Equipment fault time prediction method and device, equipment, storage medium and computer program product

Abstract

The invention relates to the technical field of fault time prediction, and discloses an equipment fault time prediction method and device, equipment, a storage medium and a computer program product, and the method comprises the steps: generating an initial prediction parameter set according to the monitoring data of target equipment and part feature information; determining a prediction precision index of each parameter combination in the initial prediction parameter set; selecting an optimal parameter combination from the initial prediction parameter set based on a prediction precision index; and inputting the optimal parameter combination into a preset time sequence model, and outputting a fault time prediction result of the target equipment. According to the method, the optimal parameter combination of the current state of the to-be-predicted equipment in the time sequence model is determined through the monitoring data and the component feature information of the target equipment, so that accurate prediction of the equipment fault time is realized.

Description

Technical Field

[0001] The present application relates to the technical field of fault time prediction, and particularly to a method, device, equipment, storage medium and computer program product for predicting the fault time of a device. Background Art

[0002] Traditional fault prediction methods for power plant equipment usually rely on time series analysis, which is suitable for single-variable prediction. However, it requires a stationarity assumption for data, and trend separation is needed for multi-factor prediction, resulting in high computational complexity and limited accuracy. Summary of the Invention

[0003] The main objective of the present application is to provide a method, device, equipment, storage medium and computer program product for predicting the fault time of a device, aiming to solve the technical problems that existing fault time prediction methods require trend separation for multi-factor prediction, resulting in high computational complexity and limited accuracy.

[0004] To achieve the above objective, the present application proposes a method for predicting the fault time of a device, which includes:

[0005] Generating an initial prediction parameter set based on the monitoring data and component feature information of the target device;

[0006] Determining the prediction accuracy index of each parameter combination in the initial prediction parameter set;

[0007] Selecting the optimal parameter combination from the initial prediction parameter set based on the prediction accuracy index;

[0008] Inputting the optimal parameter combination into a preset time series model to output the prediction result of the fault time of the target device.

[0009] Optionally, the step of generating an initial prediction parameter set based on the monitoring data and component feature information of the target device includes:

[0010] Performing multi-dimensional fault feature encoding on the component feature information based on the monitoring data of the target device to generate an initial feature set;

[0011] Constructing the initial prediction parameter set of the target device based on the initial feature set;

[0012] Optionally, the initial feature set includes a sensor failure feature subset, a mechanical wear feature subset, and an abnormal vibration feature subset;

[0013] The step of constructing the initial prediction parameter set of the target device based on the initial feature set includes:

[0014] Weightedly reorganize the sensor failure feature subset to generate a failure feature sequence containing the installation location information of the target device;

[0015] Perform time-domain degradation analysis on the mechanical wear feature subset to generate a wear trend sequence;

[0016] Perform multi-scale entropy feature extraction on the abnormal vibration feature subset to generate a vibration mode sequence containing vibration energy distribution and mutation features;

[0017] Generate an initial prediction parameter set based on the failure feature sequence, the wear trend sequence, and the vibration mode sequence.

[0018] Optionally, the step of weightedly reorganizing the sensor failure feature subset to generate a failure feature sequence containing the installation location information of the target device includes:

[0019] Divide the sensor failure feature subset according to the functional areas of the target device, and perform feature cross-validation on the monitoring data of different functional areas to obtain a validation result;

[0020] Determine the data confidence weight of the sensor failure feature subset in each functional area based on the validation result;

[0021] Dynamically weightedly reorganize the sensor failure feature subset according to the data confidence weight to generate a failure feature sequence.

[0022] Optionally, the step of determining the prediction accuracy index of each parameter combination in the initial prediction parameter set includes:

[0023] Determine the prediction accuracy quantization value of each parameter combination through a preset evaluation function;

[0024] Set a time constraint condition through the maintenance cycle of the target device;

[0025] Adjust the prediction accuracy quantization value according to the time constraint condition to obtain the prediction accuracy index of each parameter combination.

[0026] Optionally, after the step of inputting the optimal parameter combination into a preset time series model and outputting the fault time prediction result of the target device, it further includes:

[0027] Calculate the error volatility of the fault time prediction result within a continuous prediction period;

[0028] Statistically analyze the interval deviation degree between the predicted time points and the actual fault points in the fault time prediction result;

[0029] Normalize the error volatility and interval deviation to generate an adjustment and correction coefficient, and adjust the time series model based on the adjustment and correction coefficient.

[0030] In addition, to achieve the above object, the present application also provides a device failure time prediction device, which includes:

[0031] A parameter acquisition module, configured to generate an initial prediction parameter set according to the monitoring data of the target device and the component feature information;

[0032] An index determination module, configured to determine the prediction accuracy index of each parameter combination in the initial prediction parameter set;

[0033] A combination selection module, configured to select an optimal parameter combination from the initial prediction parameter set based on the prediction accuracy index;

[0034] A time prediction module, configured to input the optimal parameter combination into a preset time series model and output the prediction result of the failure time of the target device.

[0035] In addition, to achieve the above object, the present application also provides a device for predicting the failure time of a device. The device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. The computer program is configured to implement the steps of the device failure time prediction method as described above.

[0036] In addition, to achieve the above object, the present application also provides a storage medium. The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the device failure time prediction method as described above are implemented.

[0037] In addition, to achieve the above object, the present application also provides a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, the steps of the device failure time prediction method as described above are implemented.

[0038] The present application discloses generating an initial prediction parameter set according to the monitoring data of the target device and the component feature information; determining the prediction accuracy index of each parameter combination in the initial prediction parameter set; selecting an optimal parameter combination from the initial prediction parameter set based on the prediction accuracy index; inputting the optimal parameter combination into a preset time series model and outputting the prediction result of the failure time of the target device. Since the present application determines the optimal parameter combination of the current state of the device to be predicted in the time series model through the monitoring data of the target device and the component feature information, accurate prediction of the device failure time is achieved. Description of the Drawings

[0039] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with this application, and are used together with the specification to explain the principles of this application.

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0041] Figure 1 It is a schematic flowchart of the first embodiment of the method for predicting the equipment failure time of this application;

[0042] Figure 2 It is a schematic flowchart of the second embodiment of the method for predicting the equipment failure time of this application;

[0043] Figure 3 It is a schematic flowchart of the third embodiment of the method for predicting the equipment failure time of this application;

[0044] Figure 4 It is a schematic diagram of the module structure of the device for predicting the equipment failure time in the embodiments of this application;

[0045] Figure 5 It is a schematic diagram of the device structure of the hardware operating environment involved in the method for predicting the equipment failure time in the embodiments of this application.

[0046] The implementation, functional features, and advantages of the purpose of this application will be further described with reference to the embodiments and the accompanying drawings. Specific Embodiments

[0047] It should be understood that the specific embodiments described here are only used to explain the technical solutions of this application and are not used to limit this application.

[0048] To better understand the technical solutions of this application, the following will be described in detail in combination with the specification drawings and specific embodiments.

[0049] The main solution of the embodiments of this application is: generating an initial prediction parameter set according to the monitoring data and component feature information of the target device; determining the prediction accuracy indicators of each parameter combination in the initial prediction parameter set; selecting the optimal parameter combination from the initial prediction parameter set based on the prediction accuracy indicators; inputting the optimal parameter combination into a preset time series model, and outputting the prediction result of the failure time of the target device.

[0050] Current device fault prediction technologies generally adopt single - dimension data analysis, such as vibration spectrum or temperature threshold monitoring, etc., which have the defect of insufficient ability to fuse multi - source heterogeneous data. In traditional methods, when processing time - series data through the LSTM network, the spatial topological relationship of the device fails to be effectively integrated, resulting in a pseudo - correlation misjudgment rate of up to 25% for the failure characteristics of sensors; while using fixed weights to fuse multiple parameters, the prediction error fluctuates greatly when the working conditions change suddenly. The existing device fault time prediction technologies mainly have the following bottlenecks: 1. Mechanical wear analysis is mostly limited to time - domain statistics, lacking joint modeling of time - frequency domain degradation trends, resulting in an excessive missed detection rate of early pitting of bearings; 2. Abnormal vibration detection relies on fixed - band energy analysis, with a long response delay for impact faults; 3. The parameter optimization process is static, unable to adapt to the feature drift caused by device aging, and the model needs to be manually calibrated weekly, increasing the operation and maintenance costs. In addition, traditional three - dimensional visualization methods only display the current state and fail to realize the dynamic deduction of the fault propagation path, making the maintenance decision - making lack foresight.

[0051] This application provides an intelligent prediction system based on multi - dimensional coupling of space, time - series, and working conditions. By structurally processing the monitoring data and component characteristics, a complete technical closed - loop from parameter generation to prediction output is established to meet the real - time requirements of industrial sites.

[0052] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, intelligent decision - making, and program running functions, such as a power plant intelligent management platform, etc., or an electronic device capable of realizing the above functions. Hereinafter, taking the power plant equipment detection system as an example, this embodiment and the following embodiments will be described.

[0053] Based on this, the embodiment of this application provides a method for predicting the device fault time, referring to Figure 1 , Figure 1 is a schematic flowchart of the first embodiment of the device fault time prediction method of this application.

[0054] In this embodiment, the device fault time prediction method includes:

[0055] Step S10, generating an initial prediction parameter set according to the monitoring data and component feature information of the target device.

[0056] It should be noted that the target device refers to the specific device object for which the fault time needs to be predicted. In an industrial scenario, the target device can be a specific rotating machine (such as a steam turbine, compressor), production line equipment (such as a numerically controlled machine tool), or critical infrastructure (such as power transmission and transformation equipment), etc.

[0057] The monitoring data of the target device refers to the device operation status data collected in real time through various sensors, such as temperature, pressure, vibration frequency, etc.; the component characteristic information is about the inherent attributes and characteristics of each component of the target device, such as the material of the component, design parameters, service life, etc. The initial prediction parameter set is composed of a series of parameters extracted, transformed, and combined from the monitoring data and component characteristic information, and these parameters will be used as the basic input for subsequent failure time prediction.

[0058] It can be understood that the initial prediction parameter set contains feature information related to device failures extracted from multiple dimensions, and this information can more comprehensively and accurately reflect the operation status of the device.

[0059] It should be understood that during the data collection process, attention should be paid to the installation location and calibration of the sensors to ensure the accuracy of the data. For example, the accelerometer should be installed at the key vibration parts of the device and calibrated regularly. And preprocess the collected data, such as denoising and filtering.

[0060] In one example, the monitoring data of a power plant steam turbine includes bearing vibration signals, lubricating oil temperature, rotor speed, etc. Through multi-dimensional fault feature coding, features such as kurtosis (reflecting impact faults) and spectral entropy (measuring signal complexity) can be extracted from the bearing vibration signals, and features such as temperature change rate can also be extracted from the lubricating oil temperature data to form an initial feature set.

[0061] Step S20, determine the prediction accuracy index for each parameter combination in the initial prediction parameter set.

[0062] It should be noted that the prediction accuracy index is a quantitative index used to measure the difference between the prediction result and the actual value, such as root mean square error, mean absolute percentage error, etc.

[0063] It can be understood that when determining the prediction accuracy index, the influence of different parameter combinations on the failure time prediction result needs to be considered. The prediction accuracy index can be a combination of multiple evaluation indexes, such as root mean square error, mean absolute error, prediction accuracy rate, etc. In the actual implementation process, each parameter combination can be quantitatively evaluated through a preset evaluation function to obtain the prediction accuracy quantization value of each parameter combination.

[0064] Furthermore, in order to adjust the quantization value according to the time constraint conditions to obtain the prediction accuracy index, ensure that the index can more accurately reflect the prediction performance of the parameter combination within the actual device maintenance cycle, thereby improving the accuracy of selecting the optimal parameter combination. The step S20 can include:

[0065] Determine the prediction accuracy quantization value of each parameter combination through a preset evaluation function; set a time constraint condition according to the maintenance cycle of the target device; adjust the prediction accuracy quantization value according to the time constraint condition to obtain the prediction accuracy index of each parameter combination.

[0066] It should be noted that the preset evaluation function is a mathematical function that is set in advance to quantitatively evaluate the accuracy of different parameter combinations in predicting the device failure time. The design of the preset evaluation function will consider the difference between the prediction result and the actual failure time, such as the common mean square error, mean absolute error, etc. The prediction accuracy quantization value is a specific value obtained by calculating each parameter combination using the preset evaluation function, which shows the accuracy of the parameter combination in predicting the device failure time. The maintenance cycle is the time interval between the completion of one maintenance of the target device and the start of the next maintenance. The time constraint condition is a prediction time range limit set according to the device maintenance plan or the degree of impact of the failure, to avoid the situation where the predicted failure time conflicts with the maintenance arrangement.

[0067] It can be understood that the time constraint condition can be dynamically adjusted according to the importance of the device and the consequences of the failure. For example, the prediction time constraint for critical devices (such as boilers) can be more stringent.

[0068] Step S30, select the optimal parameter combination from the initial prediction parameter set based on the prediction accuracy index.

[0069] It can be understood that when selecting the optimal parameter combination, it is necessary to comprehensively consider the prediction accuracy index and the actual application requirements. For example, if the consequences of the device failure are very serious, then the parameter combination with the highest prediction accuracy may be preferentially selected, even if its computational complexity is relatively high; if the device needs to perform real-time fault prediction, then a parameter combination with higher computational efficiency can be selected while taking into account a certain prediction accuracy.

[0070] In some embodiments of the present application, intelligent optimization algorithms such as genetic algorithms and particle swarm optimization algorithms can be used to search for the optimal parameter combination from the initial prediction parameter set. These algorithms can efficiently find the optimal solution in a complex parameter space, improving the accuracy and efficiency of parameter selection.

[0071] In an example, for the fault prediction of a power plant generator, a genetic algorithm is used to select the optimal combination from the initial prediction parameter set including vibration characteristics, temperature characteristics, and pressure characteristics. First, the parameter combination is represented as a chromosome, for example, binary encoding. Then, chromosomes with high fitness are selected according to the prediction accuracy index; it is prompted to generate new chromosomes through crossover operations and randomly mutate to maintain the diversity of the parameter combinations. After several generations of evolution, a set of parameter combinations is obtained, and its prediction accuracy index reaches the optimal within the maintenance cycle.

[0072] Step S40: Input the optimal parameter combination into a preset time series model, and output the predicted fault time result of the target device.

[0073] It should be understood that the preset time series model can be various types of models, such as ARIMA model, LSTM model, Prophet model, etc. These models can analyze and predict time series data, and infer future trends based on the patterns of historical data.

[0074] It can be understood that in some embodiments of the present application, to improve the prediction accuracy and adaptability of the model, the model can be updated and optimized regularly. For example, when new monitoring data is obtained, retrain the model, or adjust the parameters of the model according to the new data.

[0075] In this embodiment, an initial prediction parameter set is generated according to the monitoring data and component feature information of the target device; the prediction accuracy indicators of each parameter combination in the initial prediction parameter set are determined; the optimal parameter combination is selected from the initial prediction parameter set based on the prediction accuracy indicators; the optimal parameter combination is input into a preset time series model, and the predicted fault time result of the target device is output. Since the present application determines the optimal parameter combination of the current state of the device to be predicted in the time series model through the monitoring data and component feature information of the target device, accurate prediction of the device fault time is realized.

[0076] Refer to Figure 2 , Figure 2 FIG.

[0077] In the second embodiment, step S10 includes:

[0078] Step S101: Perform multi-dimensional fault feature encoding on the component feature information based on the monitoring data of the target device to generate an initial feature set.

[0079] It should be noted that multi-dimensional fault feature encoding refers to analyzing and encoding the component feature information from multiple different perspectives to extract more comprehensive and representative fault features. For example, the vibration signal can be analyzed from multiple dimensions such as time domain, frequency domain, and time-frequency domain, and various features such as mean value, variance, kurtosis, and spectral entropy can be extracted.

[0080] In some embodiments of the present application, deep learning techniques can be used for multi-dimensional fault feature encoding. For example, a convolutional neural network (CNN) is used to extract features from vibration signals. The CNN can automatically learn the deep features in the data, improving the efficiency and accuracy of feature extraction.

[0081] In one example, for the boiler equipment in a power plant, the CNN is used to extract features from its vibration signals, and automatically learn the deep features that can reflect faults such as furnace coking and tube explosion. First, data such as vibration signals are converted into a format suitable for the input of the CNN (such as a two-dimensional matrix). Then, local features are extracted through convolutional kernels. In the fully connected layer, the extracted features are mapped to the fault feature space. Finally, features of different dimensions (such as vibration, temperature, pressure) are concatenated or weighted and fused to form a comprehensive fault feature, obtaining an initial feature set.

[0082] Step S102, construct an initial prediction parameter set for the target device based on the initial feature set.

[0083] It can be understood that when constructing the initial prediction parameter set, each feature in the initial feature set needs to be reasonably combined and screened to form a parameter set that can effectively reflect the fault state of the device. For example, sensor failure features, mechanical wear features, abnormal vibration features, etc. can be combined to form a comprehensive parameter set.

[0084] It should be understood that when there are many parameters in the initial feature set, dimensionality reduction techniques such as principal component analysis (PCA) and linear discriminant analysis (LDA) can also be used to process the initial feature set, reduce the dimensionality of the features, and improve the computational efficiency and prediction accuracy of the subsequent model.

[0085] Furthermore, in order to extract spatial topology, time-domain degradation, and frequency-domain mutation features through classified feature processing, and synthesize these features to generate an initial prediction parameter set, which can more comprehensively and accurately describe the fault features of the device. The step S102 may include:

[0086] Perform weighted recombination on the sensor failure feature subset to generate a failure feature sequence containing the installation location information of the target device; perform time-domain degradation analysis on the mechanical wear feature subset to generate a wear trend sequence; perform multi-scale entropy feature extraction on the abnormal vibration feature subset to generate a vibration mode sequence containing vibration energy distribution and mutation features; generate an initial prediction parameter set according to the failure feature sequence, the wear trend sequence, and the vibration mode sequence.

[0087] It should be noted that the sensor failure feature subset is anomaly features extracted from raw signals collected by sensors such as vibration and temperature, such as peak values, kurtosis, and energy distribution, reflecting the sensor's inherent health (e.g., decreased sensitivity, noise interference). Installation location information reflects the topological relationship of the equipment's physical structure, such as the location of bearings at key support points in the rotor system. This information is used to distinguish reliability differences in data from different regions. The mechanical wear feature subset is a collection of time-domain signal features reflecting the physical wear of equipment components, such as the rate of change of kurtosis and the energy fraction of defect frequencies. Time-domain degradation analysis uses statistical models to fit the time-varying trend of wear rate. The wear trend series is a continuous curve with time as the horizontal axis and wear rate as the vertical axis, used to predict remaining service life. The abnormal vibration feature subset includes non-stationary signal features such as shock pulse (SPI) and high-frequency resonant band energy. Multiscale entropy calculates signal complexity at different time scales to measure the mutation characteristics of the degradation process. The vibration pattern series characterizes the mutation patterns of the vibration signal in the time-frequency domain.

[0088] In one example, when predicting bearing failures in a power plant steam turbine unit, the equipment is first divided into multiple functional areas (such as the bearing seat area and the gearbox area), and independent sensors are deployed in each area. Correlation analysis is performed on the data from multiple areas under the same operating conditions, and the confidence weight of the data in each area is calculated:

[0089]

[0090] Among them, σ i and σ j The data variance of the sensor is divided into regions i and j. W is the position sensitivity coefficient, which can be preset according to the equipment manual. For example, the weight of the bearing seat area is set to 0.6. The formula for weighting the sensor failure feature subset is as follows:

[0091]

[0092] Among them, S i is the standardized regional characteristic value, F miss This is the reorganized failure signature sequence. By combining sensor data from various locations, we can eliminate misjudgments caused by single sensor failures (such as sensor noise interference in a specific area), improving feature reliability. For example, a turbine bearing seat sensor might generate a false alarm due to vibration attenuation, while the gearbox sensor verification data is normal. Weighted reorganization can eliminate outliers and prevent false alarms.

[0093] When performing time-domain degradation analysis on a subset of mechanical wear features, the kurtosis change rate (KR) of the vibration signal is first calculated:

[0094]

[0095] Among them, K tK is the current kurtosis, and K0 is the initial value of kurtosis.

[0096] Next, extract the energy proportion of the bearing inner ring defect frequency (BPFI):

[0097]

[0098] E BPFI represents the vibration signal energy distribution density P(f) based on the bearing component characteristic frequency at frequency f. is the integral of the vibration energy within the frequency range BPFI ± 5 centered on BPFI, representing the total energy within this specific frequency range.

[0099] Using the LSTM model trained with historical data, take KR and E BPFI as inputs and output the wear rate prediction value R(t):

[0100] R(t) = a·e bt + c

[0101] According to the mechanical wear characteristic subset, early tiny wear signs can be captured (such as the energy proportion of BPFI rising from 5% to 15%), avoiding sudden failures.

[0102] When analyzing the abnormal vibration characteristic subset, signal preprocessing is required. First, perform mean removal and normalization on the vibration signal. Then, use wavelet transform to decompose the signal into multiple scale components, and calculate the sample entropy (SE) for each scale component:

[0103]

[0104] where m is the embedding dimension, used to construct the vector dimension of the phase space, r is the tolerance threshold, defining the criterion for judging data similarity, and N is the data length of the time series. C m (r) and C m+1 (r) are respectively the proportions of the number of sample pairs satisfying the distance between two sequences less than or equal to r to the total number of sample pairs when the embedding dimensions are m and m + 1.

[0105] Finally, weight and combine the entropy values of each scale to generate a vibration mode sequence:

[0106]

[0107] where α k is the scale weight, determined by the mutual information method. n is the number of scale components.

[0108] After obtaining the failure feature sequence, wear trend sequence, and vibration mode sequence, a comprehensive input vector that fuses multi-source features is used to output the initial prediction parameter set for the time series model. The optimal feature combination can also be screened through grid search and genetic algorithm to reduce the risk of model overfitting.

[0109] First, align the three subsequences in time and splice them into a high-dimensional feature vector X:

[0110] X = [F miss (t), R(t), V(t)]

[0111] Next, define the candidate parameter space (such as the number of LSTM layers {2, 3}, learning rate {0.001, 0.01}). Use the cross-entropy loss function to evaluate the model performance L:

[0112]

[0113] where y i is the actual failure time, is the predicted value.

[0114] Furthermore, in order to determine the data confidence weight based on the verification result, so that the data in different functional areas can be reasonably weighted during the recombination process, the differences in sensor data in different functional areas can be fully considered, and the quality of the failure feature sequence can be improved. The step of weighted recombination of the sensor failure feature subset to generate a failure feature sequence including the installation position information of the target device may include:

[0115] Divide the sensor failure feature subset according to the functional area of the target device, and perform feature cross-verification on the monitoring data in different functional areas to obtain the verification result; determine the data confidence weight of the sensor failure feature subset in each functional area based on the verification result; perform dynamic weighted recombination on the sensor failure feature subset according to the data confidence weight to generate a failure feature sequence.

[0116] It can be understood that according to the functional structure of the target device (such as mechanical modules, system units), the sensor data is classified into different subsets. When performing feature cross-verification on the monitoring data in different functional areas, it can be analyzed by setting verification rules. For example, when the device is operating normally, the temperature change in the target area should match the parameter change trend in the associated area; or compare the spatio-temporal correlation of sensor data in different areas. For example, when the temperature of the steam turbine bearing rises, check whether the corresponding vibration sensor data is synchronously abnormal.

[0117] It should be understood that when quantifying the reliability of data in each region according to the cross-validation results, a high weight is assigned to the region with high reliability, and a low weight is assigned otherwise. By weighted combining the subsets of sensor failure characteristics in each region based on the confidence weights, a sequence that comprehensively reflects the overall failure state of the device can be generated.

[0118] In this embodiment, multi-dimensional fault feature encoding is performed on the component feature information based on the monitoring data of the target device to generate an initial feature set; an initial prediction parameter set of the target device is constructed based on the initial feature set. By performing multi-dimensional fault feature encoding on the component feature information based on the monitoring data to generate an initial feature set, the fault features of the device can be more comprehensively and deeply mined, improving the richness and effectiveness of the features.

[0119] Refer to Figure 3 , Figure 3 is a schematic flowchart of the third embodiment of the device fault time prediction method of the present application. Based on the above second embodiment, the third embodiment of the device fault time prediction method of the present application is proposed.

[0120] In the third embodiment, after the step S40, it further includes:

[0121] Step S501, calculating the error volatility of the fault time prediction result within a continuous prediction period.

[0122] It should be noted that the error volatility is the degree of fluctuation of the prediction error within a continuous prediction period, and can be used to evaluate the stability of the model prediction. The error volatility reflects the fluctuation of the prediction result within a period of time. Calculating the error volatility can understand the stability of the model prediction.

[0123] It can be understood that the degree of deviation of the error of each prediction time point within the continuous prediction period from the average error can be calculated, and then the error volatility can be represented in the form of a standard deviation.

[0124] In an example, the error volatility is represented in the form of a standard deviation, that is, a mathematical measure of the degree of deviation of the error from the average error. The calculation formula is:

[0125]

[0126] where e t is the prediction error at the t-th time point, is the average error, and N is the total number of time points.

[0127] For each prediction time point t, it is necessary to calculate the predicted fault time and the error with the actual fault time y i to obtain e t and the standard deviation σ.

[0128] Calculate the average error e within each window t and the standard deviation. Finally, take the average of the standard deviations of all windows as the total error volatility (Error Volatility, EV):

[0129]

[0130] where K is the total number of windows.

[0131] Step S502: Statistically analyze the deviation degree between the predicted time point and the actual fault point in the fault time prediction result.

[0132] It should be noted that the deviation degree refers to the time difference between the predicted time point and the actual fault point. Statistically analyzing the deviation degree can evaluate the accuracy and timeliness of the model prediction. For example, if the predicted time point is earlier than the actual fault point by a certain period, it indicates that the model can give an early warning; if the predicted time point lags behind the actual fault point, it means that the model's prediction has a delay.

[0133] It can be understood that when statistically analyzing the deviation degree, the timestamp of the actual fault point can be ensured to be accurate through device logs or manual confirmation. For the case of early prediction, it is necessary to further analyze whether the early time is sufficient for maintenance operations.

[0134] Step S503: Normalize the error volatility and the deviation degree to generate an adjustment and correction coefficient, and adjust the time series model based on the adjustment and correction coefficient.

[0135] It should be noted that the normalization process is to convert the error volatility and the deviation degree with different dimensions into comparable values for comprehensive evaluation. After generating the adjustment and correction coefficient, the parameters of the time series model can be adjusted according to this coefficient, such as adjusting the weights and learning rates of the model, so that the model can better adapt to the actual operating conditions of the device.

[0136] It can be understood that after normalizing the error volatility and the deviation degree respectively, weights can be set in combination with historical experience to calculate the comprehensive correction coefficient. If the comprehensive correction coefficient is greater than the preset value, it indicates that the model is unstable and has a large deviation, and the LSTM hyperparameters such as the learning rate and the number of hidden layer nodes are re-optimized using grid search. If the comprehensive correction coefficient is less than the preset value and the model is overfitting, noise data or synthetic data can be added to expand the training set.

[0137] In this embodiment, calculate the error volatility of the predicted fault time result within a continuous prediction period; count the deviation degree between the predicted time point and the actual fault point in the predicted fault time result; normalize the error volatility and the deviation degree to generate an adjustment and correction coefficient, and adjust the time series model based on the adjustment and correction coefficient. Normalizing the error volatility and the deviation degree to generate an adjustment and correction coefficient and adjusting the time series model based on this can enable the model to continuously learn and adapt to the actual operation conditions of the device, improve the prediction performance and adaptability of the model, and thus maintain a high prediction accuracy during the long-term prediction process.

[0138] It should be noted that the above examples are only for understanding the present application and do not constitute a limitation to the device fault time prediction method of the present application. Based on this technical concept, more forms of simple transformations are within the protection scope of the present application.

[0139] The present application also provides a device fault time prediction device. Please refer to Figure 4 , the device fault time prediction device includes:

[0140] A parameter acquisition module 10, configured to generate an initial prediction parameter set according to the monitoring data and component feature information of the target device;

[0141] An index determination module 20, configured to determine the prediction accuracy index of each parameter combination in the initial prediction parameter set;

[0142] A combination selection module 30, configured to select an optimal parameter combination from the initial prediction parameter set based on the prediction accuracy index;

[0143] A time prediction module 40, configured to input the optimal parameter combination into a preset time series model and output the predicted fault time result of the target device.

[0144] The device fault time prediction device provided by the present application adopts the device fault time prediction method in the above embodiment, and can solve the technical problems that the existing fault time prediction methods need to perform trend separation in multi-factor prediction, have high computational complexity, and limited accuracy. Compared with the prior art, the beneficial effects of the device fault time prediction device provided by the present application are the same as those of the device fault time prediction method provided by the above embodiment, and other technical features in the device fault time prediction device are the same as the features disclosed in the above embodiment method, and will not be elaborated here.

[0145] The present application provides a device failure time prediction device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the device failure time prediction method in the first embodiment above.

[0146] Reference is made below to Figure 5 , which shows a schematic structural diagram of a device failure time prediction device suitable for implementing the embodiments of the present application. The device failure time prediction device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistant), PADs (Portable Application Description: tablet computers), PMPs (Portable Media Player), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The device failure time prediction device shown is only an example and should not impose any limitations on the functions and scope of use of the embodiments of the present application.

[0147] As Figure 5 shown, the device failure time prediction device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in the read-only memory 1002 or a program loaded from the storage device 1003 into the random access memory 1004. In the random access memory 1004, various programs and data required for the operation of the device failure time prediction device are also stored. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. The input / output interface 1006 is also connected to the bus. Generally, the following systems may be connected to the input / output interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the device failure time prediction device to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows a device failure time prediction device having various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be alternatively implemented or had.

[0148] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes program code for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by a processing device 1001, the above functions defined in the methods of the embodiments disclosed in the present application are executed.

[0149] The equipment failure time prediction device provided by the present application adopts the equipment failure time prediction method in the above embodiment, and can solve the technical problems that the existing failure time prediction methods need to perform trend separation in multi-factor prediction, have high computational complexity, and limited accuracy. Compared with the prior art, the beneficial effects of the equipment failure time prediction device provided by the present application are the same as those of the equipment failure time prediction method provided by the above embodiment, and other technical features in the equipment failure time prediction device are the same as those disclosed in the method of the previous embodiment, and will not be elaborated here.

[0150] It should be understood that the various parts disclosed in the present application can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0151] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0152] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the equipment failure time prediction method in the above embodiment.

[0153] The computer-readable storage medium provided by this application can be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) or flash memory, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, device, or component. The program code contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0154] The above computer-readable storage medium can be included in the device failure time prediction device; or it can exist independently without being assembled into the device failure time prediction device.

[0155] The above computer-readable storage medium carries one or more programs. When the above one or more programs are executed by the device failure time prediction device, the device failure time prediction device is caused to execute the device failure time prediction method described above.

[0156] Computer program code for performing the operations of this application can be written in one or more programming languages or combinations thereof. The above programming languages include object-oriented programming languages - such as Java, Smalltalk, C++; and also include conventional procedural programming languages - such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network - including a local area network (LAN) or a wide area network (WAN) - or can be connected to an external computer (for example, by using an Internet service provider to connect through the Internet).

[0157] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a part of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0158] The modules described in the embodiments of the present application can be implemented in software or in hardware. In some cases, the name of the module does not constitute a limitation on the unit itself.

[0159] The readable storage medium provided by the present application is a computer-readable storage medium, and the computer-readable storage medium stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned device failure time prediction method, which can solve the technical problems that the existing failure time prediction methods need to perform trend separation in multi-factor prediction, have high computational complexity, and limited accuracy. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as those of the device failure time prediction method provided by the above embodiments, and will not be elaborated here.

[0160] The present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of the device failure time prediction method as described above.

[0161] The computer program product provided by the present application can solve the technical problems that the existing failure time prediction methods need to perform trend separation in multi-factor prediction, have high computational complexity, and limited accuracy. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the device failure time prediction method provided by the above embodiments, and will not be elaborated here.

[0162] The above are only some embodiments of the present application, and thus do not limit the patent scope of the present application. Any equivalent structural transformation made under the technical concept of the present application by using the content of the specification and drawings of the present application, or any direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. A method for predicting equipment failure time, characterized in that, The device failure time prediction method includes: Generating an initial prediction parameter set according to the monitoring data of the target device and the component characteristic information; Determining the prediction accuracy index of each parameter combination in the initial prediction parameter set; Selecting the optimal parameter combination from the initial prediction parameter set based on the prediction accuracy index; Inputting the optimal parameter combination into a preset time series model and outputting the failure time prediction result of the target device.

2. The method for predicting the equipment failure time according to claim 1, wherein, The step of generating an initial prediction parameter set according to the monitoring data of the target device and the component characteristic information includes: Performing multi-dimensional fault feature encoding on the component characteristic information based on the monitoring data of the target device to generate an initial feature set; Constructing an initial prediction parameter set of the target device based on the initial feature set.

3. The method for predicting equipment failure time according to claim 2, wherein, The initial feature set includes a sensor failure feature subset, a mechanical wear feature subset, and an abnormal vibration feature subset; The step of constructing an initial prediction parameter set of the target device based on the initial feature set includes: Performing weighted recombination on the sensor failure feature subset to generate a failure feature sequence including the installation position information of the target device; Performing time-domain degradation analysis on the mechanical wear feature subset to generate a wear trend sequence; Performing multi-scale entropy feature extraction on the abnormal vibration feature subset to generate a vibration mode sequence including vibration energy distribution and mutation characteristics; Generating an initial prediction parameter set according to the failure feature sequence, the wear trend sequence, and the vibration mode sequence.

4. The method for predicting the equipment failure time according to claim 3, wherein, The step of performing weighted recombination on the sensor failure feature subset to generate a failure feature sequence including the installation position information of the target device includes: Dividing the sensor failure feature subset according to the functional area of the target device, and performing feature cross-validation on the monitoring data of different functional areas to obtain a validation result; Determining the data confidence weight of the sensor failure feature subset in each functional area based on the validation result; Performing dynamic weighted recombination on the sensor failure feature subset according to the data confidence weight to generate a failure feature sequence.

5. The method for predicting the equipment failure time according to claim 1, wherein The step of determining the prediction accuracy index of each parameter combination in the initial prediction parameter set includes: Determining the prediction accuracy quantization value of each parameter combination through a preset evaluation function; Setting a time constraint condition through the maintenance cycle of the target device; Adjusting the prediction accuracy quantization value according to the time constraint condition to obtain the prediction accuracy index of each parameter combination.

6. The method for predicting the equipment failure time according to any one of claims 1 to 5, characterized in that, After the step of inputting the optimal parameter combination into a preset time series model and outputting the failure time prediction result of the target device, it further includes: Calculating the error volatility of the failure time prediction result within a continuous prediction period; Counting the interval deviation degree between the predicted time point and the actual failure point in the failure time prediction result; Normalizing the error volatility and the interval deviation degree to generate an adjustment and correction coefficient, and adjusting the time series model based on the adjustment and correction coefficient.

7. A device failure time prediction device, characterized in that, The device includes: A parameter acquisition module for generating an initial prediction parameter set according to the monitoring data of the target device and the component characteristic information; An index determination module, configured to determine the prediction accuracy index of each parameter combination in the initial prediction parameter set; A combination selection module, configured to select an optimal parameter combination from the initial prediction parameter set based on the prediction accuracy index; A time prediction module, configured to input the optimal parameter combination into a preset time series model and output the prediction result of the failure time of the target device.

8. A device failure time prediction device, characterized in that, The device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the device failure time prediction method according to any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the device failure time prediction method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps of the device failure time prediction method according to any one of claims 1 to 6.

Images