Diesel generator set bearing fault diagnosis system combining ttao-vmd and iwoa-bp

Abstract

A diesel generator set bearing fault diagnosis system combining TTAO-VMD and IWOA-BP includes: a signal acquisition module for acquiring vibration signals of the monitored diesel generator set bearings, which serve as input signals to the TTAO-VMD fault feature extraction module; a TTAO-VMD fault feature extraction module for receiving bearing vibration signals and calculating characteristic parameters of the bearing vibration signals in the time and frequency domains based on optimal components, constructing a feature vector dataset; an IWOA-BP fault mode recognition module for classifying rolling bearing faults based on the feature vector dataset; and a human-machine interface module for displaying data analysis results and fault types.

Description

Technical fields:

[0001] This invention relates to a diesel generator set bearing fault diagnosis system combining TTAO-VMD and IWOA-BP, and more particularly to a diesel generator set bearing fault diagnosis system based on variational mode decomposition and whale optimization algorithm. Background technology:

[0002] Rolling bearings are the most common critical components in mechanical equipment. Once they fail, they may cause serious industrial accidents and incalculable economic losses. Therefore, fault diagnosis of rolling bearings is of great significance. Generally speaking, fault diagnosis of rolling bearings includes two key steps: fault feature extraction and fault mode recognition.

[0003] In the extraction of fault features from rolling bearings, vibration signals are first collected, and then information containing fault features is extracted and decomposed from them. To address the signal distortion and mode aliasing problems inherent in traditional signal decomposition methods, variational mode decomposition (VMD) has been proposed, effectively solving the mode aliasing problem in traditional methods. However, during feature signal extraction, the decomposition effect of VMD is highly dependent on the decomposition order k and the penalty factor α. Since these parameters are usually set empirically, this method is prone to over-decomposition or under-decomposition, thus affecting the decomposition accuracy and fault diagnosis effect.

[0004] In the field of fault mode identification, numerous researchers have conducted studies to address the slow convergence speed and susceptibility to local optima inherent in traditional whale optimization algorithms. For example, the introduction of inertial weights from particle swarm optimization has solved this problem; combining von Neumann topology with adaptive weights has improved the convergence speed and accuracy; introducing inertial weights and adaptive convergence factors has reduced the risk of local optima and premature convergence; and using nonlinear inertial weights has improved the global search capability and convergence speed. However, all of these methods improve whale optimization performance by introducing adaptive weights. If the pace of adaptive weight adjustments is too large or too frequent, the algorithm's direction will change drastically during iteration, causing the solution to fluctuate significantly in the target space or even deviate from the optimal region, failing to converge effectively to the optimal solution. Furthermore, whale optimization also suffers from slow convergence speed when dealing with high function complexity, susceptibility to local optima in high-dimensional optimization, and a decrease in population diversity with increasing iteration count, leading to insufficient exploration capability and low global search capability in later stages.

[0005] In summary, existing methods for diagnosing bearing faults in diesel generator sets have the following shortcomings in practical applications:

[0006] 1. The decomposition order k and penalty factor α in the variational mode decomposition method have a significant impact on the decomposition effect. Since these parameters are usually set empirically, the method is prone to over-decomposition or under-decomposition, which affects the decomposition accuracy and fault diagnosis effect.

[0007] 2. When using intelligent optimization algorithms to optimize the two parameters k and penalty factor α in variational mode decomposition methods, there are drawbacks such as slow convergence speed, low global search capability, and easy getting trapped in local optima.

[0008] 3. Fault mode identification methods based on whale optimization algorithms suffer from poor stability of algorithm iteration, low optimization efficiency, and low global search capability. Summary of the Invention:

[0009] This invention provides a diesel generator set bearing fault diagnosis system combining TTAO-VMD and IWOA-BP. The system features a rationally designed structure and method. A triangular topology aggregation optimization algorithm is introduced to optimize the decomposition order k and penalty factor α of the variational mode decomposition method. A TTAO-VMD fault feature extraction module is constructed to avoid reducing the accuracy of the variational mode decomposition method due to improper parameter settings. Correspondingly, an IWOA-BP fault mode recognition module is also constructed, which improves the whale optimization algorithm based on the inertia weight strategy of the sigmoid function, enhancing the stability of algorithm iteration. Furthermore, the IWOA-BP fault mode recognition module introduces chaotic mapping initialization population and elite retention and perturbation mechanisms to further improve the optimization efficiency and global search capability of the whale optimization algorithm, solving the problems existing in the prior art.

[0010] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0011] A diesel generator set bearing fault diagnosis system combining TTAO-VMD and IWOA-BP, the diagnosis system including:

[0012] The signal acquisition module is used to acquire the vibration signal of the bearing of the monitored diesel generator set and use it as the input signal of the TTAO-VMD fault feature extraction module.

[0013] The TTAO-VMD fault feature extraction module is used to receive bearing vibration signals and calculate the feature parameters of the bearing vibration signals in the time and frequency domains based on the optimal components, thereby constructing a feature vector dataset.

[0014] The IWOA-BP fault mode recognition module is used to classify rolling bearing faults based on a feature vector dataset.

[0015] Human-machine interface module, which is used to display data analysis results and fault types;

[0016] In the TTAO-VMD fault feature extraction module, a triangular aggregation topology optimization algorithm is introduced to optimize the two parameters of the variational mode decomposition method, namely the decomposition order k and the penalty factor α. In the IWOA-BP fault mode recognition module, a chaotic mapping strategy is introduced to initialize the population position in order to improve the global search capability of the whale optimization algorithm.

[0017] The chaotic mapping strategy is as follows:

[0018]

[0019] Where, x i This indicates the position where the i-th individual was generated.

[0020] The IWOA-BP fault mode recognition module introduces an adaptive inertial weight based on the Sigmoid function to improve the whale optimization algorithm's ability to balance exploration and development, and accelerate its convergence speed. The adaptive inertial weight is as follows:

[0021]

[0022] Where, ω max and ω min These represent the maximum and minimum values ​​of the weights, respectively. The coefficient z is used to adjust the steepness of the sigmoid function; m and n are two parameter coefficients; T is the maximum number of iterations.

[0023] An elite retention and perturbation mechanism is introduced in the IWOA-BP fault mode identification module to prevent the whale optimization algorithm from getting trapped in local optima. The perturbation mechanism is as follows:

[0024]

[0025]

[0026] in, γ represents the position of the elite solution; γ is a constant that defines the perturbation amplitude.

[0027] The data displayed on the human-machine interface module includes variational mode decomposition results in the time and frequency domains, variational mode decomposition results in the frequency domain, signal reconstruction results in the time domain, signal reconstruction results in the frequency domain, and fault diagnosis results.

[0028] This invention employs the aforementioned structure and optimizes the decomposition order k and penalty factor α of the variational mode decomposition method using a triangular topology aggregation optimization algorithm, avoiding a decrease in the accuracy of the variational mode decomposition method due to improper parameter settings. It improves the stability of the algorithm iteration by using an improved whale optimization algorithm based on a sigmoid function-based inertia weight strategy. It introduces chaotic mapping to initialize the population, as well as an elite retention and perturbation mechanism, to improve the optimization efficiency and global search capability of the whale optimization algorithm. By adding a local perturbation mechanism on the basis of retaining elite individuals, it slightly perturbs the location of the elite individuals to explore potential optimal candidate solutions in their vicinity, thereby more accurately approximating the global optimal solution. This invention has the advantages of precision, efficiency, practicality, and reliability. Attached image description:

[0029] Figure 1 This is a schematic diagram of the structure of the present invention.

[0030] Figure 2 This is a schematic diagram of the time-domain vibration signal decomposition results displayed by the human-machine interface module of the present invention.

[0031] Figure 3 This is a schematic diagram of the frequency domain vibration signal decomposition results displayed by the human-machine interface module of the present invention.

[0032] Figure 4 This is a schematic diagram of the signal reconstruction result in the time domain displayed by the human-machine interface module of the present invention.

[0033] Figure 5 This is a schematic diagram of the signal reconstruction result in the frequency domain displayed by the human-machine interface module of the present invention.

[0034] Figure 6 This is the fault prediction result displayed by the human-machine interface module of the present invention. Detailed implementation method:

[0035] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific implementation methods and in conjunction with the accompanying drawings.

[0036] like Figure 1-6 As shown, the diesel generator set bearing fault diagnosis system combines TTAO-VMD and IWOA-BP, and the diagnosis system includes:

[0037] The signal acquisition module is used to acquire the vibration signal of the bearing of the monitored diesel generator set and use it as the input signal of the TTAO-VMD fault feature extraction module.

[0038] The TTAO-VMD fault feature extraction module is used to receive bearing vibration signals and calculate the feature parameters of the bearing vibration signals in the time and frequency domains based on the optimal components, thereby constructing a feature vector dataset.

[0039] The IWOA-BP fault mode recognition module is used to classify rolling bearing faults based on a feature vector dataset.

[0040] Human-machine interface module, which is used to display data analysis results and fault types;

[0041] In the TTAO-VMD fault feature extraction module, a triangular aggregation topology optimization algorithm is introduced to optimize the two parameters of the variational mode decomposition method, namely the decomposition order k and the penalty factor α. In the IWOA-BP fault mode recognition module, a chaotic mapping strategy is introduced to initialize the population position in order to improve the global search capability of the whale optimization algorithm.

[0042] The chaotic mapping strategy is as follows:

[0043]

[0044] Where, x i This indicates the position where the i-th individual was generated.

[0045] The IWOA-BP fault mode recognition module introduces an adaptive inertial weight based on the Sigmoid function to improve the whale optimization algorithm's ability to balance exploration and development, and accelerate its convergence speed. The adaptive inertial weight is as follows:

[0046]

[0047] Where, ω max and ω min These represent the maximum and minimum values ​​of the weights, respectively. The coefficient z is used to adjust the steepness of the sigmoid function; m and n are two parameter coefficients; T is the maximum number of iterations.

[0048] An elite retention and perturbation mechanism is introduced in the IWOA-BP fault mode identification module to prevent the whale optimization algorithm from getting trapped in local optima. The perturbation mechanism is as follows:

[0049]

[0050]

[0051] in, γ represents the position of the elite solution; γ is a constant that defines the perturbation amplitude.

[0052] The data displayed on the human-machine interface module includes variational mode decomposition results in the time and frequency domains, variational mode decomposition results in the frequency domain, signal reconstruction results in the time domain, signal reconstruction results in the frequency domain, and fault diagnosis results.

[0053] The working principle of the diesel generator set bearing fault diagnosis system combining TTAO-VMD and IWOA-BP in this embodiment of the invention is as follows: a triangular topology aggregation optimization algorithm is introduced to optimize the decomposition order k and penalty factor α of the variational mode decomposition method, and a TTAO-VMD fault feature extraction module is constructed to avoid reducing the accuracy of the variational mode decomposition method due to improper parameter settings. At the same time, an IWOA-BP fault mode recognition module is also constructed, which can improve the whale optimization algorithm based on the inertia weight strategy of the sigmoid function, thereby improving the stability of the algorithm iteration. Furthermore, a chaotic mapping initialization population and an elite retention and perturbation mechanism are introduced into the IWOA-BP fault mode recognition module to further improve the optimization efficiency and global search capability of the whale optimization algorithm.

[0054] Currently, to address the issue that the decomposition performance of variational mode decomposition (VMD) methods is highly dependent on the decomposition order k and the penalty factor α, existing techniques include: selecting the penalty factor α for each mode using the whale optimization algorithm; adaptively decomposing VMD using the whale optimization algorithm; optimizing the parameters of VMD using the subtractive average optimizer algorithm; optimizing the parameters of VMD using the particle swarm optimization algorithm; adaptively optimizing the parameters of VMD using the locust algorithm; proposing a sparrow search algorithm to optimize the parameters of VMD; using an improved firefly algorithm to optimize the parameters of VMD; adaptively obtaining the optimal parameter combination using the gray wolf algorithm; and optimizing the parameters of VMD using a genetic algorithm.

[0055] As the above analysis shows, intelligent swarm optimization algorithms are currently widely used to optimize the parameters of variational mode decomposition methods. However, intelligent swarm optimization algorithms themselves have many limitations. For example, the optimization performance of particle swarm optimization (PSO) is highly dependent on the setting of its inertia weights; improper parameter selection can easily cause the algorithm to converge too quickly, resulting in the inability to find the global optimum. The optimization effect of genetic algorithms is affected by crossover rate and mutation rate; different parameter combinations have a significant impact on the optimization results. Although sparrow search algorithm shows certain advantages, it is relatively weak in its ability to exploit local optima, and in the later stages of iteration, due to the rapid assimilation of individual sparrows, it is prone to local optima. Locust optimization algorithm struggles to achieve a balance between global search and local exploitation when solving complex engineering optimization problems. Whale optimization algorithm has a slow convergence speed and is prone to getting trapped in local optima when dealing with multi-peak or complex constraint problems. Subtractive average optimizer algorithm, when the search space is large, its subtraction mechanism may lead to a decrease in population diversity, thereby weakening the global exploration ability and causing it to get trapped in local optima in complex or high-dimensional optimization problems.

[0056] In comparison, this invention improves the diesel generator set bearing fault diagnosis system in the following three aspects: First, it introduces a triangular topology aggregation optimization algorithm to optimize the decomposition order k and penalty factor α of the variational mode decomposition method, avoiding a decrease in the accuracy of the variational mode decomposition method due to improper parameter settings. Second, it proposes an improved whale optimization algorithm based on an inertia weight strategy using the sigmoid function to improve the stability of algorithm iteration. Third, it introduces chaotic mapping to initialize the population, as well as elite retention and perturbation mechanisms, to improve the optimization efficiency and global search capability of the whale optimization algorithm.

[0057] In the overall scheme, the diagnostic system includes: a signal acquisition module, which acquires the vibration signal of the bearing of the monitored diesel generator set and uses it as the input signal of the TTAO-VMD fault feature extraction module; the TTAO-VMD fault feature extraction module, which receives the bearing vibration signal and calculates the feature parameters of the bearing vibration signal in the time and frequency domains based on the optimal components to construct a feature vector dataset; an IWOA-BP fault mode recognition module, which classifies the faults of the rolling bearing based on the feature vector dataset; and a human-machine interface module, which displays the data analysis results and fault types. The TTAO-VMD fault feature extraction module introduces a triangular aggregation topology optimization algorithm to optimize the two parameters of the variational mode decomposition method, k and the penalty factor α. The IWOA-BP fault mode recognition module introduces a chaotic mapping strategy to initialize the population position to improve the global search capability of the whale optimization algorithm.

[0058] Specifically, the triangular topology aggregation optimization algorithm optimizes the parameters of the variational mode decomposition method in two stages: global aggregation and local aggregation. First, the population size N and parameter dimension D need to be defined. When the design variables are k and α, and the parameter dimension D is 2, the population size N is divided into [N / 3] triangular topological units during the initialization stage, where [] indicates rounding down. Each triangular topological unit contains three vertices and one randomly generated internal vertex, for a total of four search individuals. Similar triangular topological units are generated through geometric transformation to enhance population diversity.

[0059] Taking the decomposition order k as an example, [N / 3] individuals are randomly generated within the domain of k, and the initial position of each individual is generated as follows:

[0060]

[0061] in, This represents the first searched individual in the i-th triangular topological unit; and are the upper and lower bounds of the decomposition order k, respectively; h is a random number in [0,1].

[0062] Define a direction vector from the initial position. Transform to a normal coordinate system using trigonometric functions to form the second vertex; then, the direction vector... Rotate counterclockwise by π / 3 and obtain the third vertex through coordinate system transformation.

[0063] The expressions for the second and third vertices are:

[0064]

[0065]

[0066] Where l represents the size of the triangular topological unit; and It is the direction vector of the other two edges guided by the first vertex, specifically expressed as:

[0067]

[0068]

[0069] in, It is a random number in the range [0, π].

[0070] Furthermore, in order to achieve a balance between global search and local development, the range of values ​​for the triangular topological unit l needs to be:

[0071]

[0072] Where t represents the current iteration number; T represents the maximum iteration number.

[0073] The fourth vertex, being an internal search point, can be generated using a linear weighted method, specifically:

[0074]

[0075] Among them, r1, r2 and r3 are random numbers between [0,1], and their sum is 1, thus ensuring that the generated fourth vertex is always located within the triangular topological unit.

[0076] Furthermore, in order to comprehensively search for potential optimal k values, the optimal individual of each triangular topological unit needs to be linearly combined with the optimal individual of another random unit to generate a new individual during the global aggregation phase, which can be expressed as:

[0077]

[0078] Where r4 is a random number between [0,1]. and Let i and t represent the optimal individuals of unit i and the random unit, respectively, in the t-th iteration. This is the new individual of unit i in the (t+1)th iteration.

[0079] For the step of retaining the optimal k value, it is necessary to compare the newly generated individual with the current best or second-best individual. The best or second-best individual in the next iteration is represented as:

[0080]

[0081] in, and Let represent the best and second-best individuals in unit i at the (t+1)th iteration; y is the fitness value of the individual.

[0082] To re-search for potentially excellent k-values ​​within a specific region, during the local aggregation phase, the best or second-best individual obtained in the previous phase forms a temporary triangular topological unit between two vertices in another group with better fitness values. A small perturbation is applied based on the position vector difference between the best and second-best individuals to develop each triangular topological unit, represented as:

[0083]

[0084] Where ε represents the set range parameter; This represents the second new individual of unit i in the (t+1)th iteration.

[0085] To ensure the optimality of the convergence direction, the update position is determined by comparing the fitness values ​​of the two individuals before and after aggregation, thus selecting the optimal convergence direction, as shown below:

[0086]

[0087] For the IWOA-BP fault mode identification module, to address the shortcomings of traditional WOA such as insufficient global search capability, slow convergence speed, and susceptibility to local optima, a chaotic mapping strategy is introduced to initialize the population position, improving WOA's global search capability; an adaptive inertial weight based on the sigmoid function is introduced to enhance WOA's balance between exploration and development, accelerating convergence; and an elite retention and perturbation mechanism is introduced to prevent WOA from getting trapped in local optima. Specific methods include:

[0088] (1) Initial population position based on chaotic mapping strategy

[0089] Three mapping methods, Sine, Tent, and Cosine, are used to generate chaotic sequences to initialize the whale population location, increasing the distribution range of the initial solution. This method enables whales to search more regions, helping to improve the global search capability of WOA. The chaotic mapping strategy is expressed as follows:

[0090]

[0091] Where, x i This indicates the position where the i-th individual was generated.

[0092] (2) Inertia weights based on the sigmoid function

[0093] As a dynamically adjustable parameter, the weight can guide the search direction of WOA at different stages and control the balance between global exploration and local development. In view of the problems of slow convergence speed and low solution accuracy in traditional WOA, this invention introduces adaptive weights based on the sigmoid function in the trapping and random search stages of WOA.

[0094] In the early trapping phase of the algorithm, larger weights are assigned, allowing for large-step searches within the search space and preventing premature entrapment in local optima. In the later random search phase, the weights gradually decrease and stabilize, allowing for finer-grained searches with smaller steps, thus improving solution quality. Furthermore, the introduction of the sigmoid function ensures smooth weight changes during iteration, preventing instability caused by drastic weight fluctuations. The adaptive weights are expressed as:

[0095]

[0096] Where, ω max and ωmin These represent the maximum and minimum values ​​of the weights, respectively. The coefficient z is used to adjust the steepness of the sigmoid function; m and n are two parameter coefficients; T is the maximum number of iterations.

[0097] Furthermore, the improved WOA position update formula is as follows:

[0098]

[0099]

[0100]

[0101] (3) Elite preservation and local disturbance mechanism

[0102] In the traditional Word of the Area (WOA) algorithm, each individual updates its position with the target being the best individual at the current iteration number. Due to the influence of control parameters and random numbers in the update formula, even if an individual targets the current best solution during iteration, the updated position may still deviate from the optimal direction, generating a lower-quality solution. Therefore, this invention introduces an elite retention mechanism into the traditional WOA to ensure that the optimal solution is not replaced by a lower-quality solution during iteration, thereby improving the convergence and stability of the algorithm. Furthermore, to further improve the accuracy of the algorithm, this invention adds a local perturbation mechanism on top of retaining elite individuals. This mechanism performs a small perturbation on the position of the elite individual to explore potential optimal candidate solutions in its vicinity, thus more accurately approximating the global optimum. These improvements are expressed as follows:

[0103]

[0104]

[0105] in, γ represents the position of the elite solution; γ is a constant that defines the perturbation amplitude.

[0106] Specifically, in order to verify the effectiveness of the diesel generator set bearing fault diagnosis system proposed in this invention, this application uses the bearing dataset published by Case Western Reserve University for verification, thereby analyzing and evaluating the corresponding performance.

[0107] The intrinsic mode function (IMF) components in the time and frequency domains are respectively as follows: Figure 2 and Figure 3 As shown, according to the appendix Figure 2 and attached Figure 3It can be seen that the calculated kurtosis values ​​of IMF1, IMF2, IMF3, IMF4 and IMF5 are 2.603, 3.379, 4.257, 3.042 and 3.066, respectively. According to the maximum kurtosis criterion, the component with the maximum kurtosis is selected for signal reconstruction. Since IMF3 has the largest kurtosis value, it is selected as the optimal component for signal reconstruction.

[0108] From the appendix Figure 4 It can be seen that the reconstructed signal is highly consistent with the original signal in terms of waveform amplitude and trend. Moreover, the original signal contains more high-frequency components and fluctuations, indicating that the reconstructed signal not only effectively preserves the main characteristics of the original signal, but also significantly reduces the interference of high-frequency noise. The signal reconstruction effect is quite ideal.

[0109] From the appendix Figure 5 It can be clearly seen that the rotational frequency of the rolling bearing is its second harmonic, the inner ring fault frequency is its second harmonic, indicating that the diagnostic system of the present invention can accurately extract the fault characteristic frequency, thus ensuring the accuracy of fault diagnosis.

[0110] As attached Figure 6 As shown, out of 300 fault diagnosis and detection samples, the present invention resulted in 6 misclassified samples, achieving a classification accuracy of 98%. This indicates that the improved whale optimization algorithm demonstrates stronger global search capabilities and higher adaptability in parameter optimization of BP neural network weights and biases, as well as in training the fault mode recognition model. Furthermore, to avoid the influence of random factors on the experimental results, five repeated experiments were conducted, and the average value was taken as the final result.

[0111] Experimental results show that TTAO-VMD significantly improves the accuracy and reliability of fault feature extraction. Furthermore, this invention introduces an inertial weight strategy based on the Sigmoid function, chaotic mapping initialization, elite retention and perturbation mechanism into the whale algorithm, and invents an improved whale optimization algorithm. The improved whale optimization algorithm is used to optimize the initial weights and biases of the BP neural network, and an IWOA-BP fault mode recognition module is invented, which is superior to traditional methods in terms of convergence speed and global search capability.

[0112] In summary, the diesel generator set bearing fault diagnosis system combining TTAO-VMD and IWOA-BP in this embodiment of the invention introduces a triangular topology aggregation optimization algorithm to optimize the decomposition order k and penalty factor α of the variational mode decomposition method, and constructs a TTAO-VMD fault feature extraction module to avoid reducing the accuracy of the variational mode decomposition method due to improper parameter settings. At the same time, an IWOA-BP fault mode recognition module is also constructed, which can improve the whale optimization algorithm based on the inertia weight strategy of the sigmoid function, thereby improving the stability of the algorithm iteration. Furthermore, the IWOA-BP fault mode recognition module introduces chaotic mapping initialization population and elite retention and perturbation mechanisms to further improve the optimization efficiency and global search capability of the whale optimization algorithm.

[0113] The above specific embodiments should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, any alternative improvements or modifications made to the embodiments of the present invention shall fall within the scope of protection of the present invention.

[0114] Any aspects of this invention not described in detail are well-known to those skilled in the art.

Claims

1. A diesel generator set bearing fault diagnosis system combining TTAO-VMD and IWOA-BP, characterized in that, The diagnostic system includes: The signal acquisition module is used to acquire the vibration signal of the bearing of the monitored diesel generator set and use it as the input signal of the TTAO-VMD fault feature extraction module. The TTAO-VMD fault feature extraction module is used to receive bearing vibration signals and calculate the feature parameters of the bearing vibration signals in the time and frequency domains based on the optimal components, thereby constructing a feature vector dataset. The IWOA-BP fault mode recognition module is used to classify rolling bearing faults based on a feature vector dataset. Human-machine interface module, which is used to display data analysis results and fault types; In the TTAO-VMD fault feature extraction module, a triangular aggregation topology optimization algorithm is introduced to optimize the two parameters of the variational mode decomposition method, namely the decomposition order k and the penalty factor α. In the IWOA-BP fault mode recognition module, a chaotic mapping strategy is introduced to initialize the population position in order to improve the global search capability of the whale optimization algorithm. The chaotic mapping strategy is as follows: wherein x i represents the position generated by the ith individual; The IWOA-BP fault mode recognition module introduces an adaptive inertial weight based on the Sigmoid function to improve the whale optimization algorithm's ability to balance exploration and development, and accelerate its convergence speed. The adaptive inertial weight is as follows: Where, ω max and ω min These represent the maximum and minimum values ​​of the weights, respectively; the coefficient z is used to adjust the steepness of the sigmoid function; m and n are two parameter coefficients; T is the maximum number of iterations. An elite retention and perturbation mechanism is introduced in the IWOA-BP fault mode identification module to prevent the whale optimization algorithm from getting trapped in local optima. The perturbation mechanism is as follows: wherein, is the position of the elite solution; is a constant defining the amplitude of the perturbation.

2. The diesel generator set bearing fault diagnostic system incorporating TTAO-VMD and IWOA-BP of claim 1, characterized in that: The data displayed on the human-machine interface module includes variational mode decomposition results in the time and frequency domains, variational mode decomposition results in the frequency domain, signal reconstruction results in the time domain, signal reconstruction results in the frequency domain, and fault diagnosis results.

Images