for subsequent fault diagnosis and prevention.

[0017] Furthermore, the present application provides a method of performing fault diagnosis on the multi-source monitoring signal set to generate a fault signal set, including: A multi-source fault diagnosis channel is established, wherein each fault diagnosis channel stores a healthy signal sample, each fault diagnosis channel is used to receive a corresponding monitoring signal set for fault signal identification, and output a multi-source fault signal; the multi-source fault signal is fitted to generate a fault signal set.

[0018] Preferably, in order to more accurately identify the fault signal of the die-casting mold, the system terminal determines the signal type to be monitored, such as acoustic emission data, operating parameters, etc., according to the characteristics and monitoring requirements of the die-casting mold. Subsequently, an independent fault diagnosis channel is established for each type of monitoring signal. For example, an acoustic emission data channel, an operating parameter channel, etc. These channels are composed of an input layer, a data processing layer, and an output layer. Each data processing layer has a built-in data set, which stores the healthy signal samples corresponding to the channel. These healthy signal samples clarify the healthy range of the data to be processed by the channel, which is determined based on historical data and safety standards. After that, the system terminal will build the comparison algorithm into the data processing layer, and the comparison algorithm can compare the input multi-source monitoring signal set with the internal healthy signal samples to identify the deviation signals therein, which are fault signals. After the multi-source fault diagnosis channel is established, the system terminal inputs the multi-source monitoring signal set into the corresponding channel according to the type. In each fault diagnosis channel, the system terminal calls the comparison algorithm to analyze the received monitoring signal with the healthy signal sample to determine whether the monitoring signal is within the range of the healthy signal sample. If it exceeds the range of this healthy signal sample, the system terminal marks this monitoring signal as a deviation signal. By performing the same operation on all monitoring signals in the multi-source monitoring signal set, multiple deviation signals can be obtained. Then, the system terminal outputs these deviation signals as multi-source fault signals. These fault signals may include different types of fault information, such as abnormal pressure, excessive temperature, etc. Finally, the system terminal fits these multi-source fault signals from different channels. The fitting process is to integrate and correlate these scattered fault signals, and analyze whether multiple fault signals appear in a similar time, which may indicate that these faults are related. Combined with professional knowledge, these possible causal relationships are analyzed to determine the correlation between fault signals to form a comprehensive fault signal set. This fault signal set contains various types of fault information and also reflects the inherent connection and correlation between them, thereby providing the system terminal with a comprehensive understanding of the fault status of the die-casting mold.

[0019] The model training module 3 is used to train a graph neural network model, and the graph neural network model constructs a fault graph network of the die-casting mold based on the fault signal set.

[0020] In one embodiment, in the model training module 3, the system terminal defines a graph structure according to the physical structure and functional relationship of the die-casting mold. The nodes in the graph can represent different parts or components of the mold, and the edges represent the interactions or dependencies between these parts. Subsequently, the graph neural network model is constructed using the API and functions provided by the Deep Graph Library (DGL). The model architecture is then designed, including graph convolution layers, pooling layers, fully connected layers, etc. After that, the fault signal set is divided into a training set, a validation set, and a test set. Some training parameters and hyperparameters are then configured based on experience. These parameters include learning rate, batch size, number of iterations, optimizer, etc. Then, a training loop is written using the training functions and algorithms provided by the Deep Graph Library. In the training loop, the system terminal inputs the training set into the model, calculates the output and loss function value of the model. Then, the gradient is calculated based on the loss function value, and the parameters of the model are updated. After each iteration, the performance of the model is evaluated using the validation set, and the best model, that is, the model with the best performance on the validation set, is saved. The system terminal adjusts the parameters and hyperparameters of the model to optimize the performance of the model based on the performance indicators of the validation set, such as accuracy, recall, F1 score, etc., such as trying different network structures, changing the learning rate, adjusting the batch size, etc. After the training is completed, the system terminal uses the test set to evaluate the generalization ability of the model. The test data is input into the model, the performance indicators of the model are calculated, and compared with the performance on the training set and validation set. If the model shows good performance on the test set, the system terminal outputs this graph neural network model as a fault graph network.

[0021] The fault propagation point identification module 4 is used to identify the fault propagation points of the fault graph network and output the first type of propagation fault nodes and the second type of propagation fault nodes, wherein the first type of propagation fault nodes are structural faults that propagate along the physical connection structure, and the second type of propagation fault nodes are functional faults that propagate by affecting the entire die-casting mold.

[0022] In one embodiment, in the fault propagation point identification module 4, the system terminal extracts nodes from the acquired fault graph network to obtain multiple nodes of the fault graph network. Subsequently, the fault features corresponding to these nodes are collected, and then the fault propagation points of the nodes of the fault graph network are identified using these fault features, and the first type of propagation fault nodes and the second type of propagation fault nodes are output. Among them, the first type of propagation fault nodes are structural fault nodes, which refer to nodes where the fault propagates along the physical connection structure of the die-casting mold. In the graphical representation of the die-casting mold, these nodes represent the various physical components or components of the mold, and the connection between them reflects the actual physical connection relationship. When a node (i.e., a physical component) fails, the fault may propagate along these physical connections to other adjacent nodes, thereby triggering a chain reaction and causing a larger range of faults. By obtaining the first type of propagation fault nodes, the system terminal can identify the source of the fault. The second type of propagation fault nodes are functional fault nodes, which refer to nodes that propagate faults by affecting the function of the entire die-casting mold. These nodes may not directly correspond to a specific physical component in the graphical representation, but represent a function of the mold. When these functions fail, they may have a negative impact on the operation of the entire mold, causing the mold to malfunction or degrade in performance. Unlike structural failures, functional failures are often more difficult to predict and diagnose because they may involve complex interactions and mutual influences between multiple physical components. By obtaining the second type of propagation fault nodes, the system terminal can predict the development trend of the fault. These two types of propagation fault nodes are of great significance for the fault prediction and diagnosis of die-casting molds, and can help the system terminal locate the source of the fault more accurately, thereby improving the reliability and production efficiency of the mold.

[0023] Furthermore, the present application provides a method for identifying fault propagation points on the fault graph network and outputting first-type propagation fault nodes and second-type propagation fault nodes, including: Acquire N nodes of the fault graph network; collect the fault type, fault characteristics and impact range corresponding to each of the N nodes, and output N groups of fault characteristics corresponding to the N nodes; identify fault propagation points of the N nodes according to the N groups of fault characteristics, and output first-type propagation fault nodes and second-type propagation fault nodes.

[0024] Preferably, the system terminal obtains N nodes from the fault graph network, which represent different physical components or functional units in the die-casting mold. Subsequently, for these N nodes, the fault types, fault characteristics and impact ranges corresponding to them are collected. For the collection of fault characteristics, structural faults focus on vibration and stress-related characteristics extracted from acoustic emission signals. This is because structural faults are often closely related to physical properties such as vibration and stress distribution of physical components. By analyzing these characteristics, nodes that are prone to structural faults can be identified, such as faults caused by factors such as material fatigue, loose connections or uneven force. For functional faults, fault characteristics pay more attention to characteristics related to operating procedures and system responses. Functional faults are usually caused by factors such as improper operation, control system failures or incorrect parameter settings, which may affect the function and performance of the entire die-casting mold. Therefore, when collecting fault characteristics, it is necessary to pay attention to parameters and data related to operating procedures, system responses, etc., such as operating time, temperature changes, pressure fluctuations, etc. After obtaining the fault characteristics of N nodes, the system terminal identifies the fault propagation point. By aggregating the fault features of the corresponding neighboring nodes of each node, the system terminal can capture the impact of neighboring nodes on the propagation of central node faults. Afterwards, the distinguished aggregation results are transformed according to the feature transformation layer to determine the transformation features corresponding to these nodes. Then, the fault propagation points of these nodes are identified based on these transformation features to generate the first type of propagation fault nodes and the second type of propagation fault nodes. Through fault propagation point identification, the fault characteristics and propagation risks of each node in the die-casting mold can be understood more accurately, providing an important basis for subsequent fault prediction and diagnosis.

[0025] Furthermore, the present application provides for respectively identifying fault propagation points of the N nodes according to the N groups of fault characteristics, including: Obtain N groups of neighbor node sets corresponding to the N nodes; perform feature aggregation on the fault feature set corresponding to each group of neighbor node sets, and output N aggregated features corresponding to the N nodes.

[0026] Optionally, the system terminal obtains N sets of neighboring nodes of N nodes in the graph network. These neighboring nodes may interact with the central node through physical connections or functional associations. Subsequently, for each node's neighboring node set, the system terminal collects their corresponding fault features to form a fault feature set. These fault features may include vibration and stress features related to structural faults, or operating procedures and system response features related to functional faults. Afterwards, the aggregation feature formula is used to perform feature aggregation on the fault feature set of each set of neighboring node sets. Feature aggregation is the process of merging the features of multiple nodes into a unified representation, which can help the system terminal capture the impact of neighboring nodes on the propagation of central node faults. The aggregation feature formula is as follows: ;in, Represents the aggregated features of the i-th node, which is the updated node feature obtained after the aggregation operation. Represents the set of neighboring nodes of the i-th node, that is, the set of all nodes connected to node i. The feature vector representing the neighbor node j, that is, the original feature representation of node j. The system terminal calculates the aggregated feature by inputting each set of neighbor nodes obtained into the aggregated feature formula. Finally, the above process is performed on each of the N nodes to generate N aggregated features. Each aggregated feature corresponds to a node and contains the fault feature information of the node's neighbor node set, providing more valuable information for subsequent fault propagation point identification and analysis.

[0027] The N aggregate features are processed through a feature transformation layer, and N transformation features corresponding to the N nodes are output; fault propagation points of the N nodes are identified according to the N transformation features, and first-type propagation fault nodes and second-type propagation fault nodes are output.

[0028] Optionally, in order to accurately obtain the first type of propagation fault nodes and the second type of propagation fault nodes, the system terminal integrates the transformation feature formula to construct a feature transformation layer, which is used to map the aggregated features from the original space to a new feature space, so that in this new space, the distinction between nodes is higher, making it easier to identify subsequent fault propagation points. The transformation feature formula is as follows: ;in, is the transformed feature corresponding to the i-th node, which is a new feature representation that contains the information of the aggregated features of node i after nonlinear transformation. It is a weight matrix that determines the influence of each element in the aggregated feature on the transformed feature and is obtained in advance through training with a large amount of sample data. It is a bias term used to adjust the baseline value of the transformation feature, which is obtained in advance through training with a large amount of sample data. ReLU is a nonlinear activation function, which introduces a nonlinear transformation, setting the part of the input value less than 0 to 0 and the part greater than 0 unchanged. The aggregate feature representing the i-th node is obtained by aggregating the features of the set of neighboring nodes of node i. The system terminal encapsulates this transformation feature formula to form a feature transformation layer. After obtaining the feature transformation layer, the system terminal inputs N aggregate features into the feature transformation layer, and obtains the weighted feature representation by multiplying the aggregate features of each node with the weight matrix. Subsequently, the weighted feature representation is added to the bias term to obtain the results of multiple N linear transformations. After that, the results of the linear transformation are used as the input of the nonlinear activation function to obtain N transformation features of N nodes. These transformation features contain the information of the original aggregate features, and also incorporate the patterns and relationships of the feature transformation layer, so they are more representative and discriminative. Subsequently, the system terminal identifies the fault propagation points of these N transformation features according to the fault type probability prediction formula, and obtains the first type of propagation fault nodes and the second type of propagation fault nodes.

[0029] Furthermore, the present application provides for identifying fault propagation points of the N nodes according to the N transformation features, including: The N nodes are probability predicted using the N transformation features through a fault type probability prediction formula to obtain the first type of fault probability and the second type of fault probability corresponding to each node; according to the size of the first type of fault probability corresponding to each node, the nodes with a probability greater than or equal to the expected probability are output as first type of propagation fault nodes; according to the size of the second type of fault probability corresponding to each node, the nodes with a probability greater than or equal to the expected probability are output as second type of propagation fault nodes.

[0030] Optionally, after obtaining N transformation features, the system terminal inputs these transformation features into a pre-built fault type probability prediction formula for probability prediction. For each node, the fault type probability prediction formula outputs two probability values: one is the probability of the first type of fault occurring at the node, and the other is the probability of the second type of fault occurring at the node. These two probability values ​​are obtained based on the transformation features of the node through the parameters and calculation rules in the fault type probability prediction formula. Subsequently, an expected probability threshold is set based on historical data and risk tolerance. For the first type of fault, the system terminal regards nodes with a first type of fault probability greater than or equal to the expected probability as potential first type of propagation fault nodes and outputs them. This means that these nodes have a higher risk of first type of fault and may be the source or key node of fault propagation. Similarly, for the second type of fault, the same logic is used to regard nodes with a second type of fault probability greater than or equal to the expected probability as potential second type of propagation fault nodes and output them. These nodes have a higher risk of second type of fault and may also be the key link of fault propagation.

[0031] Furthermore, the present application provides a method for performing probability prediction on the N nodes using the N transformation features through a fault type probability prediction formula, including: The expression of the fault type probability prediction formula is as follows: ;in, is the probability of the first type of failure, representing the predicted probability of the first type of failure occurring at the i-th node, The second type of failure probability represents the predicted probability of the second type of failure occurring at the i-th node. , are the learning parameters used for prediction, corresponding to the weight matrix and bias, The activation function used for classification outputs a probability distribution, and the probability of each category is between 0 and 1.

[0032] Preferably, in fault prediction and classification, it is crucial to accurately evaluate the probability of each node having different fault types. To this end, the system terminal pre-constructs a fault type probability prediction formula, which combines the transformation characteristics and learning parameters of the node and outputs the probability of each node having different fault types through the softmax function. The expression of the fault type probability prediction formula is as follows: ;in, is the probability of the first type of failure, which represents the predicted probability of the first type of failure occurring at the i-th node, and is a value between 0 and 1. is the probability of the second type of failure, which represents the predicted probability of the second type of failure occurring at the i-th node, and is also a value between 0 and 1. It is a weight matrix used to map the transformation features of the node to the predicted probability of the fault type, which is pre-trained through a large amount of sample data. It is a bias term used to adjust the baseline value of the predicted probability, which is obtained by pre-training with a large amount of sample data. It is an activation function for classification, which converts the original output into a probability distribution. The probability of each category is between 0 and 1, and the sum of the probabilities of all categories is 1. The transformation feature corresponding to the i-th node is obtained by transforming the aggregated feature through the feature transformation layer. The system terminal multiplies the transformation feature of the i-th node with the weight matrix, and then adds the bias term to obtain the original output value. Subsequently, the original output value is input into the softmax function for processing to obtain two probability values, which respectively represent the probability of the first type of fault and the second type of fault occurring at node i.

[0033] The fault prediction module 5 is used to perform fault prediction on the first type of propagation fault nodes and the second type of propagation fault nodes, and obtain a first type of fault prediction index and a second type of fault prediction index.

[0034] In one embodiment, in the fault prediction module 5, in order to understand the situation of the potential fault node, the system terminal performs fault prediction on the first type of propagation fault node and the second type of propagation fault node, aiming to evaluate the possibility and severity of future fault development of these nodes. By using the first type of propagation fault node and the second type of propagation fault node to perform weight superposition and update on the fault graph network, a fault prediction result can be obtained. This fault prediction result includes the first type of fault prediction index and the second type of fault prediction index, which provides the system terminal with detailed information on the future state of the potential fault node, helps to better understand the risk of the fault, and provides support for subsequent fault prevention.

[0035] Further, the present application provides a method for predicting the faults of the first type of propagation fault nodes and the second type of propagation fault nodes, including: According to the second type of propagation fault nodes, the fault weights of the fault graph network are superimposed and updated; according to the first type of propagation fault nodes, the superimposed and updated fault graph network is updated and predicted, and a fault prediction result is output.

[0036] Preferably, after obtaining the first type of propagation fault nodes and the second type of propagation fault nodes, the system terminal quantifies the fault impact of the second type of propagation fault nodes and superimposes and updates them to the fault graph network. Since the second type of nodes have a wider impact in fault propagation, the fault effects they produce will act on the entire network, so superimposed updates are required to accurately reflect their impact on the entire mold. First, for each second type of propagation fault node, the system terminal quantitatively evaluates the degree and scope of its fault impact. The quantitative evaluation is carried out through various indicators, such as node failure probability, fault propagation range, fault impact time, etc. These indicators can help the system terminal determine the impact of each second type of fault node on the entire network. Subsequently, according to the results of the fault impact assessment, a fault weight is assigned to each second type of fault node. This weight represents the impact of the node failure on the entire network. After determining the fault weight of each second type of fault node, the system terminal superimposes these weights on the corresponding nodes of the fault graph network, that is, adds the two weights to update the fault weight distribution of the entire network. Afterwards, for the first type of propagation fault node, the fault graph network after superimposition and update is used to query its fault weight in the updated fault graph network. This weight represents the risk level of the node in the current network state. The larger the weight, the higher the risk. Then, the system terminal regards the fault weight of the first type of propagation fault node as a quantitative indicator of the risk level, that is, the first type of fault prediction indicator. This indicator reflects the possibility of these nodes having the first type of fault in the current state. The fault weight of the second type of propagation fault node is then regarded as a quantitative indicator of their overall fault risk, that is, the second type of fault prediction indicator. This indicator not only reflects the failure risk of the node itself, but also reflects their influence in the network and the potential risk of cascading failures. Finally, the obtained first type of fault prediction indicators and second type of fault prediction indicators are sorted into fault prediction results for output, providing more accurate data support for fault prediction.

[0037] Further, if Figure 2 As shown, the present application provides a method for optimizing fault prediction, including: A pressure monitoring data set of the die-casting die during the die-casting process is collected; and an internal pressure-bearing node of the die-casting die during the die-casting process is obtained according to the pressure monitoring data set.

[0038] Preferably, in the die-casting process of the die-casting mold, in order to ensure the safety of production and the stability of product quality, the system terminal can obtain the pressure changes of various parts of the mold during the die-casting process by collecting the pressure monitoring data set of the die-casting mold during the die-casting process. Based on these pressure monitoring data, the system terminal can analyze and identify the internal pressure-bearing nodes in the die-casting mold. These internal pressure-bearing nodes are specific areas or components that are subjected to greater pressure during the die-casting process. Because they are subjected to greater pressure, the risk of cracks and failures is relatively large.

[0039] Based on the pressure index size of the internal pressure-bearing node, a fault impact index is generated; and according to the fault impact index, fault prediction optimization is performed on the first type of propagation fault nodes and the second type of propagation fault nodes.

[0040] Preferably, the system terminal extracts the size of the pressure index from the internal pressure-bearing nodes. These pressure-bearing indicators reflect the size of the pressure that the node bears during the die-casting process and are an important basis for evaluating the risk of node failure. Subsequently, based on these pressure-bearing indicators, a fault impact index can be generated. The fault impact index is used to quantify the possible impact of the internal pressure-bearing node failure on the entire die-casting process or product quality. These fault impact indicators can help the system terminal more accurately understand the risk level and potential consequences of failures at different pressure-bearing nodes. Once the fault impact index is determined, the system terminal performs fault prediction optimization on the first type of propagation fault nodes and the second type of propagation fault nodes in the die-casting mold, that is, determines the pressure weight of each internal pressure-bearing node according to the fault impact index, and adds these pressure weights to the weights of the corresponding nodes in the fault graph network, and updates the fault graph network to optimize the fault prediction results.

[0041] The fault prediction result output module 6 is used to output the fault prediction result of the die-casting mold according to the first type of fault prediction index and the second type of fault prediction index.

[0042] In one embodiment, the system terminal assigns corresponding weights to the first type of fault prediction index and the second type of fault prediction index in the obtained fault prediction results based on historical experience and expert advice, and then performs weighted summation on the two indicators to generate the final fault prediction result. This prediction result is a comprehensive assessment of the failure risk of the entire die-casting mold. It comprehensively considers the pressure-bearing conditions of each node in the mold, the possibility of failure, and the potential impact range, thereby providing a relatively comprehensive and accurate fault risk prediction.

[0043] In summary, the fault diagnosis and prediction system for a motor rotor die-casting mold provided by the present application has the following technical effects: The present application obtains a multi-source monitoring signal set of a die-casting mold through a multi-source signal monitoring module 1, and uses a fault diagnosis module 2 to diagnose the signal set and generate a fault signal set. Subsequently, the graph neural network model is trained through a model training module 3 to construct a fault graph network of the die-casting mold. In the fault propagation point identification module 4, the first type of propagation fault node and the second type of propagation fault node in the fault graph network are identified. In order to optimize fault prediction, the internal pressure-bearing node is obtained according to the pressure monitoring data set of the die-casting process of the die-casting mold, and the fault impact index is generated based on the pressure index size. In the fault prediction module 5, the first and second types of propagation fault nodes are predicted to generate corresponding fault prediction indicators. The fault prediction result output module 6 outputs the fault prediction results of the die-casting mold according to these prediction indicators. In addition, fault propagation point identification, fault type probability prediction, and update prediction of the fault graph network are also performed. These technical effects jointly solve the technical problems of inaccurate fault diagnosis and prediction due to a single data source, and the inability to effectively identify the fault propagation path, so as to achieve the technical effect of accurately identifying the fault propagation point through early warning of the fault, improving the accuracy of fault diagnosis and prediction accuracy.

[0044] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0045] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A fault diagnosis and prediction system for a motor rotor die-casting mold, characterized in that: The system comprises: A multi-source signal monitoring module, used to perform multi-source signal monitoring on the die-casting mold according to the multi-source data acquisition device to obtain a multi-source monitoring signal set; A fault diagnosis module, used for performing fault diagnosis on the multi-source monitoring signal set to generate a fault signal set; A model training module, used for training a graph neural network model, wherein the graph neural network model constructs a fault graph network of the die-casting mold according to the fault signal set; A fault propagation point identification module is used to identify fault propagation points on the fault graph network and output a first type of propagation fault node and a second type of propagation fault node, wherein the first type of propagation fault node is a structural fault that propagates along a physical connection structure, and the second type of propagation fault node is a functional fault that propagates by affecting the entire die-casting mold; A fault prediction module, used to perform fault prediction on the first type of propagation fault nodes and the second type of propagation fault nodes, and obtain a first type of fault prediction index and a second type of fault prediction index; The fault prediction result output module is used to output the fault prediction result of the die-casting mold according to the first type of fault prediction index and the second type of fault prediction index.

2. A fault diagnosis and prediction system for a motor rotor die-casting mold as claimed in claim 1, characterized in that: The system further comprises: Collecting a pressure monitoring data set of the die-casting mold during the die-casting process; According to the pressure monitoring data set, obtaining the internal pressure-bearing nodes of the die-casting mold during the die-casting process; Generate a fault impact index based on the pressure index size of the internal pressure-bearing node; Fault prediction optimization is performed on the first type of propagation fault nodes and the second type of propagation fault nodes according to the fault impact index.

3. A fault diagnosis and prediction system for a motor rotor die-casting mold as claimed in claim 1, characterized in that: Performing fault propagation point identification on the fault graph network and outputting first-type propagation fault nodes and second-type propagation fault nodes includes: Obtain N nodes of the fault graph network; Collecting the fault type, fault characteristics, and impact range corresponding to each of the N nodes, and outputting N groups of fault characteristics corresponding to the N nodes; Fault propagation points of the N nodes are identified respectively according to the N groups of fault features, and first-type propagation fault nodes and second-type propagation fault nodes are output.

4. A fault diagnosis and prediction system for a motor rotor die-casting mold as claimed in claim 3, characterized in that: According to the N groups of fault characteristics, fault propagation points are identified for the N nodes respectively, and the system includes: Obtain N sets of neighboring nodes corresponding to the N nodes; Perform feature aggregation on the fault feature set corresponding to each group of neighboring node sets, and output N aggregated features corresponding to the N nodes; Processing the N aggregated features through a feature transformation layer, and outputting N transformed features corresponding to the N nodes; Fault propagation points of the N nodes are identified according to the N transformation features, and first-type propagation fault nodes and second-type propagation fault nodes are output.

5. A fault diagnosis and prediction system for a motor rotor die-casting mold as claimed in claim 4, characterized in that: The fault propagation points of the N nodes are identified according to the N transformation features, and the system further includes: Probability prediction of the N nodes is performed using the N transformation features through a fault type probability prediction formula to obtain a first type of fault probability and a second type of fault probability corresponding to each node; According to the size of the first type of failure probability corresponding to each node, the nodes with a probability greater than or equal to the expected probability are output as the first type of propagation failure nodes; According to the magnitude of the second type of failure probability corresponding to each node, nodes with a probability greater than or equal to the expected probability are output as second type propagation failure nodes.

6. A fault diagnosis and prediction system for a motor rotor die-casting mold as claimed in claim 5, characterized in that: The probability prediction of the N nodes is performed using the N transformation features through a fault type probability prediction formula, and the expression of the fault type probability prediction formula is as follows: ； in, is the probability of the first type of failure, representing the predicted probability of the first type of failure occurring at the i-th node, is the probability of the second type of failure, representing the predicted probability of the second type of failure occurring at the i-th node, , are the learning parameters used for prediction, corresponding to the weight matrix and bias, is an activation function for classification, the output is a probability distribution, and the probability of each category is between 0 and 1.

7. A fault diagnosis and prediction system for a motor rotor die-casting mold as claimed in claim 1, characterized in that: Performing fault prediction on the first-type propagation fault node and the second-type propagation fault node includes: According to the second type of propagation fault nodes, superimposing and updating the fault weights of the fault graph network; According to the first type of propagation fault nodes, the fault graph network after superimposition and update is updated and predicted, and the fault prediction result is output.

8. A fault diagnosis and prediction system for a motor rotor die-casting mold as claimed in claim 1, characterized in that: Performing fault diagnosis on the multi-source monitoring signal set to generate a fault signal set, the system further includes: Establishing a multi-source fault diagnosis channel, wherein each fault diagnosis channel stores a healthy signal sample, and each fault diagnosis channel is used to receive a corresponding monitoring signal set to perform fault signal identification and output a multi-source fault signal; The multi-source fault signals are fitted to generate a fault signal set.

Images