Equipment fault determination method and device, computer equipment and readable storage medium

Abstract

The invention relates to an equipment fault determination method and device, computer equipment and a computer readable storage medium. The method comprises the following steps: acquiring current state data of target equipment and current data characteristics of the current state data in real time; acquiring a plurality of first historical working conditions, and comparing the current state data with historical state data corresponding to the first historical working conditions; obtaining a first target fault recognition model and a second target fault recognition model; and based on a comparison result, inputting the current data features into a first target fault recognition model to obtain a current fault type, or inputting dimensionless features in the current data features into a second target fault recognition model to obtain the current fault type. According to the method provided by the invention, the accuracy of fault diagnosis on the target equipment can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of industrial control technology, and in particular to a method and apparatus for determining equipment failure, a computer device, and a computer-readable storage medium. Background Art

[0002] In a highly automated and intelligent modern industrial system, mechanical equipment often needs to operate continuously in extremely changeable working environments, which constitute complex and diverse working conditions. Multiple working conditions not only mean that the equipment needs to withstand fluctuations in physical parameters such as temperature, pressure, vibration and load, but also mean that the combination and change rate of these parameters have extremely high uncertainty. This uncertainty poses a severe challenge to the stability and reliability of the equipment, and also greatly increases the complexity and difficulty of fault diagnosis, which may reduce the accuracy of fault diagnosis of mechanical equipment. Summary of the invention

[0003] Based on this, it is necessary to provide a method, device, computer equipment and computer-readable storage medium for determining equipment faults that can improve the accuracy of fault diagnosis in order to solve the above technical problems.

[0004] In a first aspect, the present application provides a method for determining a device failure, the method comprising:

[0005] Acquire the current status data of the target device and the current data characteristics of the current status data in real time;

[0006] Acquire a plurality of first historical operating conditions, and compare the current state data with the historical state data corresponding to the first historical operating conditions;

[0007] Acquire a first target fault identification model and a second target fault identification model, wherein the first target fault identification model is obtained by training an initial fault identification model based on historical data features of the historical state data, and the second target fault identification model is obtained by training the initial fault identification model based on dimensionless features in the historical data features;

[0008] Based on the comparison result, the current data feature is input into the first target fault identification model to obtain the current fault type, or the dimensionless feature in the current data feature is input into the second target fault identification model to obtain the current fault type.

[0009] In one embodiment, the data features include time domain features and frequency domain features; the step of obtaining the current data features of the current state data includes:

[0010] Acquire a time domain signal of the current state data, and acquire a time domain feature of the current state data based on the time domain signal;

[0011] Performing Fourier transform on the time domain signal to extract a first frequency domain feature of the current state data;

[0012] Performing Hilbert transform on the time domain signal to obtain an analytical signal corresponding to the time domain signal, and determining the modulus of the analytical signal as the envelope signal of the current state data;

[0013] Performing Fourier transform on the envelope signal to convert the envelope signal from the time domain to the frequency domain;

[0014] A second frequency domain feature of the current state data is extracted from the envelope signal in the frequency domain.

[0015] In one embodiment, the obtaining of multiple first historical operating conditions includes:

[0016] Acquire historical status data of the target device within a preset time period, and divide the preset time period into a preceding time period and a succeeding time period based on a preset ratio;

[0017] Determine the historical status data acquired in the preceding time period as first historical status data, and determine the historical status data acquired in the succeeding time period as second historical status data;

[0018] Mapping the first historical state data from an initial dimensional space to a target dimensional space using a preset dimensionality reduction model, wherein the dimension of the initial dimensional space is higher than the dimension of the target dimensional space;

[0019] Based on the hierarchical clustering model, the first historical state data after dimension reduction is clustered to divide the first historical state data into a plurality of cluster clusters, each cluster cluster corresponding to a first historical operating condition.

[0020] In one embodiment, the training process of the fault identification model includes:

[0021] Inputting the historical data features of the first historical state data or the dimensionless features in the historical data features into the initial fault identification model to train the initial fault identification model;

[0022] For each round of training, when the loss value output by the fault identification model is within a preset loss value range, or the cumulative number of rounds is greater than a preset number, the fault identification model obtained in the round of training is determined as an intermediate fault identification model;

[0023] Determine an intermediate fault identification model determined based on the historical data characteristics of the first historical state data as a first intermediate fault identification model, and determine the first target fault identification model based on the first intermediate fault identification model and the second historical state data;

[0024] An intermediate fault identification model determined based on the dimensionless features in the historical data features is determined as a second intermediate fault identification model, and the second target fault identification model is determined based on the second intermediate fault identification model and the second historical state data.

[0025] In one embodiment, determining the first target fault identification model based on the first intermediate fault identification model and the second historical state data includes:

[0026] Based on the preset dimension reduction model and the hierarchical clustering model, obtaining a plurality of second historical operating conditions corresponding to the second historical state data, and an actual fault type corresponding to each second historical operating condition;

[0027] For each second historical operating condition, when there is a first historical operating condition that is the same as the second historical operating condition, inputting historical data features of second historical state data corresponding to the second historical operating condition into the first intermediate fault identification model, and outputting an output fault type corresponding to the second historical operating condition;

[0028] Based on the difference between the output fault type and the actual fault type, obtaining an accuracy evaluation value of the first intermediate fault identification model;

[0029] When the accuracy evaluation value is within a first preset accuracy range, determining the first intermediate fault identification model as the first target fault identification model;

[0030] When the accuracy evaluation value is not within the first preset accuracy range, return to the step of inputting the historical data features of the first historical state data into the initial fault identification model, and continue to execute until the accuracy evaluation value is within the first preset accuracy range.

[0031] In one embodiment, determining the second target fault identification model based on the second intermediate fault identification model and the second historical state data includes:

[0032] Based on the preset dimension reduction model and the hierarchical clustering model, obtaining a plurality of second historical operating conditions corresponding to the second historical state data, and an actual fault type corresponding to each second historical operating condition;

[0033] For each second historical operating condition, when there is no first historical operating condition identical to the second historical operating condition, inputting the dimensionless features of the second historical state data corresponding to the second historical operating condition into the second intermediate fault identification model, and outputting the output fault type corresponding to the second historical operating condition;

[0034] Based on the difference between the output fault type and the actual fault type, obtaining an accuracy evaluation value of the second intermediate fault identification model;

[0035] When the accuracy evaluation value is within a second preset accuracy range, determining the second intermediate fault identification model as the second target fault identification model;

[0036] When the accuracy evaluation value is not within the second preset accuracy range, return to the step of inputting the dimensionless features in the historical data features into the initial fault identification model, and continue to execute until the accuracy evaluation value is within the second preset accuracy range.

[0037] In one embodiment, based on the comparison result, inputting the current data feature into the first target fault identification model to obtain the current fault type, or inputting the dimensionless feature in the current data feature into the second target fault identification model to obtain the current fault type, comprises:

[0038] In the case where there is a data difference between the historical state data corresponding to the first historical working condition and the current state data and the difference is within a preset data range, inputting the current data feature into the first target fault identification model to obtain the current fault type;

[0039] When there is no data difference between the historical state data corresponding to the first historical operating condition and the current state data and the data is within a preset data range, the dimensionless features in the current data features are input into the second target fault identification model to obtain the current fault type.

[0040] In a second aspect, the present application further provides a device for determining a device failure, the device comprising:

[0041] A first acquisition module, used to acquire the current state data of the target device and the current data characteristics of the current state data in real time;

[0042] A second acquisition module is used to acquire a plurality of first historical operating conditions and compare the current state data with the historical state data corresponding to the first historical operating conditions;

[0043] A third acquisition module is used to acquire a first target fault recognition model and a second target fault recognition model, wherein the first target fault recognition model is obtained by training an initial fault recognition model based on historical data features of the historical state data, and the second target fault recognition model is obtained by training the initial fault recognition model based on dimensionless features in the historical data features;

[0044] An input module is used to input the current data features into the first target fault identification model to obtain the current fault type based on the comparison result, or input the dimensionless features in the current data features into the second target fault identification model to obtain the current fault type.

[0045] In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method in any one of the above embodiments when executing the computer program.

[0046] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method in any one of the above embodiments are implemented.

[0047] The above-mentioned equipment fault determination method, device, computer equipment and computer-readable storage medium obtain the current state data of the target equipment and the current data features of the current state data in real time; obtain multiple first historical working conditions, and compare the current state data with the historical state data corresponding to the first historical working conditions; obtain the first target fault identification model and the second target fault identification model, the first target fault identification model is obtained by training the initial fault identification model based on the historical data features of the historical state data, and the second target fault identification model is obtained by training the initial fault identification model based on the dimensionless features in the historical data features; based on the comparison result, the current data features are input into the first target fault identification model to obtain the current fault type, or the dimensionless features in the current data features are input into the second target fault identification model to obtain the current fault type. The method provided by the present application can effectively improve the accuracy of fault diagnosis of the target equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 A schematic diagram of a flow chart of a method for determining a device failure in one embodiment;

[0050] Figure 2 A schematic diagram of the distribution of test set data and training set data in one embodiment;

[0051] Figure 3 It is a flowchart of a method for obtaining current data features in one embodiment;

[0052] Figure 4 is a spectrum diagram of a signal in one embodiment;

[0053] Figure 5 is a schematic curve diagram of the number of clusters in one embodiment;

[0054] Figure 6 A schematic diagram of a process of dimensionality reduction and hierarchical clustering in one embodiment;

[0055] Figure 7 Schematic diagram of time-frequency domain waveform of inner ring fault in one embodiment;

[0056] Figure 8 A schematic diagram of a time-frequency domain waveform of an outer ring fault in an embodiment;

[0057] Fig. 9 A schematic flow chart of a method for determining a device failure in another embodiment;

[0058] Fig.10 A schematic diagram of modeling historical data in another embodiment;

[0059] Fig.11 A schematic diagram of fault prediction in another embodiment;

[0060] Fig.12 is a structural block diagram of a device for determining a device failure in one embodiment;

[0061] Fig.13 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. 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.

[0063] In one embodiment, Figure 1As shown, a method for determining a device failure is provided. This embodiment uses the method applied to a terminal as an example for illustration. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0064] S102: Acquire current status data of the target device and current data features of the current status data in real time.

[0065] Optionally, the state data may include, but is not limited to, vibration data, temperature data, rotation speed data, pressure data, current data, and voltage data.

[0066] Optionally, after the current state data is acquired, preprocessing work such as data cleaning, denoising and standardization may be performed on the current state data to ensure the quality of the data.

[0067] Optionally, after the current data features are acquired, features irrelevant to fault identification may be screened out from the current data features, and only features relevant to fault identification are retained.

[0068] S104: Acquire a plurality of first historical operating conditions, and compare the current state data with the historical state data corresponding to the first historical operating conditions.

[0069] Among them, working conditions refer to the various working states and conditions of the equipment during actual operation.

[0070] S106. Obtain a first target fault identification model and a second target fault identification model. The first target fault identification model is obtained by training an initial fault identification model based on historical data features of historical state data, and the second target fault identification model is obtained by training the initial fault identification model based on dimensionless features in historical data features.

[0071] Among them, dimensionless features refer to the process of converting data with actual physical meaning or measurement units into numerical values ​​without units or dimensions during feature engineering and data preprocessing. Dimensionless features can include but are not limited to waveform symmetry, margin factor, kurtosis factor, pulse factor, etc. Some generalized features that are not affected by loads can also be included in the dimensionless category, such as the frequency band with the highest energy density, which is not affected by the specific size of the energy.

[0072] The current state data corresponds to the current working condition. Only when the current working condition is one of the multiple first historical working conditions, can the first target fault identification model trained according to the historical data features of the historical state data identify the fault type corresponding to the current state data. Since the second target fault identification model obtained through dimensionless feature training does not rely on a specific physical dimension, it can better adapt to the inconsistency of data distribution under different working conditions. This adaptability enables the second target fault identification model to maintain relatively good prediction and classification performance under unknown working conditions. Therefore, in the absence of the first historical working condition being consistent with the current working condition, the second target fault identification model can be used to identify the fault type corresponding to the current state data.

[0073] Optionally, in the process of training the initial fault identification model, the model is trained using the training set data, and the training result of the model is tested using the test set data. If there is no operating condition corresponding to the training set data that is consistent with the operating condition corresponding to the test set data, the dimensionless features corresponding to the training set data can be extracted, and the initial fault identification model can be trained using the extracted dimensionless features so that the feature space corresponding to the test set data is included in the feature space corresponding to the training set data, thereby overcoming the problem of inconsistent data distribution. Figure 2 As shown, Figure 2 The train in the formula represents the training set data, and the test represents the test set data. Figure 2 The left figure in the figure shows the distribution between the test set data and the training set data when the working conditions corresponding to the training set data are not consistent with the working conditions corresponding to the test set data. Figure 2 The right figure in shows the distribution between the test set data and the training set data after extracting the corresponding dimensionless features of the training set data.

[0074] S108. Based on the comparison result, input the current data feature into the first target fault identification model to obtain the current fault type, or input the dimensionless feature in the current data feature into the second target fault identification model to obtain the current fault type.

[0075] Optionally, when the first historical operating condition is consistent with the current operating condition, the current data features can be input into the first target fault identification model to obtain the current fault type; when the first historical operating condition is not consistent with the current operating condition, the dimensionless features in the current data features can be input into the second target fault identification model to obtain the current fault type.

[0076] In the above-mentioned equipment fault determination method, the current state data of the target equipment and the current data features of the current state data are obtained in real time; a plurality of first historical working conditions are obtained, and the current state data is compared with the historical state data corresponding to the first historical working conditions; a first target fault identification model and a second target fault identification model are obtained, the first target fault identification model is obtained by training the initial fault identification model based on the historical data features of the historical state data, and the second target fault identification model is obtained by training the initial fault identification model based on the dimensionless features in the historical data features; based on the comparison result, the current data features are input into the first target fault identification model to obtain the current fault type, or the dimensionless features in the current data features are input into the second target fault identification model to obtain the current fault type. The method provided by the present application can effectively improve the accuracy of fault diagnosis of the target equipment.

[0077] In some embodiments, Figure 3 As shown, the data features include time domain features and frequency domain features; the current data features of the current state data are obtained, including:

[0078] S302: Acquire a time domain signal of current state data, and acquire a time domain feature of the current state data based on the time domain signal.

[0079] S304: Perform Fourier transform on the time domain signal to extract the first frequency domain feature of the current state data.

[0080] S306 , performing Hilbert transform on the time domain signal to obtain an analytical signal corresponding to the time domain signal, and determining the modulus of the analytical signal as the envelope signal of the current state data.

[0081] S308: Perform Fourier transform on the envelope signal to convert the envelope signal from the time domain to the frequency domain.

[0082] S310: Extract a second frequency domain feature of current state data from an envelope signal in the frequency domain.

[0083] Among them, the time domain signal is used to describe the change of the current state data over time; the time domain feature refers to the characteristics of the signal on the time axis, which describes the law of the signal changing over time. For example, the time domain feature can be the maximum value, minimum value, mean, median, kurtosis, skewness, anomaly number, waveform factor, etc.; the frequency domain feature refers to the characteristics of the frequency component obtained after the time domain signal is converted to the frequency domain through Fourier transform. For example, the frequency domain feature can be the frequency centroid, frequency variance, frequency root mean square, energy entropy, etc.; Fourier transform is a method of converting a time domain signal into a frequency domain signal. It can decompose a signal that is continuous or discrete in time into a superposition of sine waves and cosine waves of different frequencies; Hilbert transform is a generalization of Fourier transform. It converts a real signal into its dual part of the analytical signal, so as to analyze the time domain characteristics and frequency domain characteristics of the signal; envelope signal refers to a curve describing the amplitude change of a complex signal obtained by analyzing it in signal processing.

[0084] Optionally, the time domain features and the frequency domain features may be used as data features respectively, or the features obtained by fusing the time domain features and the frequency domain features may be used as data features, which is not specifically limited in the embodiments of the present application.

[0085] Since the impact vibration signal caused by the fault is often masked by high-frequency noise, it is necessary to use envelope technology to separate the low-frequency fault information modulated on the high-frequency carrier, so as to more clearly identify and analyze the fault characteristics and improve the accuracy and reliability of fault diagnosis. The essence of the technology of extracting envelope is to separate the profile of amplitude change over time from the original signal through mathematical transformation.

[0086] Optionally, the extraction process of the envelope signal is as follows:

[0087] For the time domain signal x(t), its Hilbert transform is as follows:

[0088]

[0089] In the formula, * is convolution, is the Hilbert transform of x(t).

[0090] The analytical signal z(t) is shown as follows:

[0091]

[0092] The modulus of the analytical signal is the envelope of the original signal. The envelope signal is shown as follows:

[0093]

[0094] Take a bearing vibration data as an example. The spectrum diagram of the time domain signal without envelope analysis is unable to diagnose the fault because the fault signal is submerged in the noise. Therefore, it is necessary to perform envelope analysis on the time domain waveform of the bearing's time domain signal, and then perform corresponding spectrum analysis on the envelope signal to obtain the inner ring fault characteristic frequency of the bearing. According to the inner ring fault characteristic frequency, it can be known that the bearing may have an inner ring fault. Figure 4 As shown, Figure 4 The left picture in the figure is the spectrum diagram of the original time domain signal, and the right picture is the spectrum diagram after the envelope is extracted.

[0095] In this embodiment, the frequency domain feature is determined by extracting the envelope of the time domain signal, so that the accuracy of fault diagnosis based on the determined frequency domain feature is higher.

[0096] In some embodiments, multiple first historical operating conditions are obtained, including: obtaining historical status data of a target device within a preset time period, and dividing the preset time period into a former time period and a latter time period based on a preset ratio; determining the historical status data obtained in the former time period as first historical status data, and determining the historical status data obtained in the latter time period as second historical status data; mapping the first historical status data from an initial dimensional space to a target dimensional space using a preset dimensionality reduction model, the dimensionality of the initial dimensional space being higher than the dimensionality of the target dimensional space; and clustering the first historical status data after dimensionality reduction based on a hierarchical clustering model to divide the first historical status data into multiple cluster clusters, each cluster cluster corresponding to a first historical operating condition.

[0097] Among them, the dimensionality reduction model is a data preprocessing technology, which aims to reduce the complexity of data by reducing the characteristic dimensions of the data. For example, the preset dimensionality reduction model can be a T-SNE (t-Distributed Stochastic Neighbor Embedding) dimensionality reduction model. T-SNE is a nonlinear model for dimensionality reduction. The hierarchical clustering model is an unsupervised learning algorithm used to classify or group a set of data points according to their similarities. Figure 5 and Figure 6 As shown, Figure 5 is a graph showing the number of clusters. Figure 6 Schematic diagram of the process of dimensionality reduction and hierarchical clustering.

[0098] Alternatively, take rolling bearing data as an example. Extract the vibration waveform of the rolling bearing during continuous operation. The vibration waveform covers data of four common state types: normal state, inner ring fault, outer ring fault, and rolling element fault. Due to continuous operation in a changing working environment, complex and diverse working conditions are caused. Figure 7 and Figure 8As shown in the figure, under different working conditions, the time-frequency domain waveforms of the same type of faults are also very different. Figure 7 The inner fault in the equation indicates an inner ring fault. Figure 8 The outer fault in the figure indicates an outer race fault.

[0099] In this embodiment, the first historical operating condition is determined based on the preset dimensionality reduction model and the hierarchical clustering model in sequence, so that the determined first historical operating condition is more accurate.

[0100] In some embodiments, the training process of the fault identification model includes: inputting the historical data features of the first historical state data or the dimensionless features in the historical data features into the initial fault identification model to train the initial fault identification model; for each round of training process, when the loss value output by the fault identification model is within a preset loss value range, or the cumulative number of rounds is greater than the preset number, the fault identification model obtained in the round training is determined as an intermediate fault identification model; the intermediate fault identification model determined based on the historical data features of the first historical state data is determined as a first intermediate fault identification model, and a first target fault identification model is determined based on the first intermediate fault identification model and the second historical state data; the intermediate fault identification model determined based on the dimensionless features in the historical data features is determined as a second intermediate fault identification model, and a second target fault identification model is determined based on the second intermediate fault identification model and the second historical state data.

[0101] Optionally, the fault identification model may be a classification model constructed by xgboost. XGBoost (Extreme Gradient Boosting) is an efficient gradient boosting algorithm. The loss value is within a preset loss value range, indicating that the loss function is in a convergence state, and the obtained model meets the training requirements.

[0102] In this embodiment, when the loss value output by the fault identification model is within a preset loss value range, or the cumulative number of rounds is greater than a preset number, the model training is determined to be completed, so that the intermediate fault identification model obtained by training is more accurate.

[0103] In some embodiments, a first target fault identification model is determined based on a first intermediate fault identification model and second historical state data, including: based on a preset dimensionality reduction model and a hierarchical clustering model, obtaining a plurality of second historical operating conditions corresponding to the second historical state data, and an actual fault type corresponding to each second historical operating condition; for each second historical operating condition, in the presence of a first historical operating condition identical to the second historical operating condition, inputting the historical data features of the second historical state data corresponding to the second historical operating condition into the first intermediate fault identification model, and outputting an output fault type corresponding to the second historical operating condition; based on the difference between the output fault type and the actual fault type, obtaining an accuracy evaluation value of the first intermediate fault identification model; in the case where the accuracy evaluation value is within a first preset accuracy range, determining the first intermediate fault identification model as the first target fault identification model; in the case where the accuracy evaluation value is not within the first preset accuracy range, returning to the step of inputting the historical data features of the first historical state data into the initial fault identification model, and continuing to execute until the accuracy evaluation value is within the first preset accuracy range.

[0104] Optionally, the process of acquiring the second historical operating condition based on the preset dimensionality reduction model and the hierarchical clustering model is consistent with the process of acquiring the first historical operating condition.

[0105] Optionally, during the model training process, the first intermediate fault identification model is first trained using the historical data features of the first historical state data. Subsequently, the fault type output by the validation set of the model is compared with the actual fault type to evaluate the accuracy of the model. If the accuracy of the model reaches or exceeds a preset acceptable level (i.e., the difference between the predicted result of the model and the actual result is within a tolerable range), the model is determined to be the first target fault identification model and is ready to be applied to the actual fault identification task. If the accuracy of the model has not yet met the requirements, the iterative optimization phase will be entered. In this phase, the model training step will be re-executed, but before that, the parameters of the model (such as the maximum depth of the tree, the learning rate, the L1 and L2 regularization parameters, etc.) will be adjusted based on the previous training results and the model performance analysis to improve the performance of the model. The adjusted model will be trained again using the historical data features of the first historical state data, and its accuracy will be evaluated again based on the validation set. This process will be iterative, and each iteration may involve further adjustment of the model parameters until the accuracy of the model is no longer improved. In this way, it can be ensured that the fault identification model finally obtained is both accurate and reliable and can meet the needs of practical applications.

[0106] In this embodiment, the first target fault identification model is determined based on the second historical state data corresponding to the second historical operating condition, so that the determined first target fault identification model is more accurate.

[0107] In some embodiments, a second target fault identification model is determined based on a second intermediate fault identification model and second historical state data, including: based on a preset dimensionality reduction model and a hierarchical clustering model, obtaining a plurality of second historical operating conditions corresponding to the second historical state data, and an actual fault type corresponding to each second historical operating condition; for each second historical operating condition, in the absence of a first historical operating condition identical to the second historical operating condition, inputting the dimensionless features of the second historical state data corresponding to the second historical operating condition into the second intermediate fault identification model, and outputting an output fault type corresponding to the second historical operating condition; based on the difference between the output fault type and the actual fault type, obtaining an accuracy evaluation value of the second intermediate fault identification model; when the accuracy evaluation value is within a second preset accuracy range, determining the second intermediate fault identification model as the second target fault identification model; when the accuracy evaluation value is not within the second preset accuracy range, returning to the step of inputting the dimensionless features in the historical data features into the initial fault identification model, and continuing to execute until the accuracy evaluation value is within the second preset accuracy range.

[0108] Optionally, optionally, the process of acquiring the second historical operating condition based on the preset dimensionality reduction model and the hierarchical clustering model is consistent with the process of acquiring the first historical operating condition.

[0109] In this embodiment, the second target fault identification model is determined based on the dimensionless characteristics of the second historical state data corresponding to the second historical operating condition, so that the determined second target fault identification model is more accurate.

[0110] In some embodiments, based on the comparison results, the current data features are input into the first target fault identification model to obtain the current fault type, or the dimensionless features in the current data features are input into the second target fault identification model to obtain the current fault type, including: when there is a data difference between the historical state data corresponding to the first historical working condition and the current state data and the data is within a preset data range, the current data features are input into the first target fault identification model to obtain the current fault type; when there is no data difference between the historical state data corresponding to the first historical working condition and the current state data and the data is within the preset data range, the dimensionless features in the current data features are input into the second target fault identification model to obtain the current fault type.

[0111] Optionally, the data difference between the historical status data corresponding to the first historical operating condition and the current status data is within a preset data range, indicating that the first historical operating condition is consistent with the current operating condition; the data difference between the historical status data corresponding to the first historical operating condition and the current status data is not within the preset data range, indicating that the first historical operating condition is inconsistent with the current operating condition.

[0112] In this embodiment, based on the consistency between the first historical operating condition and the current operating condition, different data features and models are selected to determine the fault type, so that the determined fault type is more accurate.

[0113] In one embodiment, Fig. 9 , Fig.10 and Fig.11 As shown, another method for determining device failure is provided. Fig. 9 The flowchart of this method is shown in Figure 2. Fig.10 A schematic diagram of the historical data modeling in this method, Fig.11 The schematic diagram of fault prediction in this method is shown in Figure 1. The method includes the following contents:

[0114] In the model application phase, we built an efficient architecture that integrates Kafka data queues and the Apache Flink streaming computing framework. The advantage of this architecture is that it can seamlessly connect to massive data streams, ensuring that data enters the processing flow with extremely low latency (seconds or even milliseconds). Through Flink's powerful streaming processing capabilities, we have achieved real-time computing of sensor data, refined feature extraction and fusion, and real-time working condition analysis.

[0115] In order to ensure the continued efficiency and accuracy of the diagnostic model, we have introduced a model drift monitoring mechanism. This mechanism intelligently evaluates model performance by continuously comparing the deviation between the model prediction results and the actual observed data. Once a significant decline in model performance is detected, the model update process is automatically triggered. In this process, we use the latest data resources to retrain or fine-tune the model to ensure that the diagnostic model can quickly adapt to changes in data distribution and always remain in the optimal state. This dynamic adjustment mechanism not only improves the accuracy of equipment diagnosis, but also significantly enhances the flexibility and robustness of the diagnostic system.

[0116] For a single prediction task, the system first receives real-time data from sensors. Subsequently, these data are efficiently processed by Flink's streaming processing engine, including key steps such as data cleaning, feature extraction and fusion. On this basis, the system further analyzes the current working condition and accurately determines whether it belongs to a known working condition or an unknown working condition. For known working conditions, the working condition label is added, and a pre-trained fault diagnosis model optimized for the working condition is called for rapid diagnosis. For unknown working conditions, the system switches to a dimensionless model for diagnosis to ensure accurate and reliable diagnostic results even under complex and changeable working conditions. This design not only improves the pertinence and efficiency of diagnosis, but also enhances the system's adaptability to unknown working conditions.

[0117] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0118] Based on the same inventive concept, the embodiment of the present application also provides a device for implementing the device failure determination method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more device failure determination device embodiments provided below can refer to the limitations of the device failure determination method above, and will not be repeated here.

[0119] In an exemplary embodiment, Fig.12 As shown, a device 1200 for determining a device failure is provided, comprising: a first acquisition module 1201, a second acquisition module 1202, a third acquisition module 1203 and an input module 1204, wherein:

[0120] The first acquisition module 1201 is used to acquire the current state data of the target device and the current data features of the current state data in real time.

[0121] The second acquisition module 1202 is used to acquire multiple first historical operating conditions and compare the current state data with the historical state data corresponding to the first historical operating conditions.

[0122] The third acquisition module 1203 is used to obtain a first target fault identification model and a second target fault identification model, wherein the first target fault identification model is obtained by training an initial fault identification model based on historical data features of the historical state data, and the second target fault identification model is obtained by training the initial fault identification model based on dimensionless features in the historical data features.

[0123] The input module 1204 is used to input the current data features into the first target fault identification model to obtain the current fault type based on the comparison result, or input the dimensionless features in the current data features into the second target fault identification model to obtain the current fault type.

[0124] In some embodiments, the data features include time domain features and frequency domain features; the first acquisition module 1201 is also used to acquire the time domain signal of the current state data, and acquire the time domain features of the current state data based on the time domain signal; perform Fourier transform on the time domain signal to extract the first frequency domain features of the current state data; perform Hilbert transform on the time domain signal to obtain an analytical signal corresponding to the time domain signal, and determine the modulus of the analytical signal as the envelope signal of the current state data; perform Fourier transform on the envelope signal to convert the envelope signal from the time domain to the frequency domain; and extract the second frequency domain features of the current state data from the envelope signal in the frequency domain.

[0125] In some embodiments, the second acquisition module 1202 is also used to obtain historical status data of the target device within a preset time period, and divide the preset time period into a previous time period and a subsequent time period based on a preset ratio; determine the historical status data obtained in the previous time period as the first historical status data, and determine the historical status data obtained in the subsequent time period as the second historical status data; use a preset dimensionality reduction model to map the first historical status data from an initial dimensional space to a target dimensional space, the dimension of the initial dimensional space is higher than the dimension of the target dimensional space; based on a hierarchical clustering model, cluster the first historical status data after dimensionality reduction to divide the first historical status data into multiple cluster clusters, each cluster cluster corresponding to a first historical operating condition.

[0126] In some embodiments, the device failure determination apparatus 1200 further includes:

[0127] A training module is used to input the historical data features of the first historical state data or the dimensionless features in the historical data features into the initial fault identification model to train the initial fault identification model.

[0128] The first determination module is used to determine the fault identification model obtained in each round of training as an intermediate fault identification model when the loss value output by the fault identification model is within a preset loss value range or the cumulative number of rounds is greater than a preset number.

[0129] The second determination module is used to determine the intermediate fault identification model determined based on the historical data characteristics of the first historical state data as the first intermediate fault identification model, and determine the first target fault identification model based on the first intermediate fault identification model and the second historical state data.

[0130] The third determination module is used to determine the intermediate fault identification model determined based on the dimensionless characteristics in the historical data characteristics as the second intermediate fault identification model, and determine the second target fault identification model based on the second intermediate fault identification model and the second historical state data.

[0131] In some embodiments, the second determination module is also used to obtain multiple second historical operating conditions corresponding to the second historical state data and the actual fault type corresponding to each second historical operating condition based on the preset dimensionality reduction model and the hierarchical clustering model; for each second historical operating condition, when there is a first historical operating condition that is the same as the second historical operating condition, the historical data features of the second historical state data corresponding to the second historical operating condition are input into the first intermediate fault identification model, and the output fault type corresponding to the second historical operating condition is output; based on the difference between the output fault type and the actual fault type, the accuracy evaluation value of the first intermediate fault identification model is obtained; when the accuracy evaluation value is within the first preset accuracy range, the first intermediate fault identification model is determined as the first target fault identification model; when the accuracy evaluation value is not within the first preset accuracy range, the step of inputting the historical data features of the first historical state data into the initial fault identification model is returned, and the execution is continued until the accuracy evaluation value is within the first preset accuracy range.

[0132] In some embodiments, the third determination module is also used to obtain multiple second historical operating conditions corresponding to the second historical state data and the actual fault type corresponding to each second historical operating condition based on the preset dimensionality reduction model and the hierarchical clustering model; for each second historical operating condition, in the absence of a first historical operating condition identical to the second historical operating condition, the dimensionless features of the second historical state data corresponding to the second historical operating condition are input into the second intermediate fault identification model, and the output fault type corresponding to the second historical operating condition is output; based on the difference between the output fault type and the actual fault type, the accuracy evaluation value of the second intermediate fault identification model is obtained; when the accuracy evaluation value is within the second preset accuracy range, the second intermediate fault identification model is determined as the second target fault identification model; when the accuracy evaluation value is not within the second preset accuracy range, the step of inputting the dimensionless features in the historical data features into the initial fault identification model is returned, and the execution is continued until the accuracy evaluation value is within the second preset accuracy range.

[0133] In some embodiments, the input module 1204 is also used to input the current data features into the first target fault identification model to obtain the current fault type when there is a data difference between the historical status data corresponding to the first historical operating condition and the current status data and the data difference is within a preset data range; and to input the dimensionless features in the current data features into the second target fault identification model to obtain the current fault type when there is no data difference between the historical status data corresponding to the first historical operating condition and the current status data and the data difference is within a preset data range.

[0134] Each module in the above-mentioned device fault determination apparatus can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module above.

[0135] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.13 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a method for determining a device fault is implemented.

[0136] Those skilled in the art will understand that Fig.13 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0137] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0138] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0139] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0140] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0141] The technical features of the above embodiments may be combined arbitrarily. 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 application.

[0142] The above-described 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 scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for determining equipment failure, characterized in that: The method comprises: Acquire the current status data of the target device and the current data characteristics of the current status data in real time; Acquire a plurality of first historical operating conditions, and compare the current state data with the historical state data corresponding to the first historical operating conditions; Acquire a first target fault identification model and a second target fault identification model, wherein the first target fault identification model is obtained by training an initial fault identification model based on historical data features of the historical state data, and the second target fault identification model is obtained by training the initial fault identification model based on dimensionless features in the historical data features; Based on the comparison result, the current data feature is input into the first target fault identification model to obtain the current fault type, or the dimensionless feature in the current data feature is input into the second target fault identification model to obtain the current fault type.

2. The method according to claim 1, characterized in that Data features include time domain features and frequency domain features; The obtaining of the current data feature of the current state data includes: Acquire a time domain signal of the current state data, and acquire a time domain feature of the current state data based on the time domain signal; Performing Fourier transform on the time domain signal to extract a first frequency domain feature of the current state data; Performing Hilbert transform on the time domain signal to obtain an analytical signal corresponding to the time domain signal, and determining the modulus of the analytical signal as the envelope signal of the current state data; Performing Fourier transform on the envelope signal to convert the envelope signal from the time domain to the frequency domain; A second frequency domain feature of the current state data is extracted from the envelope signal in the frequency domain.

3. The method according to claim 1, characterized in that The obtaining of multiple first historical operating conditions includes: Acquire historical status data of the target device within a preset time period, and divide the preset time period into a preceding time period and a succeeding time period based on a preset ratio; Determine the historical status data acquired in the preceding time period as first historical status data, and determine the historical status data acquired in the succeeding time period as second historical status data; Mapping the first historical state data from an initial dimensional space to a target dimensional space using a preset dimensionality reduction model, wherein the dimension of the initial dimensional space is higher than the dimension of the target dimensional space; Based on the hierarchical clustering model, the first historical state data after dimension reduction is clustered to divide the first historical state data into a plurality of cluster clusters, each cluster cluster corresponding to a first historical operating condition.

4. The method according to claim 3, characterized in that The training process of the fault identification model includes: Inputting the historical data features of the first historical state data or the dimensionless features in the historical data features into the initial fault identification model to train the initial fault identification model; For each round of training, when the loss value output by the fault identification model is within a preset loss value range, or the cumulative number of rounds is greater than a preset number, the fault identification model obtained in the round of training is determined as an intermediate fault identification model; Determine an intermediate fault identification model determined based on the historical data characteristics of the first historical state data as a first intermediate fault identification model, and determine the first target fault identification model based on the first intermediate fault identification model and the second historical state data; An intermediate fault identification model determined based on the dimensionless features in the historical data features is determined as a second intermediate fault identification model, and the second target fault identification model is determined based on the second intermediate fault identification model and the second historical state data.

5. The method according to claim 4, characterized in that The determining the first target fault identification model based on the first intermediate fault identification model and the second historical state data includes: Based on the preset dimension reduction model and the hierarchical clustering model, obtaining a plurality of second historical operating conditions corresponding to the second historical state data, and an actual fault type corresponding to each second historical operating condition; For each second historical operating condition, when there is a first historical operating condition that is the same as the second historical operating condition, inputting historical data features of second historical state data corresponding to the second historical operating condition into the first intermediate fault identification model, and outputting an output fault type corresponding to the second historical operating condition; Based on the difference between the output fault type and the actual fault type, obtaining an accuracy evaluation value of the first intermediate fault identification model; When the accuracy evaluation value is within a first preset accuracy range, determining the first intermediate fault identification model as the first target fault identification model; When the accuracy evaluation value is not within the first preset accuracy range, return to the step of inputting the historical data features of the first historical state data into the initial fault identification model, and continue to execute until the accuracy evaluation value is within the first preset accuracy range.

6. The method according to claim 4, characterized in that The determining the second target fault identification model based on the second intermediate fault identification model and the second historical state data includes: Based on the preset dimension reduction model and the hierarchical clustering model, obtaining a plurality of second historical operating conditions corresponding to the second historical state data, and an actual fault type corresponding to each second historical operating condition; For each second historical operating condition, when there is no first historical operating condition identical to the second historical operating condition, inputting the dimensionless features of the second historical state data corresponding to the second historical operating condition into the second intermediate fault identification model, and outputting the output fault type corresponding to the second historical operating condition; Based on the difference between the output fault type and the actual fault type, obtaining an accuracy evaluation value of the second intermediate fault identification model; When the accuracy evaluation value is within a second preset accuracy range, determining the second intermediate fault identification model as the second target fault identification model; When the accuracy evaluation value is not within the second preset accuracy range, return to the step of inputting the dimensionless features in the historical data features into the initial fault identification model, and continue to execute until the accuracy evaluation value is within the second preset accuracy range.

7. The method according to claim 1, characterized in that The step of inputting the current data feature into the first target fault identification model to obtain the current fault type based on the comparison result, or inputting the dimensionless feature in the current data feature into the second target fault identification model to obtain the current fault type, comprises: In the case where there is a data difference between the historical state data corresponding to the first historical working condition and the current state data and the difference is within a preset data range, inputting the current data feature into the first target fault identification model to obtain the current fault type; When there is no data difference between the historical state data corresponding to the first historical operating condition and the current state data and the data is within a preset data range, the dimensionless features in the current data features are input into the second target fault identification model to obtain the current fault type.

8. A device for determining equipment failure, characterized in that: The device comprises: A first acquisition module, used for acquiring the current state data of the target device and the current data characteristics of the current state data in real time; A second acquisition module is used to acquire a plurality of first historical operating conditions and compare the current state data with the historical state data corresponding to the first historical operating conditions; A third acquisition module is used to acquire a first target fault recognition model and a second target fault recognition model, wherein the first target fault recognition model is obtained by training an initial fault recognition model based on historical data features of the historical state data, and the second target fault recognition model is obtained by training the initial fault recognition model based on dimensionless features in the historical data features; An input module is used to input the current data features into the first target fault identification model to obtain the current fault type based on the comparison result, or input the dimensionless features in the current data features into the second target fault identification model to obtain the current fault type.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images