Diesel generating set intelligent vibration reduction regulation and control method based on vibration signal analysis

Abstract

The invention discloses an intelligent vibration reduction regulation and control method for a diesel generating set based on vibration signal analysis. The intelligent vibration reduction regulation and control method comprises the following steps: S1, collecting multi-channel vibration signals of the diesel generating set; s2, performing preprocessing such as denoising and normalization on the vibration signal; s3, extracting time domain, frequency domain and time-frequency domain characteristic parameters; s4, optimizing the feature weight by adopting a parrot search algorithm, and constructing a recognition model to judge the operation state; s5, after abnormity is recognized, the fuel injection quantity, the injection angle, the air inlet pressure, the rotating speed and other operation control parameters are collected; s6, optimizing the control parameters through a whale optimization algorithm, and minimizing the main vibration amplitude; s7, applying the optimal parameters to a control system to execute adjustment; and S8, collecting an adjusted vibration signal for feedback, updating the model, and realizing closed-loop control and self-learning. According to the invention, intelligent vibration identification and self-adaptive regulation and control of the diesel generating set are realized, and the operation stability and the vibration reduction efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent vibration reduction and control of diesel generator sets, and in particular to an intelligent vibration reduction and control method of a diesel generator set based on vibration signal analysis. Background Art

[0002] As an efficient and reliable energy supply device, diesel generator sets are widely used in key sectors such as communications, electricity, transportation, mining, and healthcare. During operation, they are prone to periodic or aperiodic mechanical vibrations due to factors such as their complex internal structure, high-frequency fluctuations in the combustion system, and interactions between mechanical components. Long-term vibration not only causes wear on the equipment itself, loose connections, and fatigue damage to components, but can also be transmitted through the foundation to the building structure, causing resonance, noise pollution, and structural safety hazards. Therefore, how to effectively identify and suppress abnormal vibration during diesel generator set operation has become a key research topic in equipment operation and maintenance.

[0003] Currently, mainstream approaches to vibration suppression rely on passive vibration isolation devices and fixed-threshold vibration monitoring technologies. While simple in structure, passive vibration isolation devices, such as spring dampers, rubber pads, and metal brackets, are significantly affected by load, frequency, and environmental factors, and lack adaptive capabilities. The latter primarily uses fixed alarm thresholds based on vibration amplitude and frequency, making it difficult to address nonlinear vibration issues under complex, multi-source operating conditions. Furthermore, existing monitoring systems, which mostly rely on single-point, single-channel data acquisition, lack comprehensive awareness of the overall structural state and are unable to accurately locate the source and type of faults. This leads to delayed maintenance decisions and untimely fault warnings.

[0004] On the other hand, some research has begun to explore the use of machine learning methods to identify vibration signals, but most of these efforts focus on model classification accuracy, failing to establish a closed-loop control chain of "identification-control-feedback." Existing methods often lack a mechanism for dynamically optimizing control parameters. Once an abnormal state is identified, manual decision-making or reliance on preset parameters is still required, and diesel engine operating parameters cannot be adaptively adjusted based on actual vibration changes. Furthermore, current optimization methods, which often employ conventional particle swarm optimization and genetic algorithms, have limited global convergence capabilities for multi-objective problems and fail to fully consider the differential weighting of vibration characteristics across multiple channels.

[0005] Based on the above problems, an intelligent vibration reduction and control method for diesel generator sets based on vibration signal analysis is proposed. It can integrate multi-channel vibration signal characteristics and combine with the parrot search algorithm for weight optimization to improve the accuracy and stability of abnormal state recognition; the whale optimization algorithm is used to globally optimize key operating parameters such as injection quantity, injection angle, intake pressure and speed to achieve active control under abnormal vibration conditions; and finally a complete closed-loop control system with vibration recognition, parameter optimization, feedback adjustment and self-learning update is constructed to solve the problems of recognition lag, insufficient control response and non-adaptation in the existing technology, thereby improving the intelligent level of vibration control and operational stability of diesel generator sets.

[0006] Therefore, how to provide an intelligent vibration reduction and control method for diesel generator sets based on vibration signal analysis is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0007] One objective of the present invention is to provide an intelligent vibration reduction and control method for diesel generator sets based on vibration signal analysis. This method fully utilizes multi-channel vibration signal acquisition technology, feature extraction and optimization algorithms, intelligent recognition models, and adaptive parameter control mechanisms. It describes in detail how to optimize vibration feature weights using the Parrot search algorithm and how to dynamically optimize diesel engine operating control parameters using the Whale optimization algorithm, thereby achieving active vibration reduction and control under abnormal vibration conditions. This method offers the advantages of high recognition accuracy, fast response speed, and strong control adaptability, effectively improving the operational stability and intelligence level of diesel generator sets under complex operating conditions.

[0008] According to an embodiment of the present invention, a method for intelligent vibration reduction and control of a diesel generator set based on vibration signal analysis includes the following steps:

[0009] S1. Acquire multi-channel vibration signal data of a diesel generator set under different operating conditions, wherein the vibration signal data includes an acceleration signal and a velocity signal;

[0010] S2. performing denoising, normalization, and wavelet threshold filtering on the vibration signal data to obtain standardized vibration data;

[0011] S3. Extracting characteristic parameters of the standardized vibration data, wherein the characteristic parameters include root mean square value, crest factor, main frequency component and wavelet packet energy distribution;

[0012] S4. Optimize the weights of the characteristic parameters using the Parrot search algorithm, build a vibration recognition model, and identify the operating status of the diesel generator set;

[0013] S5. If the abnormal vibration state is identified, the current operating control parameters of the diesel generator set are collected, wherein the operating control parameters include injection amount, injection angle, intake pressure and speed;

[0014] S6. Setting the minimization of the amplitude of the main vibration frequency as the goal, optimizing the operation control parameters using the whale optimization algorithm to generate the optimal control parameter configuration;

[0015] S7. Apply the optimal control parameter configuration to the diesel generator set control system, perform adjustment operations, and collect the adjusted vibration signal for feedback update model to achieve adaptive vibration reduction control.

[0016] Optionally, the S1 specifically includes:

[0017] S11. Install a multi-channel vibration sensor group at the key structural position of the diesel generator set. The sensor group includes acceleration sensors and velocity sensors installed on the engine cylinder, base, bracket and foundation platform, respectively collecting acceleration signals a at the corresponding positions. i (t) and velocity signal v i (t), where i∈{1,2,…,n} represents the number of the i-th measurement point and t represents the time;

[0018] S12, the original vibration signal a collected at each measuring point i (t), v i (t) Composition of vibration signal data set

[0019] S13, using a unified sampling frequency f s , where f s ≥10kHz, synchronously sample and time-align the vibration signal dataset D to form a time-series continuous original vibration signal matrix Where T is the number of discrete time points corresponding to the sampling duration;

[0020] S14, the vibration signal matrix M D Store and transmit to the vibration feature extraction module for subsequent processing.

[0021] Optionally, the S2 specifically includes:

[0022] S21, the vibration signal matrix obtained in step S1 Perform discrete processing to obtain a discrete vibration signal sequence Where k∈{1,2,…,T} represents a discrete time index in units of sampling intervals

[0023] S22, discrete vibration signal sequence for each channel The denoising process is performed using a soft threshold wavelet denoising method, and the soft threshold function is defined as

[0024]

[0025] Where x is the wavelet transform coefficient and λ is the preset threshold;

[0026] S23, normalize the denoised signal using the maximum and minimum normalization formula

[0027]

[0028] in is the normalized signal;

[0029] S24, the normalized signal sequence Combined into a normalized vibration signal matrix Input to the feature extraction module for subsequent processing.

[0030] Optionally, the S3 specifically includes:

[0031] S31, the normalized vibration signal matrix Performing feature extraction processing, the extracted feature parameter set includes time domain feature parameters, frequency domain feature parameters and time-frequency domain feature parameters;

[0032] S32, time domain characteristic parameters include: root mean square value Crest Factor Kurtosis coefficient where μ i is the mean, σ i is the standard deviation;

[0033] S33, frequency domain characteristic parameters are obtained by Perform fast Fourier transform (FFT) calculation to obtain the spectrum X i (f), calculate the main frequency amplitude A max,i =max|X i (f)|, spectrum center frequency

[0034] S34, time-frequency domain characteristic parameters are obtained by Perform discrete wavelet packet decomposition to obtain the energy of each frequency band where w i,j [k] is the wavelet packet coefficient of the jth layer of the i-th channel, T j is the number of sampling points in this frequency band;

[0035] S35, the characteristic parameters of all channels are combined into a characteristic vector set F = {F1, F2, ..., F n}, where F i Contains all time domain, frequency domain, and time-frequency domain features of the corresponding channel, which are used as training input for subsequent vibration recognition models.

[0036] Optionally, the S4 specifically includes:

[0037] S41, the obtained feature vector set F={F1,F2,…,F n Input to the vibration recognition model building module as the input data set for model training;

[0038] S42. For each eigenvector F i The characteristic component f in i,m Assign initial weight Where m∈{1,2,…,M} represents the feature number, and M is the total number of features for each channel;

[0039] S43, using the parrot search algorithm to Perform global optimization iteration according to the fitness function

[0040]

[0041] Minimize the classification error rate, where W = {w i,m} is the feature weight matrix, y j is the true label of the sample, Output labels for the model, is the indicator function, N is the total number of samples;

[0042] S44. In each iteration, the population memory update mechanism of the parrot search algorithm is used to perform adaptive weight adjustment based on the similarity between each feature individual and the current optimal individual to generate a new generation of feature weight set W. (t+1) ;

[0043] S45. When the maximum iteration number is reached or the convergence condition is met, the optimal feature weight matrix W is output. * , which is used to weight the feature vector set F to generate the weighted feature input set F * ;

[0044] S46, inputting the weighted features into set F * The data is input into a classification model for training. The classification model is one of a support vector machine model, a convolutional neural network model, or a long short-term memory network model, and is used to identify the operating status category of the diesel generator set, including normal operating status, unbalanced status, structural looseness status, and bearing damage status.

[0045] Optionally, the S5 specifically includes:

[0046] S51, the weighted feature input set F received by the vibration recognition model trained in claim 5 * Input into the recognition model and output the corresponding running status prediction label Where j∈{1,2,…,N} represents the sample number, Respectively represent normal state, unbalanced state, loose state and bearing damage state;

[0047] S52, the running state label output by the recognition model is compared with the set abnormal state set Y abn ={y imb ,y loose ,y brg} to compare, if It is determined to be an abnormal vibration state, triggering the control parameter acquisition process;

[0048] S53, after determining that it is in an abnormal vibration state, collect the current diesel generator set operation control parameter set P = {q, θ, p in ,n}, where: q represents the fuel injection amount per unit time; θ represents the fuel injection advance angle; p in Indicates the intake pressure; n indicates the diesel engine speed;

[0049] S54: Input the operation control parameter set P as the parameters to be optimized into the parameter optimization model module.

[0050] Optionally, the S6 specifically includes:

[0051] S61, set the operation control parameter set P = {q, θ, p in ,n} is input as the variable to be optimized into the whale optimization algorithm model to construct the search space

[0052] S62, set the optimization objective function to

[0053]

[0054] Where J(P) is the evaluation function, A main (P) represents the amplitude of the main vibration frequency component of the diesel generator set under the parameter combination P;

[0055] S63. Initialize the number of whale population individuals N w , randomly generate an initial parameter vector for each individual where i∈{1,2,…,N w};

[0056] S64. In each iteration, the parameters are updated according to the position update formula of the whale optimization algorithm, simulating the spiral bubble predation behavior and global search mechanism. The update rules include:

[0057] Exploration Phase Updates:

[0058] Development phase updates: Among them, P * is the current optimal individual, A, C, b, l are algorithm control parameters;

[0059] S65. When the maximum number of iterations or error convergence conditions are met, the optimal parameter solution is output.

[0060] Optionally, the S7 specifically includes:

[0061] S71, solve the optimal parameter P * Input the diesel generator control system to perform corresponding control and adjustment operations, including fuel injection quantity adjustment, injection angle adjustment, intake pressure regulation and speed control;

[0062] S72. After the adjustment is completed, collect new vibration signals Calculate the change in main frequency amplitude Based on the results of ΔA, the data is fed back to the vibration identification model and parameter optimization model to update the sample library and perform model adaptive correction.

[0063] Optionally, the S8 specifically includes:

[0064] S81, configure the optimal operation control parameters P * The data is sent to the diesel generator control system to perform control and adjustment operations, including fuel injection quantity adjustment, injection angle adjustment, intake pressure adjustment and speed control;

[0065] S82: After the control adjustment is completed, re-collect the adjusted vibration signal data to generate an updated normalized vibration signal matrix.

[0066] The beneficial effects of the present invention are:

[0067] (1) The present invention introduces multi-channel vibration signal acquisition and multi-dimensional feature extraction technology, and combines it with the Parrot search algorithm to perform global optimization of feature weights, effectively overcoming the problem of traditional methods relying on single features or fixed weights. It can achieve high-precision recognition of abnormal vibration states (such as imbalance, looseness, bearing damage, etc.) under complex working conditions, and improve the generalization ability and stability of the recognition model.

[0068] (2) The present invention uses the whale optimization algorithm to dynamically optimize key operating control parameters such as injection quantity, injection angle, intake pressure and speed, with minimizing the amplitude of the main vibration frequency as the objective function, ensuring that the control strategy can respond quickly and intervene effectively according to the current recognition state, thereby achieving real-time and active adjustment of abnormal vibrations, overcoming the limitations of traditional control that relies on fixed parameters or manual intervention.

[0069] (3) The present invention forms a closed-loop system structure of "acquisition-identification-optimization-control-feedback-update". Through the continuous updating and self-learning capabilities of the vibration identification model and the optimization model, the system has dynamic adaptability in long-term operation, which not only improves the sustainability of the control effect, but also enhances the system's adaptability to new working conditions, significantly improving the overall intelligence and reliability of the diesel generator set. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0071] Figure 1 This is a flow chart of an intelligent vibration reduction and control method for a diesel generator set based on vibration signal analysis proposed by the present invention;

[0072] Figure 2 This is a structural diagram of a multi-channel vibration signal acquisition and preprocessing module of an intelligent vibration reduction and control method for a diesel generator set based on vibration signal analysis proposed by the present invention;

[0073] Figure 3 This is a schematic diagram of the vibration feature extraction and feature weight optimization processing flow of the intelligent vibration reduction and control method for diesel generator sets based on vibration signal analysis proposed by the present invention. DETAILED DESCRIPTION

[0074] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0075] refer to Figure 1-3 , an intelligent vibration reduction and control method for a diesel generator set based on vibration signal analysis, comprising the following steps:

[0076] S1. Acquire multi-channel vibration signal data of a diesel generator set under different operating conditions, wherein the vibration signal data includes an acceleration signal and a velocity signal;

[0077] S2. performing denoising, normalization, and wavelet threshold filtering on the vibration signal data to obtain standardized vibration data;

[0078] S3. Extracting characteristic parameters of the standardized vibration data, wherein the characteristic parameters include root mean square value, crest factor, main frequency component and wavelet packet energy distribution;

[0079] S4. Optimize the weights of the characteristic parameters using the Parrot search algorithm, build a vibration recognition model, and identify the operating status of the diesel generator set;

[0080] S5. If the abnormal vibration state is identified, the current operating control parameters of the diesel generator set are collected, wherein the operating control parameters include the injection amount, injection angle, intake pressure and speed;

[0081] S6. Setting the minimization of the amplitude of the main vibration frequency as the goal, optimizing the operation control parameters using the whale optimization algorithm to generate the optimal control parameter configuration;

[0082] S7. Apply the optimal control parameter configuration to the diesel generator set control system, perform adjustment operations, and collect the adjusted vibration signal for feedback to update the model to achieve adaptive vibration reduction control.

[0083] In this embodiment, S1 specifically includes:

[0084] S11. Install a multi-channel vibration sensor group at the key structural position of the diesel generator set. The sensor group includes acceleration sensors and velocity sensors installed on the engine cylinder, base, bracket and foundation platform, respectively collecting acceleration signals a at the corresponding positions. i (t) and velocity signal v i (t), where i∈{1,2,…,n} represents the number of the i-th measurement point and t represents the time;

[0085] S12, the original vibration signal a collected at each measuring point i (t), v i (t) Composition of vibration signal data set

[0086] S13, using a unified sampling frequency f s , where f s ≥10kHz, synchronously sample and time-align the vibration signal dataset D to form a time-series continuous original vibration signal matrix Where T is the number of discrete time points corresponding to the sampling duration;

[0087] S14, the vibration signal matrix M D Store and transmit to the vibration feature extraction module for subsequent processing.

[0088] In this embodiment, S2 specifically includes:

[0089] S21, the vibration signal matrix obtained in step S1 Perform discrete processing to obtain a discrete vibration signal sequence Where k∈{1,2,…,T} represents a discrete time index in units of sampling intervals

[0090] S22, discrete vibration signal sequence for each channel The denoising process is performed using a soft threshold wavelet denoising method, and the soft threshold function is defined as

[0091]

[0092] Where x is the wavelet transform coefficient and λ is the preset threshold;

[0093] S23, normalize the denoised signal using the maximum and minimum normalization formula

[0094]

[0095] in is the normalized signal;

[0096] S24, the normalized signal sequence Combined into a normalized vibration signal matrix Input to the feature extraction module for subsequent processing.

[0097] In this embodiment, S3 specifically includes:

[0098] S31, the normalized vibration signal matrix Performing feature extraction processing, the extracted feature parameter set includes time domain feature parameters, frequency domain feature parameters and time-frequency domain feature parameters;

[0099] S32, time domain characteristic parameters include: root mean square value Crest Factor Kurtosis coefficient where μ i is the mean, σ i is the standard deviation;

[0100] S33, frequency domain characteristic parameters are obtained by Perform fast Fourier transform (FFT) calculation to obtain the spectrum X i (f), calculate the main frequency amplitude A max,i =max|X i (f)|, spectrum center frequency

[0101] S34, time-frequency domain characteristic parameters are obtained by Perform discrete wavelet packet decomposition to obtain the energy of each frequency band where w i,j [k] is the wavelet packet coefficient of the jth layer of the i-th channel, T j is the number of sampling points in this frequency band;

[0102] S35, the characteristic parameters of all channels are combined into a characteristic vector set F = {F1, F2, ..., F n}, where F i Contains all time domain, frequency domain, and time-frequency domain features of the corresponding channel, which are used as training input for subsequent vibration recognition models.

[0103] In this embodiment, the S4 specifically includes:

[0104] S41, the obtained feature vector set F={F1,F2,…,F n Input to the vibration recognition model building module as the input data set for model training;

[0105] S42. For each eigenvector F i The characteristic component f in i,m Assign initial weight Where m∈{1,2,…,M} represents the feature number, and M is the total number of features for each channel;

[0106] S43, using the parrot search algorithm to Perform global optimization iteration according to the fitness function

[0107]

[0108] Minimize the classification error rate, where W = {w i,m} is the feature weight matrix, y j is the true label of the sample, Output labels for the model, is the indicator function, N is the total number of samples;

[0109] S44. In each iteration, the population memory update mechanism of the parrot search algorithm is used to perform adaptive weight adjustment based on the similarity between each feature individual and the current optimal individual to generate a new generation of feature weight set W. (t+1) ;

[0110] S45. When the maximum iteration number is reached or the convergence condition is met, the optimal feature weight matrix W is output. * , which is used to weight the feature vector set F to generate the weighted feature input set F * ;

[0111] S46, inputting the weighted features into set F * The data is input into a classification model for training. The classification model is one of a support vector machine model, a convolutional neural network model, or a long short-term memory network model, and is used to identify the operating status category of the diesel generator set, including normal operating status, unbalanced status, structural looseness status, and bearing damage status.

[0112] In this embodiment, the S5 specifically includes:

[0113] S51, the weighted feature input set F received by the vibration recognition model trained in claim 5 * Input into the recognition model and output the corresponding running status prediction label Where j∈{1,2,…,N} represents the sample number, Respectively represent normal state, unbalanced state, loose state and bearing damage state;

[0114] S52, the running state label output by the recognition model is compared with the set abnormal state set Y abn ={y imb ,y loose ,y brg} to compare, if It is determined to be an abnormal vibration state, triggering the control parameter acquisition process;

[0115] S53, after determining that it is in an abnormal vibration state, collect the current diesel generator set operation control parameter set P = {q, θ, p in ,n}, where: q represents the fuel injection amount per unit time; θ represents the fuel injection advance angle; p in Indicates the intake pressure; n indicates the diesel engine speed;

[0116] S54: Input the operation control parameter set P as the parameters to be optimized into the parameter optimization model module.

[0117] In this embodiment, S6 specifically includes:

[0118] S61, set the operation control parameter set P = {q, θ, p in ,n} is input as the variable to be optimized into the whale optimization algorithm model to construct the search space

[0119] S62, set the optimization objective function to

[0120]

[0121] Where J(P) is the evaluation function, A main(P) represents the amplitude of the main vibration frequency component of the diesel generator set under the parameter combination P;

[0122] S63. Initialize the number of whale population individuals N w , randomly generate an initial parameter vector for each individual where i∈{1,2,…,N w};

[0123] S64. In each iteration, the parameters are updated according to the position update formula of the whale optimization algorithm, simulating the spiral bubble predation behavior and global search mechanism. The update rules include:

[0124] Exploration Phase Updates:

[0125] Development phase updates: Among them, P * is the current optimal individual, A, C, b, l are algorithm control parameters;

[0126] S65. When the maximum number of iterations or error convergence conditions are met, the optimal parameter solution is output.

[0127] In this embodiment, the S7 specifically includes:

[0128] S71, solve the optimal parameter P * Input the diesel generator control system to perform corresponding control and adjustment operations, including fuel injection quantity adjustment, injection angle adjustment, intake pressure regulation and speed control;

[0129] S72. After the adjustment is completed, collect new vibration signals Calculate the change in main frequency amplitude Based on the results of ΔA, the data is fed back to the vibration identification model and parameter optimization model to update the sample library and perform model adaptive correction.

[0130] In this embodiment, S8 specifically includes:

[0131] S81, configure the optimal operation control parameters P * The data is sent to the diesel generator control system to perform control and adjustment operations, including fuel injection quantity adjustment, injection angle adjustment, intake pressure adjustment and speed control;

[0132] S82: After the control adjustment is completed, re-collect the adjusted vibration signal data to generate an updated normalized vibration signal matrix.

[0133] Example 1:

[0134] To verify the feasibility of this invention, the invention was applied to the vibration state identification and adaptive vibration reduction control system of a standby diesel generator set in a mining area. The equipment was installed in the power supply system of a dispatching center at an open-pit coal mine in Ordos, Inner Mongolia. The generator set was a Cummins QST30-G4 diesel generator with a rated power of 1000kW. The project was based on the fact that during peak mining operations, the mine relied on a diesel generator system to provide a stable power source for communication dispatching and electric transportation systems. However, during long-term operation, the equipment experienced frequent structural resonance and abnormal bearing wear, seriously affecting operational efficiency and system stability.

[0135] Traditional equipment monitoring systems are based solely on threshold judgments, such as setting acceleration to exceed 12 mm / s. 2 Alarms are triggered, but in complex operating conditions with overlapping multiple vibration modes, the false alarm rate is high. Furthermore, manual identification of the fault type and execution of shutdown procedures are still required after the alarm is triggered, significantly reducing equipment availability. Based on this, the present invention is implemented in this scenario, aiming to achieve automatic identification of vibration states, optimal control of operating parameters, and closed-loop learning capabilities for the system, thereby comprehensively improving operational safety and maintenance efficiency.

[0136] During the implementation process, four groups of acceleration and velocity sensors were respectively arranged at four key parts of the diesel generator set, namely the side cover, the front of the cylinder, the output flange and the base platform, to form an eight-channel synchronous acquisition system. The sampling frequency is 20kHz and the sampling time is 30 seconds / cycle. The operation data is collected every 2 hours and uploaded to the edge computing processing unit. The signal is denoised and normalized by wavelet threshold in the processing module to extract 24-dimensional composite feature parameters such as RMS, peak factor, main frequency, and spectrum energy. The parrot search algorithm described in the present invention is used to optimize the feature weights, and the weighted features are input into the LSTM recognition model. The classification results show that the state recognition accuracy is stably maintained at 96.3%, which is much higher than the 84.7% of the traditional KNN model.

[0137] After the system identified three consecutive instances of imbalance + bearing damage, the whale optimization algorithm module was immediately activated to optimize the four parameters of injection volume (q), injection angle (θ), intake pressure (p_in) and speed (n). The optimization goal is to minimize the main frequency amplitude and control the response cycle within 10 seconds. After the system was adjusted, data was collected again, and the feedback vibration main frequency dropped from the original 53.2Hz to 47.8Hz, corresponding to the main frequency amplitude of 8.43mm / s 2 Down to 3.25mm / s 2 , the vibration reduction effect is obvious.

[0138] During its 72 hours of continuous operation, the system completed 38 state recognition and automatic closed-loop control responses, without any false stops or over-vibration alarms. Compared to pre-implementation data, the average bearing housing temperature rise dropped from 64.5°C to 57.1°C, extending the projected service life of key components by 23.4%. Inspection frequency for maintenance personnel was reduced from twice per shift to once per day, reducing the manual intervention rate by over 50%.

[0139] Table 1 Comparison of vibration characteristics and operating status of diesel generator sets before and after the implementation of the present invention

[0140]

[0141] With respect to the data results in Table 1 "Comparative Data Table of Vibration Characteristics and Operating Status of Diesel Generator Sets Before and After the Implementation of the Invention", the following analysis is conducted in combination with the actual operating results:

[0142] After the implementation of this invention, the main vibration frequency of the diesel generator set dropped from the original 53.2Hz to 47.8Hz, a decrease of about 10.1%. This frequency shift is significantly away from the original structural resonance range of the equipment, indicating that by adjusting the operating parameters, high vibration risk areas can be effectively avoided, reducing the possibility of resonance at the source. More importantly, the main frequency amplitude has dropped from 8.43mm / s 2 Down to 3.25mm / s 2 , the vibration intensity is reduced by 61.4%, which fully demonstrates that the "identification + optimization + regulation" mechanism proposed in the present invention has a significant effect on vibration control.

[0143] In terms of temperature, the bearing temperature dropped from 64.5°C to 57.1°C, a decrease of 11.5%. This reflects that after the vibration is reduced, the friction and wear of the equipment are reduced, and the operating load is more stable, which leads to a decrease in heat accumulation. This effect helps to extend the service life of mechanical components.

[0144] In terms of recognition models, the LSTM model optimized by the Parrot search algorithm used in this invention achieved a recognition accuracy of 96.3%, a 13.7% improvement over the traditional model. This high-accuracy recognition capability provides a more reliable basis for subsequent control decisions, avoiding the risk of misadjustment or miscontrol.

[0145] In terms of operational stability, the system previously experienced five vibration over-limit alarms within 72 hours, but these were completely eliminated after implementing the present invention, significantly improving stability. Furthermore, the average adjustment response time was reduced from ≥45 seconds with manual operation to ≤10 seconds with automated control, a 77.8% reduction. This demonstrates the present invention's high efficiency and timeliness at the control execution level, enabling rapid response to abnormal conditions.

[0146] Improvements in operations and maintenance have also been significant. The frequency of manual intervention has been reduced by 50%, from twice per shift to once per day. This significantly reduces operational pressure and labor costs. Furthermore, based on a fatigue damage accumulation theory model, the estimated lifespan of key components has increased by 23.4%, directly reducing long-term equipment repair and replacement costs.

[0147] Comprehensive analysis of the above indicators shows that the present invention not only achieves the functional goals of accurate identification, effective control, and closed-loop stability in the technical dimension, but also demonstrates high reliability, good engineering adaptability, and significant economic benefits in actual operation, and has strong promotion and application value.

[0148] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A diesel generator set intelligent vibration reduction control method based on vibration signal analysis, characterized in that: The steps include: S1. Acquire multi-channel vibration signal data of a diesel generator set under different operating conditions, wherein the vibration signal data includes an acceleration signal and a velocity signal; S2. performing denoising, normalization, and wavelet threshold filtering on the vibration signal data to obtain standardized vibration data; S3. Extracting characteristic parameters of the standardized vibration data, wherein the characteristic parameters include root mean square value, crest factor, main frequency component and wavelet packet energy distribution; S4. Optimize the weights of the characteristic parameters using the Parrot search algorithm, build a vibration recognition model, and identify the operating status of the diesel generator set; S5. If the abnormal vibration state is identified, the current operating control parameters of the diesel generator set are collected, wherein the operating control parameters include the injection amount, injection angle, intake pressure and speed; S6. Setting the minimization of the amplitude of the main vibration frequency as the goal, optimizing the operation control parameters using the whale optimization algorithm to generate the optimal control parameter configuration; S7. Apply the optimal control parameter configuration to the diesel generator set control system, perform adjustment operations, and collect the adjusted vibration signal for feedback to update the model to achieve adaptive vibration reduction control.

2. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: Said S1 specifically includes: S11. Install a multi-channel vibration sensor group at the key structural position of the diesel generator set. The sensor group includes acceleration sensors and velocity sensors installed on the engine cylinder, base, bracket and foundation platform, respectively collecting acceleration signals a at the corresponding positions. i (t) and velocity signal v i (t), where i∈{1,2,…,n} represents the number of the i-th measurement point and t represents the time; S12, the original vibration signal a collected at each measuring point i (t), v i (t) Composition of vibration signal data set S13, using a unified sampling frequency f s , where f s ≥10kHz, synchronously sample and time-align the vibration signal dataset D to form a time-series continuous original vibration signal matrix Where T is the number of discrete time points corresponding to the sampling duration; S14, the vibration signal matrix M D Store and transmit to the vibration feature extraction module for subsequent processing.

3. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: The S2 specifically includes: S21, the vibration signal matrix obtained in step S1 Perform discrete processing to obtain a discrete vibration signal sequence Where k∈{1,2,…,T} represents a discrete time index in units of sampling intervals S22, discrete vibration signal sequence for each channel The denoising process is performed using a soft threshold wavelet denoising method, and the soft threshold function is defined as Where x is the wavelet transform coefficient and λ is the preset threshold; S23, normalize the denoised signal using the maximum and minimum normalization formula in is the normalized signal; S24, the normalized signal sequence Combined into a normalized vibration signal matrix Input to the feature extraction module for subsequent processing.

4. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: The S3 specifically includes: S31, the normalized vibration signal matrix Performing feature extraction processing, the extracted feature parameter set includes time domain feature parameters, frequency domain feature parameters and time-frequency domain feature parameters; S32, time domain characteristic parameters include: root mean square value Crest Factor Kurtosis coefficient where μ i is the mean, σ i is the standard deviation; S33, frequency domain characteristic parameters are obtained by Perform fast Fourier transform (FFT) calculation to obtain the spectrum X i (f), calculate the main frequency amplitude A max,i =max|X i (f)|, spectrum center frequency S34, time-frequency domain characteristic parameters are obtained by Perform discrete wavelet packet decomposition to obtain the energy of each frequency band where w i,j [k] is the wavelet packet coefficient of the jth layer of the i-th channel, T j is the number of sampling points in this frequency band; S35, the characteristic parameters of all channels are combined into a characteristic vector set F = {F1, F2, ..., F n }, where F i Contains all time domain, frequency domain, and time-frequency domain features of the corresponding channel, which are used as training input for subsequent vibration recognition models.

5. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: The S4 specifically includes: S41, the obtained feature vector set F={F1,F2,…,F n Input to the vibration recognition model building module as the input data set for model training; S42. For each eigenvector F i The characteristic component f in i,m Assign initial weight Where m∈{1,2,…,M} represents the feature number, and M is the total number of features for each channel; S43, using the parrot search algorithm to Perform global optimization iteration according to the fitness function Minimize the classification error rate, where W = {w i,m } is the feature weight matrix, y j is the true label of the sample, Output labels for the model, is the indicator function, N is the total number of samples; S44. In each iteration, the population memory update mechanism of the parrot search algorithm is used to perform adaptive weight adjustment based on the similarity between each feature individual and the current optimal individual to generate a new generation of feature weight set W. (t+1) ; S45. When the maximum iteration number is reached or the convergence condition is met, the optimal feature weight matrix W is output. * , which is used to weight the feature vector set F to generate the weighted feature input set F * ; S46, inputting the weighted features into set F * The data is input into a classification model for training. The classification model is one of a support vector machine model, a convolutional neural network model, or a long short-term memory network model, and is used to identify the operating status category of the diesel generator set, including normal operating status, unbalanced status, structural looseness status, and bearing damage status.

6. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: The S5 specifically includes: S51, the weighted feature input set F received by the vibration recognition model trained in claim 5 * Input into the recognition model and output the corresponding running status prediction label Where j∈{1,2,…,N} represents the sample number, Respectively represent normal state, unbalanced state, loose state and bearing damage state; S52, the running state label output by the recognition model is compared with the set abnormal state set Y abn ={y imb ,y loose ,y brg } for comparison, if It is determined to be an abnormal vibration state, triggering the control parameter acquisition process; S53, after determining that it is in an abnormal vibration state, collect the current diesel generator set operation control parameter set P = {q, θ, p in ,n}, where: q represents the fuel injection amount per unit time; θ represents the fuel injection advance angle; p in Indicates the intake pressure; n indicates the diesel engine speed; S54: Input the operation control parameter set P as the parameters to be optimized into the parameter optimization model module.

7. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: The S6 specifically includes: S61, set the operation control parameter set P = {q, θ, p in ,n} is input as the variable to be optimized into the whale optimization algorithm model to construct the search space S62, set the optimization objective function to Where J(P) is the evaluation function, A main (P) represents the amplitude of the main vibration frequency component of the diesel generator set under the parameter combination P; S63. Initialize the number of whale population individuals N w , randomly generate an initial parameter vector for each individual where i∈{1,2,…,N w }; S64. In each iteration, the parameters are updated according to the position update formula of the whale optimization algorithm, simulating the spiral bubble predation behavior and global search mechanism. The update rules include: Exploration Phase Updates: Development phase updates: Among them, P * is the current optimal individual, A, C, b, l are algorithm control parameters; S65. When the maximum number of iterations or error convergence conditions are met, the optimal parameter solution is output.

8. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: The S7 specifically includes: S71, solve the optimal parameter P * Input the diesel generator control system to perform corresponding control and adjustment operations, including fuel injection quantity adjustment, injection angle adjustment, intake pressure regulation and speed control; S72. After the adjustment is completed, collect new vibration signals Calculate the change in main frequency amplitude Based on the results of ΔA, the data is fed back to the vibration identification model and parameter optimization model to update the sample library and perform model adaptive correction.

9. The intelligent vibration reduction control method for diesel generator sets based on vibration signal analysis according to claim 1 is characterized in that: The S8 specifically includes: S81, configure the optimal operation control parameters P * The data is sent to the diesel generator control system to perform control and adjustment operations, including fuel injection quantity adjustment, injection angle adjustment, intake pressure adjustment and speed control; S82: After the control adjustment is completed, re-collect the adjusted vibration signal data to generate an updated normalized vibration signal matrix.

Images