Method and device for determining effectiveness of diesel generator set fault preventive test

Abstract

The present invention discloses a method and device for determining the effectiveness of a diesel generator set fault preventive test. The method comprises obtaining vibration data, temperature data, vibration time series, temperature time series and test data of the diesel generator set; performing denoising and dimensionality reduction processing on the vibration data and temperature data to obtain a fused feature matrix, and inputting it into a fault classification model to obtain the fault type; performing time series alignment on the vibration time series and the temperature time series to obtain time series data, and inputting it into a prediction model to obtain vibration prediction values ​​and temperature prediction values; performing deviation correction on the test data, and inputting the corrected data into a pre-built test prediction model to obtain prediction test conditions; and performing generator fault prevention operations and prevention effectiveness evaluation according to the fault type, vibration prediction value, temperature prediction value and prediction test conditions. The method solves the problem of being unable to accurately perform fault diagnosis and preventive maintenance of diesel generators.

Description

Technical Field

[0001] The invention relates to the technical field of generator fault prevention, and in particular to a method and a device for determining the effectiveness of a diesel generator set fault preventive test. Background Art

[0002] At present, diesel generator sets play a vital role in various facilities, especially as emergency backup power supply in the case of unstable power supply or emergency power outage. Its reliability is essential to ensure the normal operation of key equipment. However, since diesel generator sets are composed of a series of complex mechanical components, there are not only close physical connections between these components, but also complex coupling relationships, which makes fault diagnosis and preventive maintenance a major technical challenge. First, the interaction between the components may cause the problem of a single point of failure to spread to the entire system, increasing the difficulty of identifying and locating the specific source of the fault. Secondly, factors such as stress changes, wear and tear, and material aging under different working environments will affect the working state of the generator set, further increasing the complexity of maintenance work. In addition, with the increase of service life, some potential problems inside the generator set may not appear immediately, but gradually accumulate, eventually leading to sudden failures. Therefore, it is necessary to find a way to combine traditional mechanical engineering knowledge with modern information technology to more effectively solve the fault diagnosis and preventive maintenance problems faced by diesel generator sets.

[0003] In traditional diesel generator maintenance, technicians first put on personal protective equipment and prepare the necessary tools. They will begin with a visual inspection of the generator set, looking for any obvious physical damage or leaks, and ensuring that all connections are properly tightened. Next, technicians will check the oil level, the cleanliness of the fuel filter, and the condition of the coolant, and adjust or replace them if necessary. The electrical system inspection includes disconnecting the negative terminal of the battery to avoid accidental starting, then cleaning the battery terminals and testing their voltage, and using a multimeter to detect the resistance value in the circuit to confirm that there are no short circuits or open circuits. For mechanical parts, technicians will manually rotate the generator shaft, check the condition of the fan blades and radiator, and adjust the valve clearance or replace worn parts as needed. After completing these steps, reconnect the battery, turn on the fuel supply and try to start the generator set, observing the various indicator lights and instrument readings during the startup process. After the generator set is running stably, record key parameters such as speed and temperature, and check the exhaust. Finally, shut down the engine after running for a period of time at no load, record all inspection results in detail, and fill out a maintenance report. Such a set of processes helps technicians identify basic problems and take appropriate measures.

[0004] The defects in the maintenance process of diesel generator sets are mainly reflected in the lack of foresight for potential failures, the inability to accurately identify early wear of internal components or subtle cracks in hidden parts, and the lack of systematic analysis tools to predict failure trends, resulting in the inability to accurately perform fault diagnosis and preventive maintenance of diesel generators. Summary of the invention

[0005] The invention provides a method and a device for determining the effectiveness of a diesel generator set fault preventive test, so as to solve the problem of accurately performing fault diagnosis and preventive maintenance of the diesel generator.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a method for determining the effectiveness of a diesel generator set fault preventive test, comprising: Obtain vibration data, temperature data, vibration time series and temperature time series of the diesel generator set, and use the Monte Carlo simulation method to randomly sample the test conditions to obtain test data; Using Dobesi wavelet to preprocess the vibration data and the temperature data to obtain vibration characteristics and temperature characteristics; Perform data merging and dimensionality reduction processing according to the vibration characteristics and the temperature characteristics to obtain a fusion feature matrix; Inputting the fused feature matrix into a pre-built fault classification model to obtain the fault type; Using a dynamic time warping algorithm to perform time series alignment on the vibration time series and the temperature time series to obtain time series data; Inputting the time series data into a pre-trained prediction model to obtain a vibration prediction value and a temperature prediction value; Using a Kalman filter algorithm to correct the deviation of the test data, and inputting the corrected data into a pre-built test prediction model to obtain predicted test conditions; Generator fault prevention operations and prevention effectiveness evaluation are performed based on the fault type, the vibration prediction value, the temperature prediction value and the prediction test conditions.

[0007] In an optional implementation, the using of Dobesi wavelet to preprocess the vibration data and the temperature data to obtain vibration characteristics and temperature characteristics includes: Decomposing the vibration data and the temperature data into wavelet transform coefficients of different frequency components by using Dobesi wavelet, and removing high-frequency data in the wavelet transform coefficients based on a soft threshold method to obtain denoised wavelet coefficients; Reconstructing denoised vibration data and denoised temperature data through inverse wavelet transform according to the denoised wavelet coefficients; The denoised vibration data and the denoised temperature data are subjected to time domain and frequency domain feature extraction to obtain vibration features and temperature features.

[0008] In an optional implementation, the data merging and dimensionality reduction processing is performed according to the vibration feature and the temperature feature to obtain a fusion feature matrix, including: Combining the vibration features and the temperature features into a feature matrix; The feature matrix is ​​subjected to dimensionality reduction processing by principal component analysis to obtain a low-dimensional feature matrix; The low-dimensional feature matrix is ​​cleaned according to a preset cleaning threshold to obtain a fused feature matrix.

[0009] In an optional implementation, the process of constructing the fault classification model includes: Divide the pre-stored historical fusion feature matrix and fault feature matrix into a test set, a training set and a validation set; Select support vector regression as the initial model, radial basis function as the kernel function, and minimize the loss function as the original optimization problem; The original optimization problem is converted into a Lagrange multiplier solution according to the radial basis function, and the dual form of the original problem is solved by quadratic programming to obtain a minimized objective function, a regularization parameter, and a kernel function parameter; Using a cross-validation method to evaluate the performance indicators of the initial model on the validation set, wherein the performance indicators include mean square error and mean absolute error; When the performance indicator does not meet the preset performance standard, adjusting the regularization parameter and the kernel function parameter, and resolving the original optimization problem; When the performance indicator meets the preset performance standard, a fault classification model of the corresponding parameters is obtained.

[0010] In an optional implementation, the using a dynamic time warping algorithm to perform time series alignment on the vibration time series and the temperature time series to obtain time series data includes: constructing a distance matrix based on the vibration time series and the temperature time series; Constructing an initial cumulative distance matrix, and optimizing the initial cumulative distance matrix according to the distance matrix to obtain a cumulative distance matrix; Backtracking according to the cumulative distance matrix to find the best path with the minimum cumulative distance; Calculating a time offset according to the optimal path, adjusting a vibration time series and a temperature time series according to the offset, and combining the vibration time series and the temperature time series with the corresponding vibration data and the temperature data to obtain time series data; Among them, the elements in the distance matrix are calculated by the following formula: Among them, the construction of the initial cumulative distance matrix is ​​as follows: Among them, the elements in the cumulative distance matrix are calculated by the following formula: in, Indicates the distance matrix The elements at the position, Indicates the vibration time series data, Indicates the temperature time series data, represents the initial cumulative distance matrix, Indicates the cumulative distance matrix The elements at the position, Indicates the cumulative distance matrix The elements at the position, Indicates the cumulative distance matrix The element at position.

[0011] In an optional implementation, the training process of the prediction model includes: Perform data cleaning and feature extraction on pre-stored historical time series data to obtain data features; Inputting the data features into the input layer of the LSTM neural network model for training, and obtaining the predicted value data output by the output layer of the LSTM neural network model; Substituting the predicted value data and the pre-stored actual value into the loss function to calculate the loss value, thereby obtaining the loss value data; Calculate the gradient of the output layer output of the LSTM neural network model according to the loss value data, and pass the gradient forward layer by layer through the chain rule to calculate the gradient of each layer parameter to obtain gradient data; Update the parameters of each layer of the convolutional neural network model according to the gradient data and the preset learning rate; The parameters of each layer are updated repeatedly and iteratively until the training times of the convolutional neural network model are greater than the preset times, or when the loss value data of the convolutional neural network model is less than the preset loss threshold, the training is determined to be completed and a prediction model is obtained.

[0012] In an optional implementation, the performing of generator fault prevention operation and prevention effectiveness evaluation according to the fault type, the vibration prediction value, the temperature prediction value and the prediction test condition includes: If the fault type belongs to a pre-stored known fault, the specific fault type is output; If the fault type does not belong to the pre-stored known faults, the fault type is determined to be an unknown fault, and the fault classification model is updated according to the unknown fault; If the vibration prediction value or the temperature prediction value exceeds a preset warning threshold, a warning signal is triggered; The grey correlation analysis method is used to calculate the correlation between the predicted test conditions and the preset standard data to obtain a correlation matrix; Extract representative indicators and coverage indicators of the prediction test conditions according to the correlation matrix to generate preliminary evaluation results; Matching the preliminary evaluation results with a pre-stored evaluation index library to determine the evaluation level of the predicted test conditions; If the evaluation level is lower than the preset level threshold, the test data is corrected for the second time and the level evaluation is re-performed; If the evaluation level is higher than a preset level threshold, the prevention effectiveness evaluation is completed.

[0013] In a second aspect, the present invention provides a device for determining the effectiveness of a diesel generator set fault preventive test, comprising: The data acquisition module is used to obtain the vibration data, temperature data, vibration time series and temperature time series of the diesel generator set, and randomly sample the test conditions using the Monte Carlo simulation method to obtain the test data; A data processing module, used for preprocessing the vibration data and the temperature data using Dobesi wavelet to obtain vibration characteristics and temperature characteristics; A data dimension reduction module, used for performing data merging and dimension reduction processing according to the vibration characteristics and the temperature characteristics to obtain a fusion feature matrix; A fault analysis module, used for inputting the fused feature matrix into a pre-built fault classification model to obtain a fault type; A time alignment module, used for performing time series alignment on the vibration time series and the temperature time series using a dynamic time warping algorithm to obtain time series data; A data prediction module, used for inputting the time series data into a pre-trained prediction model to obtain vibration prediction values ​​and temperature prediction values; A test analysis module, used to use a Kalman filter algorithm to perform deviation correction on the test data, and input the corrected data into a pre-built test prediction model to obtain predicted test conditions; A prevention evaluation module is used to perform generator fault prevention operations and prevention effectiveness evaluation according to the fault type, the vibration prediction value, the temperature prediction value and the prediction test conditions.

[0014] In a third aspect, the present invention further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for determining the effectiveness of any one of the above-mentioned diesel generator set fault preventive tests is implemented.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium, the computer-readable storage medium comprising a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned methods for determining the effectiveness of the diesel generator set fault preventive test.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The invention discloses a method for determining the effectiveness of a diesel generator set fault preventive test, comprising obtaining vibration data, temperature data, vibration time series and temperature time series of the diesel generator set, and randomly sampling test conditions by using a Monte Carlo simulation method to obtain test data; preprocessing the vibration data and the temperature data by using a Dobesi wavelet to obtain vibration characteristics and temperature characteristics; merging and reducing dimensions according to the vibration characteristics and the temperature characteristics to obtain a fusion feature matrix; inputting the fusion feature matrix into a pre-built fault classification model to obtain a fault type; aligning the vibration time series and the temperature time series by using a dynamic time warping algorithm to obtain time series data; inputting the time series data into a pre-trained prediction model to obtain a vibration prediction value and a temperature prediction value; correcting the deviation of the test data by using a Kalman filtering algorithm, and inputting the corrected data into a pre-built test prediction model to obtain a prediction test condition; and performing generator fault prevention operations and prevention effectiveness evaluation according to the fault type, the vibration prediction value, the temperature prediction value and the prediction test condition. The present invention collects vibration and temperature data of diesel generator sets and their time series, generates test data using Monte Carlo simulation, and extracts features using Dobesi wavelet. After constructing the fusion feature matrix, the fault classification model is input to identify the fault type. At the same time, the dynamic time warping algorithm is used to align the time series data and calculate the predicted value. The Kalman filter is applied to correct the data deviation and optimize the test condition prediction. Finally, combined with the fault type, predicted value and test conditions, fault prevention measures are implemented and the effectiveness is evaluated, and the data is systematically analyzed to prevent faults, solving the problem of being unable to accurately perform fault diagnosis and preventive maintenance of diesel generators. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1It is a schematic flow chart of a method for determining the effectiveness of a diesel generator set fault preventive test provided by a first embodiment of the present invention; Figure 2 It is a schematic structural diagram of a device for determining the effectiveness of a diesel generator set fault preventive test provided by a second embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Reference Figure 1 The first embodiment of the present invention provides a method for determining the effectiveness of a diesel generator set fault preventive test, comprising the following steps: S11, obtaining vibration data, temperature data, vibration time series and temperature time series of the diesel generator set, and randomly sampling the test conditions using a Monte Carlo simulation method to obtain test data; S12, performing data preprocessing on the vibration data and the temperature data using Dobesi wavelet to obtain vibration characteristics and temperature characteristics; S13, performing data merging and dimensionality reduction processing according to the vibration characteristics and the temperature characteristics to obtain a fusion feature matrix; S14, inputting the fused feature matrix into a pre-built fault classification model to obtain a fault type; S15, using a dynamic time warping algorithm to perform time series alignment on the vibration time series and the temperature time series to obtain time series data; S16, inputting the time series data into a pre-trained prediction model to obtain a vibration prediction value and a temperature prediction value; S17, using a Kalman filter algorithm to correct the deviation of the test data, and inputting the corrected data into a pre-built test prediction model to obtain predicted test conditions; S18, performing generator fault prevention operations and prevention effectiveness evaluation according to the fault type, the vibration prediction value, the temperature prediction value and the prediction test conditions.

[0020] In step S11, vibration data, temperature data, vibration time series and temperature time series of the diesel generator set are obtained, and the test conditions are randomly sampled using a Monte Carlo simulation method to obtain test data.

[0021] Accelerometers and temperature sensors are installed on the key mechanical components of diesel generator sets. These sensors are used to monitor the vibration and temperature changes of each component in real time. For example, accelerometers are installed on the crankshaft, connecting rod, and cylinder to record the dynamic characteristics of the equipment during operation; temperature sensors are installed on the cylinder head, cooling system, and exhaust pipe to record the temperature changes of each part. Through these sensors, the vibration data and temperature data of the diesel generator set can be obtained, and time series data that changes over time can be formed. Specifically, vibration and temperature data are collected once a second to form a continuous time series for subsequent analysis and modeling.

[0022] In order to ensure that the test data can fully reflect the actual working conditions, the Monte Carlo simulation method is used to randomly sample the test conditions. First, the probability distribution model of the key parameters of the diesel generator set is defined. For example, the speed R follows a normal distribution. , the load L is uniformly distributed , the temperature T follows a normal distribution After determining the specific distribution parameters of these parameters based on historical data or expert knowledge, randomly select samples from the above probability distribution. Assuming the number of sampling is N, N groups of test condition data can be obtained: For each sampling, the speed Ri changes from The load Li is extracted from The temperature Ti is extracted from Based on these randomly sampled test condition data, a test data set close to the actual working conditions is generated. Assuming that the engine performance curve is known, the corresponding power P, torque M, fuel consumption F and other data can be calculated based on the sampled speed and load conditions.

[0023] The Monte Carlo simulation method is used to generate test data to improve the comprehensiveness and representativeness of the test. Traditional fixed condition tests can only cover limited working conditions, while Monte Carlo simulation can simulate various situations under actual working conditions, thereby more comprehensively evaluating the performance and reliability of diesel generator sets.

[0024] In step S12, the vibration data and the temperature data are preprocessed using Dobesi wavelet to obtain vibration characteristics and temperature characteristics.

[0025] In a specific implementation, the use of Dobesi wavelet to preprocess the vibration data and the temperature data to obtain vibration characteristics and temperature characteristics includes: Decomposing the vibration data and the temperature data into wavelet transform coefficients of different frequency components by using Dobesi wavelet, and removing high-frequency data in the wavelet transform coefficients based on a soft threshold method to obtain denoised wavelet coefficients; Reconstructing denoised vibration data and denoised temperature data through inverse wavelet transform according to the denoised wavelet coefficients; The denoised vibration data and the denoised temperature data are subjected to time domain and frequency domain feature extraction to obtain vibration features and temperature features.

[0026] Specifically, first, the vibration data and temperature data are decomposed into wavelet transform coefficients of different frequency components using Dobesi wavelet. Dobesi wavelet is a commonly used wavelet basis function with good time-frequency localization characteristics and is suitable for signal denoising and feature extraction. Assume that the vibration data is , the temperature data is ,These data can be decomposed into wavelet coefficients of different scales through Dobesi wavelet transform.

[0027] For vibration data and temperature data , respectively, continuous wavelet transform (CWT) or discrete wavelet transform (DWT), here we take discrete wavelet transform as an example: in, represents the wavelet transform coefficients, is the preset scale parameter, is the translation parameter, is the Dobesis wavelet basis function, express or .

[0028] In actual operation, assume that the length of the vibration data is , the length of the temperature data is also , it can be decomposed into wavelet coefficients of multiple scales by discrete wavelet transform. Layer wavelet transform, we can get the approximate coefficients and detail factor .

[0029] Based on the soft threshold method, high-frequency data in the wavelet transform coefficients are removed to obtain the denoised wavelet coefficients. The soft threshold method sets a threshold , process the wavelet coefficients: in, is a symbolic function that returns Symbols; return and The larger value of Represents the denoised wavelet coefficients.

[0030] Specifically, for each wavelet coefficient , if its absolute value is less than the threshold , then set it to zero; otherwise, adjust it to And preserve its sign. This can effectively remove noise while retaining the useful signal.

[0031] According to the denoised wavelet coefficients, the denoised vibration data and temperature data are reconstructed through inverse wavelet transform. The inverse wavelet transform formula is: For discrete wavelet transform, the specific reconstruction process is to recombine the approximate coefficients and detail coefficients of each layer to restore the original signal.

[0032] For example, in discrete wavelet transform, assuming that Layer decomposition, the approximate coefficient is and the detail coefficients of each layer , then the reconstruction formula is:

[0033] The denoised vibration data and temperature data Perform time domain and frequency domain feature extraction to obtain vibration features and temperature features.

[0034] Time domain features include mean, standard deviation, and peak factor. The specific calculation method is as follows: in, and represents the average value of the signal, and Indicates the degree of signal fluctuation. and Indicates the sharpness of the signal. Indicates the quantity of vibration data or temperature data.

[0035] Frequency domain features include power spectrum density and center frequency. The specific calculation method is as follows: in, represents the Fourier transform, and represents the distribution of signal energy in frequency, and Indicates where the signal energy is concentrated.

[0036] In step S13, data merging and dimension reduction processing are performed according to the vibration characteristics and the temperature characteristics to obtain a fusion feature matrix.

[0037] In a specific implementation manner, the data merging and dimensionality reduction processing is performed according to the vibration feature and the temperature feature to obtain a fusion feature matrix, including: Combining the vibration features and the temperature features into a feature matrix; The feature matrix is ​​subjected to dimensionality reduction processing by principal component analysis to obtain a low-dimensional feature matrix; The low-dimensional feature matrix is ​​cleaned according to a preset cleaning threshold to obtain a fused feature matrix.

[0038] Specifically, first, the vibration features and temperature features are combined into a feature matrix. Assume that the vibration feature is , the temperature characteristic is , these features are combined into a feature matrix , whose dimensions are ,in is the amount of vibration data or temperature data.

[0039] Next, the feature matrix is ​​reduced in dimension by principal component analysis (PCA) to reduce redundant information and extract the main features. The core idea of ​​principal component analysis is to project the original high-dimensional data into a new low-dimensional space through linear transformation so that the data in the new space has the largest variance, thereby retaining as much information as possible.

[0040] Specifically, the vibration features and temperature features are combined into a feature matrix : in, and are the matrix forms of vibration characteristics and temperature characteristics respectively, Represents a horizontal splicing operation.

[0041] The main steps of principal component analysis (PCA) include calculating the covariance matrix, solving eigenvalues ​​and eigenvectors, selecting principal components and performing data projection.

[0042] For the feature matrix , first standardize it so that its mean is 0 and its variance is 1. The standardized matrix is ​​recorded as . Then calculate its covariance matrix : in, represents the standardized feature matrix, The normalized transposed feature matrix, Indicates the number of vibration data or temperature data, represents the covariance matrix.

[0043] Covariance matrix Perform eigenvalue decomposition and obtain the eigenvalue and the corresponding eigenvector :

[0044] Among them, the eigenvalue Indicates the variance explained by each principal component, the eigenvector Indicates the new coordinate axis direction.

[0045] Sort by eigenvalue from large to small and select the first The eigenvalues ​​corresponding to the eigenvectors form a new projection matrix .here It is the preset number of dimensions after dimensionality reduction.

[0046] The original feature matrix Project to a new low-dimensional space to obtain a low-dimensional feature matrix : in, It is from the previous The matrix consists of feature vectors, with dimension , the matrix after projection The dimension is .

[0047] According to the preset threshold, the low-dimensional feature matrix is ​​cleaned to remove outliers or data points with large noise to obtain the final fusion feature matrix. Assume that the threshold is set to , then for each element in the feature matrix , if its absolute value is greater than the threshold , then set it to zero or remove the data point, otherwise keep the element.

[0048] After the above steps, the denoised fusion feature matrix is ​​obtained This matrix not only contains the comprehensive information of vibration and temperature, but also reduces redundant features through dimensionality reduction processing, improving the efficiency and accuracy of subsequent model training.

[0049] This method can effectively integrate multi-source data, reduce data dimensions while retaining key information, thereby improving the performance of the fault diagnosis model. Through a systematic data processing process, the condition monitoring and fault warning capabilities of diesel generator sets can be significantly improved to ensure the reliability and stability of the equipment. At the same time, the data cleaning step further ensures the quality of the data, avoids the impact of outliers on the model, and improves the robustness of the overall system.

[0050] In step S14, the fused feature matrix is ​​input into a pre-built fault classification model to obtain the fault type.

[0051] In a specific implementation, the process of constructing the fault classification model includes: Divide the pre-stored historical fusion feature matrix and fault feature matrix into a test set, a training set and a validation set; Select support vector regression as the initial model, radial basis function as the kernel function, and minimize the loss function as the original optimization problem; The original optimization problem is converted into a Lagrange multiplier solution according to the radial basis function, and the dual form of the original problem is solved by quadratic programming to obtain a minimized objective function, a regularization parameter, and a kernel function parameter; Using a cross-validation method to evaluate the performance indicators of the initial model on the validation set, wherein the performance indicators include mean square error and mean absolute error; When the performance indicator does not meet the preset performance standard, adjusting the regularization parameter and the kernel function parameter, and resolving the original optimization problem; When the performance indicator meets the preset performance standard, a fault classification model of the corresponding parameters is obtained.

[0052] Specifically, first, in order to build a fault classification model, the pre-stored historical fusion feature matrix and fault feature matrix need to be divided into a test set, a training set, and a validation set. Assume that the historical data set is , randomly divide it into training sets , validation set and test set .

[0053] Next, we select the support vector machine (SVM) as the initial model and use the radial basis function (RBF) as the kernel function. The goal of SVM is to minimize the loss function, which is the original optimization problem. Specifically, the objective function of SVM can be expressed as: in, is the weight vector, is the bias term, is a slack variable used to handle the inseparable case, is the regularization parameter that controls the trade-off between model complexity and error.

[0054] The radial basis function (RBF) is used as the kernel function, which has the form: in, is the kernel function parameter, which controls the influence range of the kernel function. and is the eigenvector in the historical fusion feature matrix and the fault feature matrix, represents radial basis function.

[0055] The original optimization problem is transformed into the form of solving Lagrange multipliers. By introducing Lagrange multipliers , we get the dual form of the optimization problem: , in, and is the sample label, and is the Lagrange multiplier, is the number of elements in the history fusion feature matrix and the fault feature matrix, represents the regularization parameter, represents the radial basis function, and It is the eigenvector in the history fusion feature matrix and the fault feature matrix.

[0056] The dual optimization problem is solved by quadratic programming method to obtain the minimized objective function and regularization parameter and kernel function parameters .

[0057] The cross-validation method is used to evaluate the performance indicators of the initial model on the validation set. Cross-validation is performed by further dividing the training set into multiple subsets, using one of the subsets as the validation set and the remaining subsets as the training set in turn, and repeatedly evaluating the model performance. Common performance indicators include mean square error (MSE) and mean absolute error (MAE): in, is the true label, is the predicted label and N is the number of subset samples.

[0058] When the performance index does not meet the preset performance standard, adjust the regularization parameter and kernel function parameters , re-solve the original optimization problem. This process can be done by tuning parameters through grid search or random search until the optimal parameter combination is found.

[0059] When the performance index meets the preset performance standard, the fault classification model of the corresponding parameters is obtained. At this time, the new fusion feature matrix is ​​classified using the model to output the fault type.

[0060] Specifically, the fusion feature matrix Input into the trained fault classification model, the model predicts the fault type based on the input features. Assume that the fault types are Class, the model outputs the probability distribution of each class ,in . Finally, the category with the highest probability is selected as the prediction result.

[0061] This approach can effectively utilize multi-source data for fault diagnosis, and improve the accuracy and reliability of fault identification through systematic data processing and model training processes. Through cross-validation and parameter tuning, the generalization ability of the model on different data sets is ensured, reducing the risk of overfitting.

[0062] In step S15, a dynamic time warping algorithm is used to align the vibration time series and the temperature time series to obtain time series data.

[0063] In a specific implementation, the use of a dynamic time warping algorithm to perform time series alignment on the vibration time series and the temperature time series to obtain time series data includes: constructing a distance matrix based on the vibration time series and the temperature time series; Constructing an initial cumulative distance matrix, and optimizing the initial cumulative distance matrix according to the distance matrix to obtain a cumulative distance matrix; Backtracking according to the cumulative distance matrix to find the best path with the minimum cumulative distance; Calculating a time offset according to the optimal path, adjusting a vibration time series and a temperature time series according to the offset, and combining the vibration time series and the temperature time series with the corresponding vibration data and the temperature data to obtain time series data; Among them, the elements in the distance matrix are calculated by the following formula: Among them, the construction of the initial cumulative distance matrix is ​​as follows: Among them, the elements in the cumulative distance matrix are calculated by the following formula: in, Indicates the distance matrix The elements at the position, Indicates the vibration time series data, Indicates the temperature time series data, represents the initial cumulative distance matrix, Indicates the cumulative distance matrix The elements at the position, Indicates the cumulative distance matrix The elements at the position, Indicates the cumulative distance matrix The element at position.

[0064] Specifically, first, according to the vibration time series and temperature time series Constructing distance matrix . Distance Matrix Each element in represents the first The data point in the temperature time series The Euclidean distance between data points: Next, construct the initial cumulative distance matrix , whose dimensions are , and initialized to: in, , the remaining boundary conditions are set to infinity to ensure that the path can only extend from the upper left corner to the lower right corner.

[0065] Then, according to the distance matrix Optimize the initial cumulative distance matrix , and get the cumulative distance matrix For each element in the cumulative distance matrix , calculated using the following formula: In this way, the cumulative distance matrix records the minimum cumulative distance from the starting point to each location.

[0066] Next, according to the cumulative distance matrix Backtrack to find the best path with the smallest cumulative distance. Start back to , select the direction with the smallest cumulative distance in each step. Specifically, compare , and , select the path direction corresponding to the smallest one, and record all the points on the path to form the optimal path.

[0067] The time offset is calculated based on the optimal path, and the vibration time series and temperature time series are adjusted based on the offset. For example, if the optimal path shows that the vibration time series leads the temperature time series by 2 time units, the temperature time series is moved forward by 2 units. The specific adjustment method is as follows: Assuming the optimal path is , the time offset of each position can be calculated The position of the temperature time series is then adjusted according to these offsets.

[0068] Finally, the adjusted vibration time series and temperature time series are combined with the corresponding vibration data and temperature data to obtain the final time series data. This process not only ensures the alignment between different time series, but also retains the characteristic information of the original data, which is helpful for subsequent analysis and fault diagnosis.

[0069] This method can effectively align data from different time series and eliminate the timing misalignment problem caused by sampling frequency differences or changes in equipment operating status. Through a systematic data processing process, the consistency and reliability of time series data are improved, providing a high-quality data foundation for subsequent status monitoring and fault warning. At the same time, the application of the DTW algorithm makes data alignment more flexible and accurate, improving the robustness and performance of the overall system.

[0070] In step S16, the time series data is input into a pre-trained prediction model to obtain vibration prediction values ​​and temperature prediction values.

[0071] In a specific implementation, the training process of the prediction model includes: Perform data cleaning and feature extraction on pre-stored historical time series data to obtain data features; Inputting the data features into the input layer of the LSTM neural network model for training, and obtaining the predicted value data output by the output layer of the LSTM neural network model; Substituting the predicted value data and the pre-stored actual value into the loss function to calculate the loss value, thereby obtaining the loss value data; Calculate the gradient of the output layer output of the LSTM neural network model according to the loss value data, and pass the gradient forward layer by layer through the chain rule to calculate the gradient of each layer parameter to obtain gradient data; Update the parameters of each layer of the convolutional neural network model according to the gradient data and the preset learning rate; The parameters of each layer are updated repeatedly and iteratively until the training times of the convolutional neural network model are greater than the preset times, or when the loss value data of the convolutional neural network model is less than the preset loss threshold, the training is determined to be completed and a prediction model is obtained.

[0072] Specifically, first, the pre-stored historical time series data is cleaned and features are extracted to obtain data features. Assuming that the historical time series data includes vibration time series and temperature time series, outliers and noise are removed through data cleaning, and time domain and frequency domain features are extracted to form a feature matrix.

[0073] Next, the feature matrix is ​​input into the input layer of the LSTM neural network model for training. LSTM (Long Short-Term Memory Network) is a special type of recurrent neural network (RNN) that is suitable for processing time series data with long dependencies. The structure of the LSTM model includes an input layer, multiple LSTM layers, and an output layer. Assume that the input layer receives data from the past 24 hours as features, and the output layer predicts the vibration and temperature values ​​for the next hour.

[0074] Specifically, the LSTM model contains multiple parameters, including the weight matrix and the bias term . Assume that the LSTM model has Layer, each layer has neurons, and the input dimension is , then the parameters of each layer can be expressed as , , , Indicates layer, Represents the different gates in an LSTM cell.

[0075] For each time step, the LSTM unit is calculated as follows: in, are input features, is the hidden state at the previous time step, is the cell state at the previous time step, is the sigmoid function, is the hyperbolic tangent function, represents element-wise multiplication, is the input gate, It's the Gate of Forgetfulness. is the candidate state, is the cell state, is the output gate, is the hidden state, , , and Indicates the preset paranoia item.

[0076] The predicted value output by the LSTM model and actual value Substitute into the loss function and calculate the loss value. The loss function is the mean square error (MSE): in, represents all the parameters of the model, is the sample size.

[0077] The gradient of the output layer of the LSTM model is calculated based on the loss value, and the gradient is passed forward layer by layer through the chain rule to calculate the gradient of each layer parameter. Specifically, for each parameter , its gradient can be calculated by the following formula:

[0078] in, is the derivative of the loss function with respect to the predicted value is the derivative of the predicted value with respect to the hidden state, is the derivative of the hidden state with respect to the parameters.

[0079] According to the gradient data and the preset learning rate , update the parameters of each layer of the LSTM model. The specific formula is as follows: here, is the learning rate, which controls the step size of parameter updates, represents all the parameters of the model, Represents all parameters of the updated model.

[0080] Iterate and update each layer parameter repeatedly until the training completion conditions are met. Common training completion conditions include: When the number of LSTM model training times exceeds the preset number; When the loss value data of the LSTM model is less than the preset loss threshold.

[0081] When any of the above conditions is met, the training is determined to be complete and the final prediction model is obtained. At this time, the time series data is input into the prediction model, and the model predicts the future vibration value and temperature value based on the input features.

[0082] The significance of this method is that it can effectively use time series data for prediction, and improve the accuracy and reliability of prediction through systematic data processing and model training processes. The LSTM model can capture the long-term dependencies in time series data, making the prediction results more accurate. At the same time, the model parameters are optimized through the back propagation algorithm and gradient descent method, ensuring the generalization ability of the model on different data sets and reducing the risk of overfitting.

[0083] In step S17, the test data is corrected for deviation using a Kalman filter algorithm, and the corrected data is input into a pre-built test prediction model to obtain predicted test conditions.

[0084] Specifically, Kalman filtering is a recursive algorithm that is suitable for the state estimation problem of linear dynamic systems. It estimates the true state of the system by fusing the predicted value and the measured value. Assume that the system state is , the observed value is , the process of Kalman filtering is as follows: According to the state estimate at the previous moment And the system state transfer matrix , predict the current state: in, is the state transition matrix, is the control input matrix, is the preset control vector, Indicates the prediction status at the current moment.

[0085] According to the covariance matrix of the previous moment and the system noise covariance matrix , predict the covariance at the current moment: According to the prediction covariance matrix and the observation noise covariance matrix , calculate the Kalman gain : in, is the preset observation matrix, is the observation noise covariance matrix.

[0086] According to the observed value and Kalman gain , update the current state estimate: According to the Kalman gain , update the covariance matrix at the current moment: in, is the default identity matrix.

[0087] After being processed by the above Kalman filter algorithm, the corrected vibration data and temperature data are obtained. Next, these corrected data are input into the pre-built test prediction model to obtain the predicted test conditions.

[0088] The hypothetical test prediction model is a model built based on machine learning or statistical methods, such as support vector machine (SVM), random forest or long short-term memory network (LSTM). For each time step, the corrected vibration data and temperature data are input, and the model outputs the predicted test conditions.

[0089] This method can effectively correct the deviation in the test data and improve the accuracy and reliability of the data. By fusing the predicted value and the measured value, the Kalman filter can dynamically adjust the state estimation to make the result more accurate. In addition, inputting the corrected data into the pre-built test prediction model can further improve the accuracy and reliability of the prediction and ensure the validity of the test results. Through the systematic data processing and model training process, the fault diagnosis and preventive maintenance capabilities of diesel generator sets are improved, ensuring the reliability and stability of the equipment.

[0090] In step S18, generator fault prevention operations and prevention effectiveness evaluation are performed according to the fault type, the vibration prediction value, the temperature prediction value and the prediction test conditions.

[0091] In a specific implementation, the performing of generator fault prevention operation and prevention effectiveness evaluation according to the fault type, the vibration prediction value, the temperature prediction value and the prediction test condition includes: If the fault type belongs to a pre-stored known fault, the specific fault type is output; If the fault type does not belong to the pre-stored known faults, the fault type is determined to be an unknown fault, and the fault classification model is updated according to the unknown fault; If the vibration prediction value or the temperature prediction value exceeds a preset warning threshold, a warning signal is triggered; The grey correlation analysis method is used to calculate the correlation between the predicted test conditions and the preset standard data to obtain a correlation matrix; Extract representative indicators and coverage indicators of the prediction test conditions according to the correlation matrix to generate preliminary evaluation results; Matching the preliminary evaluation results with a pre-stored evaluation index library to determine the evaluation level of the predicted test conditions; If the evaluation level is lower than the preset level threshold, the test data is corrected for a second time and the level evaluation is re-performed; If the evaluation level is higher than a preset level threshold, the prevention effectiveness evaluation is completed.

[0092] Specifically, first, a judgment is made based on the fault type output by the fault classification model. If the fault type belongs to a known fault, the specific fault type is directly output. Otherwise, the fault type is determined to be an unknown fault, and the fault classification model is updated.

[0093] Next, based on the vibration prediction value and temperature prediction Compare with the preset threshold to decide whether to trigger the warning signal. Assume that the preset vibration threshold is , the temperature threshold is , the trigger condition is: if any predicted value exceeds the preset threshold, an early warning signal is triggered and relevant data is recorded for subsequent analysis.

[0094] In order to evaluate the effectiveness of the predicted test conditions, the grey correlation analysis method is used to calculate the correlation between the predicted test conditions and the preset standard data to obtain the correlation matrix. Assume that the preset standard data is , the predicted test conditions are , then the correlation The calculation formula is: in, is the preset resolution factor.

[0095] All the correlation Composition correlation matrix : According to the correlation matrix Extract representative indicators and coverage indicators of the predicted test conditions to generate preliminary evaluation results. The representative indicator can be the maximum value or average value in the correlation matrix, and the coverage indicator can be the minimum value or variance in the correlation matrix. For example, the representative indicator and coverage indicators It can be expressed as: The preliminary evaluation results are matched with the pre-stored evaluation index library to determine the evaluation level of the predicted test conditions.

[0096] If the evaluation level is lower than the preset level threshold, the test data is revised for the second time and the level evaluation is re-performed. Assuming that the test data after the second revision is obtained, the correlation matrix is ​​recalculated and new representative indicators and coverage indicators are extracted, and then matching and evaluation are performed again.

[0097] If the evaluation level is higher than the preset level threshold, the prevention effectiveness evaluation is completed, confirming that the current test conditions and prediction results are valid and can be used for further status monitoring and fault warning.

[0098] The significance of this method is that it can effectively combine fault type, vibration prediction value, temperature prediction value and prediction test conditions to conduct comprehensive fault prevention operations and prevention effectiveness evaluation. Through systematic data processing and model training process, the accuracy and reliability of fault identification are improved. The application of grey correlation analysis method makes the evaluation process more flexible and accurate, and improves the robustness and performance of the overall system.

[0099] The following describes the working process of the present invention using a relatively common scenario as an example. Figure 1 , a method for determining the effectiveness of a diesel generator set fault preventive test, comprising the following steps: The diesel generator set experienced abnormal vibration and temperature fluctuation during operation. In order to ensure the safe operation of the equipment and prevent potential failures in advance, it was decided to use Dobesi wavelet to pre-process the vibration data and temperature data, build a multi-source data fusion model, and classify the faults through the support vector machine (SVM) model.

[0100] First, the vibration data and temperature data of the diesel generator set are obtained through sensors. The data are decomposed into wavelet transform coefficients of different frequency components using Dobesi wavelet, and high-frequency noise is removed based on the soft threshold method to obtain denoised vibration data and temperature data.

[0101] Next, the denoised vibration data and temperature data are extracted in the time domain and frequency domain to obtain vibration features and temperature features. These features are then merged into a feature matrix. The principal component analysis (PCA) method is used to reduce the dimension of the feature matrix to obtain a low-dimensional feature matrix. The low-dimensional feature matrix is ​​cleaned according to the preset threshold to finally obtain a fused feature matrix.

[0102] The fused feature matrix is ​​input into the pre-built support vector machine (SVM) fault classification model. The model is trained based on historical data sets, including training sets, validation sets, and test sets. After multiple iterations and parameter tuning, the optimal fault classification model is finally obtained. The current fused feature matrix is ​​input, and the model outputs the specific fault type. Assume that the output fault type is "bearing wear".

[0103] In order to further improve the consistency and reliability of the data, the dynamic time warping (DTW) algorithm is used to align the vibration time series and the temperature time series. The distance matrix is ​​constructed and the initial cumulative distance matrix is ​​optimized to obtain the cumulative distance matrix. The optimal path with the minimum cumulative distance is backtracked, and the time offset is calculated based on the optimal path. The vibration time series and the temperature time series are adjusted to finally obtain the aligned time series data.

[0104] The aligned time series data is input into the pre-trained long short-term memory network (LSTM) model to obtain the vibration prediction value and temperature prediction value. If any of the prediction values ​​exceeds the preset threshold, an early warning signal is triggered. For example, assuming that the preset vibration threshold is 0.5 and the temperature threshold is 85°C, and the actual prediction values ​​are 0.6 and 90°C respectively, an early warning signal is triggered.

[0105] The grey correlation analysis method is used to calculate the correlation between the predicted test conditions and the preset standard data to obtain the correlation matrix. The representative indicators and coverage indicators are extracted according to the correlation matrix to generate preliminary evaluation results. The preliminary evaluation results are matched with the pre-stored evaluation indicator library to determine the evaluation level of the predicted test conditions. Assuming that the evaluation level is "good", which is higher than the preset level threshold, the preventive effectiveness evaluation is completed, confirming that the current test conditions and prediction results are valid, and solving the problem of being unable to accurately perform fault diagnosis and preventive maintenance of diesel generators.

[0106] Reference Figure 2 The second embodiment of the present invention provides a device for determining the effectiveness of a diesel generator set fault preventive test, comprising: The data acquisition module is used to obtain the vibration data, temperature data, vibration time series and temperature time series of the diesel generator set, and randomly sample the test conditions using the Monte Carlo simulation method to obtain the test data; A data processing module, used for preprocessing the vibration data and the temperature data using Dobesi wavelet to obtain vibration characteristics and temperature characteristics; A data dimension reduction module, used for performing data merging and dimension reduction processing according to the vibration characteristics and the temperature characteristics to obtain a fusion feature matrix; A fault analysis module, used for inputting the fused feature matrix into a pre-built fault classification model to obtain a fault type; A time alignment module, used for performing time series alignment on the vibration time series and the temperature time series using a dynamic time warping algorithm to obtain time series data; A data prediction module, used for inputting the time series data into a pre-trained prediction model to obtain vibration prediction values ​​and temperature prediction values; A test analysis module, used to use a Kalman filter algorithm to perform deviation correction on the test data, and input the corrected data into a pre-built test prediction model to obtain predicted test conditions; A prevention evaluation module is used to perform generator fault prevention operations and prevention effectiveness evaluation according to the fault type, the vibration prediction value, the temperature prediction value and the prediction test conditions.

[0107] It should be noted that the device for determining the effectiveness of a diesel generator set fault preventive test provided in an embodiment of the present invention is used to execute all the process steps of the method for determining the effectiveness of a diesel generator set fault preventive test in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be repeated here.

[0108] The embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a program for determining the effectiveness of a diesel generator set fault preventive test. When the processor executes the computer program, the steps of the above-mentioned methods for determining the effectiveness of the diesel generator set fault preventive test are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the effectiveness determination module of the diesel generator set fault preventive test.

[0109] Exemplarily, the computer program may be divided into one or more modules / units, one or more modules / units are stored in a memory and executed by a processor to implement the present invention. One or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in an electronic device.

[0110] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0111] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of an electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0112] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0113] Among them, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electric carrier signals and telecommunication signals.

[0114] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0115] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for determining the effectiveness of a diesel generator set fault preventive test, characterized in that: Executed by the controller, including: Obtain vibration data, temperature data, vibration time series and temperature time series of the diesel generator set, and use the Monte Carlo simulation method to randomly sample the test conditions to obtain test data; Dobesi wavelet is used to preprocess the vibration data and temperature data to obtain vibration characteristics and temperature characteristics; Data merging and dimension reduction processing are performed according to vibration characteristics and temperature characteristics to obtain a fusion feature matrix; Input the fused feature matrix into the pre-built fault classification model to obtain the fault type; The dynamic time warping algorithm is used to align the vibration time series and the temperature time series to obtain the time series data; Input the time series data into the pre-trained prediction model to obtain the vibration prediction value and temperature prediction value; Use the Kalman filter algorithm to correct the deviation of the test data, and input the corrected data into the pre-built test prediction model to obtain the predicted test conditions; Generator fault prevention operations and prevention effectiveness evaluation are performed based on fault types, vibration prediction values, temperature prediction values ​​and prediction test conditions.

2. The method for determining the effectiveness of a diesel generator set fault preventive test according to claim 1, characterized in that: Dobesi wavelet is used to preprocess the vibration data and temperature data to obtain vibration characteristics and temperature characteristics, including: The vibration data and temperature data are decomposed into wavelet transform coefficients of different frequency components by using Dobesi wavelet, and the high-frequency data in the wavelet transform coefficients are removed based on the soft threshold method to obtain the denoised wavelet coefficients. According to the denoised wavelet coefficients, the denoised vibration data and the denoised temperature data are reconstructed by inverse wavelet transform; The denoised vibration data and denoised temperature data are subjected to time domain and frequency domain feature extraction to obtain vibration features and temperature features.

3. The method for determining the effectiveness of a diesel generator set fault preventive test according to claim 1, characterized in that: Data merging and dimensionality reduction are performed based on vibration characteristics and temperature characteristics to obtain a fusion feature matrix, including: Combine vibration features and temperature features into a feature matrix; The feature matrix is ​​reduced in dimension by principal component analysis to obtain a low-dimensional feature matrix; The low-dimensional feature matrix is ​​cleaned according to the preset cleaning threshold to obtain a fused feature matrix.

4. The method for determining the effectiveness of a diesel generator set fault preventive test according to claim 1, characterized in that: The construction process of the fault classification model includes: Divide the pre-stored historical fusion feature matrix and fault feature matrix into a test set, a training set and a validation set; Select support vector regression as the initial model, radial basis function as the kernel function, and minimize the loss function as the original optimization problem; The original optimization problem is converted into solving Lagrange multipliers according to radial basis functions, and the dual form of the original problem is solved by quadratic programming to obtain the minimized objective function, regularization parameters and kernel function parameters; The cross-validation method is used to evaluate the performance indicators of the initial model on the validation set, where the performance indicators include mean square error and mean absolute error; When the performance index does not meet the preset performance standard, the regularization parameters and kernel function parameters are adjusted to solve the original optimization problem again; When the performance index meets the preset performance standard, a fault classification model of the corresponding parameters is obtained.

5. The method for determining the effectiveness of a diesel generator set fault preventive test according to claim 1, characterized in that: The dynamic time warping algorithm is used to align the vibration time series and the temperature time series to obtain time series data, including: Construct a distance matrix based on the vibration time series and the temperature time series; Construct an initial cumulative distance matrix, and optimize the initial cumulative distance matrix according to the distance matrix to obtain a cumulative distance matrix; Backtrack according to the cumulative distance matrix to find the best path with the minimum cumulative distance; Calculating a time offset according to the optimal path, adjusting the vibration time series and the temperature time series according to the offset, and combining the vibration time series and the temperature time series with the corresponding vibration data and temperature data to obtain time series data; Among them, the elements in the distance matrix are calculated by the following formula: Among them, the construction of the initial cumulative distance matrix is ​​as follows: Among them, the elements in the cumulative distance matrix are calculated by the following formula: in, Indicates the distance matrix The elements at the position, Indicates the vibration time series data, Indicates the temperature time series data, represents the initial cumulative distance matrix, Indicates the cumulative distance matrix The elements at the position, Indicates the cumulative distance matrix The elements at the position, Indicates the cumulative distance matrix The element at position.

6. The method for determining the effectiveness of a diesel generator set fault preventive test according to claim 1, characterized in that: The training process of the prediction model includes: Perform data cleaning and feature extraction on pre-stored historical time series data to obtain data features; Input the data features into the input layer of the LSTM neural network model for training, and obtain the predicted value data output by the output layer of the LSTM neural network model; Substitute the predicted value data and the pre-stored actual value into the loss function to calculate the loss value and obtain the loss value data; The gradient of the output layer of the LSTM neural network model is calculated based on the loss value data, and the gradient is passed forward layer by layer through the chain rule to calculate the gradient of each layer parameter and obtain the gradient data; Update the parameters of each layer of the convolutional neural network model according to the gradient data and the preset learning rate; The parameters of each layer are updated repeatedly until the training times of the convolutional neural network model are greater than the preset times, or when the loss value data of the convolutional neural network model is less than the preset loss threshold, the training is determined to be completed and the prediction model is obtained.

7. The method for determining the effectiveness of a diesel generator set fault preventive test according to claim 1, characterized in that: Generator fault prevention operations and prevention effectiveness evaluation are performed based on fault types, vibration prediction values, temperature prediction values, and prediction test conditions, including: If the fault type belongs to a pre-stored known fault, the specific fault type is output; If the fault type does not belong to the pre-stored known faults, the fault type is determined to be an unknown fault, and the fault classification model is updated according to the unknown fault; If the vibration prediction value or temperature prediction value exceeds the preset warning threshold, a warning signal is triggered; The grey correlation analysis method is used to calculate the correlation between the predicted test conditions and the preset standard data to obtain the correlation matrix; Extract representative indicators and coverage indicators of the predicted test conditions according to the correlation matrix and generate preliminary evaluation results; Matching the preliminary evaluation results with the pre-stored evaluation index library to determine the evaluation level of the predicted test conditions; If the evaluation level is lower than the preset level threshold, the test data will be revised for the second time and the level evaluation will be re-performed; If the evaluation level is higher than the preset level threshold, the prevention effectiveness evaluation is completed.

8. A device for determining the effectiveness of a diesel generator set fault preventive test, characterized in that: include: The data acquisition module is used to obtain the vibration data, temperature data, vibration time series and temperature time series of the diesel generator set, and randomly sample the test conditions using the Monte Carlo simulation method to obtain the test data; A data processing module is used to pre-process the vibration data and the temperature data using Dobesi wavelet to obtain vibration characteristics and temperature characteristics; The data dimension reduction module is used to merge and reduce the dimension of data according to the vibration characteristics and temperature characteristics to obtain a fusion feature matrix; The fault analysis module is used to input the fused feature matrix into the pre-built fault classification model to obtain the fault type; A time alignment module is used to align the vibration time series and the temperature time series using a dynamic time warping algorithm to obtain time series data; A data prediction module is used to input time series data into a pre-trained prediction model to obtain vibration prediction values ​​and temperature prediction values; The test analysis module is used to correct the deviation of the test data using the Kalman filter algorithm, and input the corrected data into the pre-built test prediction model to obtain the predicted test conditions; The prevention evaluation module is used to perform generator fault prevention operations and prevention effectiveness evaluation based on fault types, vibration prediction values, temperature prediction values ​​and prediction test conditions.

9. An electronic device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for determining the effectiveness of a diesel generator set fault preventive test as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for determining the effectiveness of a diesel generator set fault preventive test as claimed in any one of claims 1 to 7.

Images