Gear performance accurate evaluation method based on voiceprint analysis

Abstract

The invention discloses a gear performance accurate evaluation method based on voiceprint analysis, and relates to the field of gear performance evaluation. Determining the optimal arrangement position of the acoustic sensor according to different gears, and adjusting the sampling frequency and duration in real time according to the operation state change of the gears to collect voiceprint data; de-noising by using a composite de-noising network based on deep learning, and processing data by using a new normalization method; extracting a plurality of innovative features to construct a feature vector, and combining a deep forest with model fusion to establish an evaluation model; and inputting a to-be-evaluated gear feature vector into the model to evaluate the performance, verifying an evaluation result through a virtual verification mode and the like, and optimizing the model based on feedback. The gear performance is accurately evaluated, voiceprint data are intelligently collected, and the evaluation accuracy is improved through feature extraction and model construction; real-time monitoring and early warning are achieved, and different gears are adapted; the block chain ensures data reliability, the evaluation model evolves continuously, the maintenance cost is reduced, and stable operation of equipment is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of gear performance evaluation, and in particular to a method for accurately evaluating gear performance based on voiceprint analysis. Background Art

[0002] In modern industrial production, gears, as key transmission components of mechanical equipment, are widely used in numerous fields, including automotive, aerospace, shipbuilding, and machine tools. Gear performance is directly related to the operational stability, reliability, and service life of the entire equipment. Gear failure can not only cause equipment downtime and impact production efficiency, but can also lead to serious safety accidents. Therefore, accurately evaluating gear performance, identifying potential problems promptly, and implementing appropriate measures are crucial for ensuring safe and efficient industrial production.

[0003] Traditional gear performance evaluation methods have numerous limitations. Methods based on physical inspection, such as direct observation of gear wear and measurement of tooth profile accuracy, often require equipment disassembly and downtime, which is cumbersome and destructive, making real-time online monitoring impossible. Methods that rely on vibration signal analysis, while able to reflect the operating status of the gear to a certain extent, are susceptible to factors such as overall equipment vibration and noise interference, resulting in inaccurate evaluation results. Furthermore, these methods can typically only detect obvious gear faults that have already occurred, making it difficult to effectively diagnose and predict performance of potential faults in the early stages.

[0004] With the development of industrial intelligence, higher requirements are placed on the accuracy, real-time nature and intelligence of gear performance evaluation. Voiceprint analysis technology, as a means of non-contact, real-time monitoring, has gradually attracted attention. However, the gear performance evaluation technology based on voiceprint analysis is still in the development stage. In terms of data acquisition, the existing voiceprint analysis methods have difficulty in comprehensively and accurately obtaining voiceprint signals that can reflect the actual operating status of the gears, and the acquisition process is easily affected by environmental noise. In terms of feature extraction and model establishment, they lack sufficient pertinence and innovation, and cannot fully tap the rich information contained in the voiceprint signals, resulting in insufficient accuracy and generalization capabilities of the evaluation model, making it difficult to meet the needs of complex and changing industrial application scenarios. Therefore, it is urgent to develop a more efficient and accurate gear performance evaluation method based on voiceprint analysis. Summary of the Invention

[0005] The present invention proposes a method for accurately evaluating gear performance based on soundprint analysis to solve the problems mentioned in the above-mentioned prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for accurately evaluating gear performance based on soundprint analysis, comprising: Voiceprint data collection steps: An intelligent sensor layout algorithm is used to determine the optimal position of the acoustic sensor for different gears. An adaptive sampling strategy is introduced into the collection process to adjust the sampling frequency and duration in real time according to the gear operating status; Voiceprint data preprocessing steps: Use convolutional neural networks (CNN) combined with recurrent neural networks (RNN) for denoising. The network is trained on noisy and noise-free data to adaptively identify and remove noise. An improved normalization method is proposed, using a quantile normalization formula to process voiceprint data. Voiceprint feature extraction steps: Extract time domain, frequency domain, time-frequency domain, and phase space reconstruction features, reconstruct the voiceprint time series through the delay embedding method, extract features to capture early fault information, introduce multi-scale entropy features, and decompose the voiceprint signal at multiple scales to calculate the entropy value; Steps for constructing feature vectors: adopt a feature importance ranking method, calculate the correlation between features and gear performance indicators, determine the score based on mutual information, select high-scoring features to form feature vectors, and remove redundant and irrelevant features; Steps to establish the evaluation model: Build an evaluation model based on deep forest, combining decision trees and deep learning ideas, use an adaptive forest growth strategy to train the model, introduce a model fusion mechanism to integrate deep forest and traditional machine learning models, and use a cross-validation optimization algorithm to dynamically adjust weights; Gear performance evaluation steps: Before inputting the feature vector into the model, the feature vector is compared with the historical feature vector to calculate the similarity. The evaluation model outputs the gear performance status, remaining life and confidence interval. The uncertainty of the evaluation results is quantified through Monte Carlo simulation and Bayesian inference methods. Result verification and feedback steps: Introduce virtual verification technology, compare simulated and actual voiceprint features, establish an adaptive optimization mechanism based on feedback control theory, and adjust the evaluation model parameters and structure according to the deviation.

[0007] Furthermore, the voiceprint data collection step adopts acoustic metamaterial-assisted collection technology. By optimizing the acoustic metamaterial structure and parameters to match the target gear voiceprint characteristic frequency, the data is fused and processed by combining the sound, vibration, temperature and stress sensor fusion technology. The information entropy fusion algorithm is used to determine the weight according to the size of the sensor data information entropy. The fusion formula is: , S fused is the fused signal, S i is the ith sensor signal, , H i is the information entropy of the i-th sensor data.

[0008] Furthermore, the voiceprint feature extraction step adopts a feature extraction method based on sparse representation. By solving the sparse optimization problem, the signal sparse representation coefficient is obtained, and the higher-order cumulants are calculated by combining the high-order cumulants feature. The high-order cumulants calculation formula is: , E is the mathematical expectation, and k is the order of the cumulant.

[0009] Furthermore, the step of constructing the feature vector adopts a feature fusion method based on graph neural network (GNN), uses GNN learning to process feature graphs, aggregates node information, explores potential relationships between features, introduces a dynamic feature update mechanism, regularly updates and optimizes feature vectors, calculates feature importance scores based on collected data, adds valuable features, and removes irrelevant features.

[0010] Furthermore, the step of establishing the evaluation model adopts the generative adversarial network (GAN) model enhancement technology. GAN consists of a generator and a discriminator. The generator generates simulated voiceprint feature vectors and performance indicators, and the discriminator distinguishes real data from generated data. The training data set is expanded by training the generator and discriminator, and the evaluation model is trained using the data set. The meta-learning mechanism is introduced to learn the optimization strategy through training.

[0011] Furthermore, the gear performance evaluation step adopts a multi-dimensional evaluation strategy to evaluate the overall performance status, remaining life, tooth surface contact strength, tooth root bending strength, and lubrication status of the gear. A corresponding sub-evaluation model is established for each dimension, and the evaluation results of the sub-evaluation model are comprehensively analyzed. Trend analysis and prediction technology are introduced. Based on historical evaluation data and current evaluation results, a time series analysis method is used to predict the trend of gear performance changes.

[0012] Furthermore, the result verification and feedback steps adopt a blockchain-based verification and feedback system, recording voiceprint data, feature vectors, evaluation results and actual test results on the blockchain, and using the blockchain smart contract function to verify and feedback the evaluation results. If there is a deviation between the evaluation results and the actual test results, the smart contract automatically triggers the evaluation model optimization process, and shares and exchanges evaluation data and experience through the blockchain network.

[0013] Furthermore, the sampling frequency adjustment formula in the voiceprint data collection step is: , f s (t) is the sampling frequency at time t, f s0 is the initial sampling frequency, k is the adjustment coefficient, Δv is the change in speed or load, and v0 is the initial speed or load; A distributed voiceprint collection network is adopted, and multiple acoustic sensors are arranged in different parts of the gear system to form a distributed collection network. The data collected by each sensor is transmitted to the central processing unit through wireless communication technology. Synchronous collection technology is used to collect data simultaneously, and encryption algorithms are used to encrypt data during the data transmission process.

[0014] Furthermore, the voiceprint feature extraction step adopts an adaptive spectral clustering feature classification method, divides the data points through the adaptive spectral clustering algorithm, determines the number of clusters and cluster centers, dynamically adjusts according to the feature distribution, identifies the gear operating status and fault type by analyzing the category features, combines the fuzzy feature extraction method, introduces the fuzzy membership function, maps the voiceprint feature values to different fuzzy sets to obtain the fuzzy feature vector; the evaluation model establishment step adopts the federated learning method to train the model, aggregates the local model parameters through the federated learning algorithm to obtain the global evaluation model, and introduces the model compression technology to compress the evaluation model.

[0015] Furthermore, when the equipment changes the gear supplier, the voiceprint data of the new gear after installation is collected through voiceprint analysis, and the feature vectors of time domain, frequency domain, time-frequency domain and phase space reconstruction are extracted, and the similarity is matched with the original gear feature vector in the historical database. The similarity calculation uses the cosine distance formula , is the new gear eigenvector, is the feature vector of the original gear; if the similarity is higher than the set threshold and the performance status and remaining life indicators output by the evaluation model are not significantly different from those of the original gear, it is determined that the new gear replaces the original gear, supporting the supplier switching decision.

[0016] Compared with the existing technology, the beneficial effects of the present invention are: In terms of assessment accuracy, intelligent sensor layout and adaptive sampling strategies enable the collection of more representative and accurate voiceprint data. Furthermore, innovative feature extraction methods, such as those combining phase space reconstruction, multiscale entropy, and sparse representation-based feature extraction, can deeply explore the nonlinear, non-Gaussian, and complex dynamic characteristics of voiceprint signals, comprehensively reflecting the operating status of gears. These features, combined with assessment models based on deep forest and model fusion technologies, significantly improve assessment accuracy, enabling precise assessment of gear health, prediction of remaining life, and identification of potential faults, reducing equipment failures and production interruptions caused by misjudgments.

[0017] In terms of real-time monitoring and early warning capabilities, the use of a distributed voiceprint collection network and real-time monitoring technology enables comprehensive, real-time collection of gear voiceprint signals. If a voiceprint signal anomaly occurs, the system quickly issues an alert, prompting staff to take timely action. This shifts equipment maintenance from traditional, scheduled maintenance to precise maintenance based on actual operating conditions. This not only reduces maintenance costs but also effectively avoids the serious consequences of untimely fault detection, ensuring the safe and stable operation of the equipment.

[0018] In terms of versatility and adaptability, the evaluation method of this invention, by incorporating technologies such as transfer learning and meta-learning, can rapidly adapt to the needs of gear performance evaluation under different types and operating conditions. Whether it is new gear equipment or existing equipment with changed operating conditions, its performance can be accurately assessed, improving the versatility of the evaluation method and reducing the technical barriers and costs for enterprises in equipment monitoring and maintenance.

[0019] Furthermore, the use of blockchain technology for result verification and feedback ensures data authenticity and traceability, enhancing the credibility of the assessment. By establishing a historical database and employing an adaptive optimization mechanism, the assessment model can continuously learn and evolve. As data accumulates, the accuracy and reliability of the assessment will continue to improve, providing long-term and stable technical support for gear performance evaluation in industrial production. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a schematic block diagram of a method for accurately evaluating gear performance based on soundprint analysis proposed in the present invention; Figure 2 This is a schematic diagram of voiceprint feature comparison; Figure 3 A schematic diagram for comparing the accuracy of the evaluation model; Figure 4 This is a schematic diagram for predicting the remaining life of the gear; Figure 5 Schematic diagram of the change in sample entropy of voiceprint signals at different scales. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0022] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0023] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below with reference to the accompanying drawings.

[0024] Reference Figures 1 to 5 : A method for accurately evaluating gear performance based on soundprint analysis, including: Voiceprint data collection steps: For gears of different types and application scenarios, an intelligent sensor layout algorithm is used to determine the optimal layout position of the sound sensor. This algorithm comprehensively considers factors such as the geometric structure of the gear, transmission mode, sound propagation path, and the distribution of ambient noise. By establishing a sound propagation model and a noise interference model, it calculates the signal-to-noise ratio and feature integrity of the soundprint signal collected at each potential position, thereby selecting the position where the most representative soundprint signal can be obtained. During the collection process, an adaptive sampling strategy is introduced. According to the dynamic changes in the gear operating state, such as fluctuations in speed and load, the sampling frequency and sampling duration are adjusted in real time. When the gear operating state is relatively stable, the sampling frequency is appropriately reduced to reduce the amount of data; when the operating state changes drastically, the sampling frequency is increased to capture more detailed information. The sampling frequency adjustment formula is: , where f s (t) is the sampling frequency at time t, f s0 is the initial sampling frequency, k is the adjustment coefficient, Δv is the change in speed or load, and v0 is the initial speed or load.

[0025] Voiceprint data preprocessing steps: De-noising is performed using a composite denoising network based on deep learning. This network combines the advantages of convolutional neural networks (CNNs) and recurrent neural networks (RNNs). CNNs are used to extract local features of voiceprint signals, while RNNs are used to process the signal's temporal information. The network is trained on a large amount of noisy and noise-free voiceprint data, enabling it to adaptively identify and remove different types of noise. An improved normalization method is proposed that considers not only the amplitude range of the voiceprint signal but also the statistical distribution characteristics of the signal. The quantile-based normalization formula is used: , where Q1 and Q3 are the first quartile and the third quartile of the voiceprint data respectively, x is the original voiceprint data, x norm This method can better handle voiceprint data with outliers.

[0026] Voiceprint feature extraction steps: In addition to traditional time domain, frequency domain and time-frequency domain features, features based on phase space reconstruction are also extracted. The phase space reconstruction theory is applied to process the voiceprint signal, and the one-dimensional voiceprint time series is reconstructed into a high-dimensional phase space through the delayed embedding method. Features such as correlation dimension and maximum Lyapunov exponent are extracted from the phase space. These features can reflect the nonlinear dynamic characteristics of the voiceprint signal and sensitively capture early gear fault information. Multi-scale entropy features are introduced to perform multi-scale decomposition of the voiceprint signal and calculate the sample entropy or fuzzy entropy at each scale. Multi-scale entropy describes the complexity and irregularity of the voiceprint signal at different scales, and comprehensively reflects the operating status of the gear. The formula for calculating multi-scale entropy is: , , where m is the embedding dimension, r is the similarity tolerance, T is the delay time, k is the scale factor, and SampEn is the sample entropy function.

[0027] Steps for constructing feature vectors: During the feature extraction phase, various features are extracted from multiple sources of gear operation data, including vibration signals collected by vibration sensors, temperature change data recorded by temperature sensors, and torque values output by torque sensors. The extracted signal features include time domain features (such as mean, variance, and peak value), frequency domain features (such as frequency content and power spectral density), and time-frequency domain features (such as wavelet transform coefficients). The Pearson correlation coefficient is used to calculate correlation metrics, accurately measuring the strength of the linear correlation between each feature and gear performance indicators (such as wear level and transmission efficiency). Values closer to 1 or -1 indicate stronger correlation, while values closer to 0 indicate weaker correlation. The calculation of the mutual information metric is based on the principles of information theory. By calculating the information entropy and conditional entropy between the feature and the performance indicator, the amount of information provided by the feature to the performance indicator is determined. Higher mutual information values indicate that the feature contains more effective information in describing gear performance. After evaluation, each feature is assigned an importance score based on the calculated correlation and mutual information metrics, and a strict ranking is performed. When selecting features to construct the final feature vector, a reasonable threshold is set to filter out features with importance scores above the threshold. Furthermore, cluster analysis and other methods are used to further eliminate important but redundant features. This not only removes redundant features irrelevant to gear performance, avoiding the "curse of dimensionality" during model training, but also significantly improves the quality of the feature vector, thereby significantly enhancing the training efficiency and prediction accuracy of the evaluation model.

[0028] Evaluation model building steps: Build an evaluation model based on Deep Forest, an ensemble learning model that combines decision trees and deep learning. The initial number of forest layers and decision trees is determined based on the dimensionality and complexity of the gear feature vector. Each layer of the decision tree forest acts as a feature processor, performing a progressively more refined analysis of the input feature vector. During training, Deep Forest employs an adaptive forest growth strategy, dynamically adjusting the forest structure by calculating the training error. If the training error decreases slowly at a certain stage, the system automatically increases the number of forest layers to further explore the hidden information in the feature vector. Simultaneously, the number of decision trees is appropriately increased based on data distribution and feature correlation to ensure that the complex mapping relationship between feature vectors and gear performance indicators is fully captured. This adaptive adjustment mechanism significantly improves the model's generalization and accuracy, making it more adaptable to performance evaluation of gears under different operating conditions and types. To further enhance the reliability of the evaluation results, a model fusion mechanism is introduced, integrating the Deep Forest model with traditional machine learning models such as support vector machines and neural networks. Support vector machines excel at finding the optimal classification hyperplane in high-dimensional space and perform well for linear data or data that can be mapped to high-dimensional linear separability using kernel functions. Neural networks, with their powerful nonlinear fitting capabilities, can learn deep, abstract features in the data. During the fusion process, predictions from multiple models are combined using weighted voting or weighted averaging. Cross-validation-based optimization algorithms play a key role in determining weights. Specifically, the training dataset is divided into multiple subsets, with one subset used as the validation set and the remaining subsets used as the training set. The performance metrics (such as accuracy, recall, and mean squared error) of each model on the different validation sets are evaluated. Based on the evaluation results, the weights of each model during the fusion process are dynamically adjusted, giving models that perform better in the validation set a greater voice in the final evaluation. This results in more accurate and stable final evaluation results, providing a more reliable basis for gear performance evaluation.

[0029] Gear performance evaluation steps: Using classic metrics such as Euclidean distance or cosine similarity, the feature vector of the gear to be evaluated is compared one by one with feature vectors of known performance conditions in a historical database. Euclidean distance calculates the linear distance between two vectors in multidimensional space, intuitively reflecting the degree of difference between feature vectors. Cosine similarity measures the directional similarity between two vectors, focusing more on the direction of the feature vectors rather than their length. The calculation process iterates over each set of feature vectors in the historical database and calculates a similarity score with the feature vector to be evaluated. A reasonable threshold is pre-set; if the similarity score falls below this threshold, it indicates that the operating condition of the gear to be evaluated has deviated from the normal range and may be abnormal. Further in-depth analysis can be performed using methods such as fault tree analysis and expert systems, or more complex and sophisticated evaluation strategies such as deep neural network evaluation models can be employed to accurately determine the actual operating condition of the gear. The richness and reliability of the evaluation model's output are key to ensuring scientifically sound gear maintenance decisions. Utilizing Monte Carlo simulation and Bayesian inference, the evaluation model not only provides information on the gear's performance status and remaining life, but also determines confidence intervals for each performance indicator. Monte Carlo simulation uses a large number of random samples to simulate various possible variations in gear performance indicators, thereby quantifying uncertainty. Bayesian reasoning, based on prior knowledge and new observational data, continuously updates the probabilistic estimate of gear performance status. The combination of these two methods provides a more comprehensive and reliable basis for gear maintenance decisions, making maintenance plans more targeted and forward-looking, effectively reducing the risk of equipment failure, and improving the stability and cost-effectiveness of production operations.

[0030] Result Verification and Feedback Steps: The verification phase transcends the limitations of traditional single-site physical testing and incorporates virtual verification technology. Using advanced computer-aided engineering (CAE) software, a highly realistic virtual simulation model of the gear is constructed based on precise gear geometric parameters (such as module, number of teeth, pressure angle), material properties (elastic modulus, Poisson's ratio, density), and manufacturing process information. This model accurately simulates the gear's operating conditions under various operating conditions, such as speed, load, and lubrication conditions. During the simulation, an acoustic simulation module is used to capture the soundprint characteristics of the gear during operation. This data includes information such as sound pressure level, frequency content, and intensity distribution. Simultaneously, high-precision acoustic sensors are used to collect the soundprint characteristics of the gear under actual operating conditions. Signal processing and pattern recognition techniques are used to meticulously compare the soundprint characteristics obtained from the virtual simulation with the actual soundprint characteristics. A feature matching algorithm is employed to calculate the similarity between the two characteristics in multiple domains, including time, frequency, and cepstrum domains, to verify the accuracy of the evaluation results. Based on feedback control theory, an adaptive optimization mechanism for the evaluation model is constructed. The deviation between the verification result and the evaluation result is used as a feedback signal. When the deviation exceeds the pre-set range, the system automatically triggers the optimization process. In terms of parameter adjustment, intelligent optimization algorithms such as gradient descent and genetic algorithms are used to iteratively update the weight parameters, threshold parameters, etc. in the evaluation model. If it involves the optimization of the model structure, a pruning algorithm may be used to remove redundant neurons or connections, or by adding new network layers, adjusting the network topology, etc., so that the evaluation model can continuously learn and adapt to changes in gear performance over time, working conditions and other factors, thereby continuously improving the accuracy and reliability of the evaluation and providing a solid guarantee for the efficient operation and maintenance of the gears.

[0031] In the present invention, acoustic metamaterial-assisted acquisition technology is used in the voiceprint data acquisition stage. Acoustic metamaterials have special acoustic properties, such as negative refraction and acoustic focusing. When they are arranged around the acoustic sensor, they can enhance the voiceprint signal within a specific frequency range and suppress signals of other interfering frequencies. By optimizing the structure and parameters of the acoustic metamaterial, it is matched with the voiceprint characteristic frequency of the target gear, thereby improving the acquisition quality of the voiceprint signal. At the same time, combined with multi-sensor fusion technology, in addition to acoustic sensors, vibration sensors and temperature sensors, stress sensors and strain sensors are also introduced. The data of these sensors are fused and processed, and a fusion algorithm based on information entropy is used. The weight of each sensor data in the fusion process is determined according to its information entropy size to obtain more comprehensive and accurate gear operation status information. The fusion formula is , where S fused is the fused signal, S i is the signal of the i-th sensor, , H i is the information entropy of the i-th sensor data.

[0032] In the present invention, the voiceprint feature extraction step proposes a feature extraction method based on sparse representation. The voiceprint signal is represented as a linear combination of a set of overcomplete dictionaries, and the sparse representation coefficients of the signal are obtained by solving the sparse optimization problem. Features are extracted from these coefficients, such as sparsity, the distribution of non-zero coefficients, etc. Sparse representation can highlight the important features in the voiceprint signal and reduce the influence of noise and redundant information. At the same time, combined with the high-order cumulant features, in addition to the third-order moment and the fourth-order moment, higher-order cumulants are also calculated. The high-order cumulants more comprehensively describe the non-Gaussian characteristics of the voiceprint signal and improve the sensitivity to gear fault characteristics. The high-order cumulant calculation formula is , where E is the mathematical expectation and k is the order of the cumulative amount.

[0033] In this invention, the feature vector construction step utilizes a feature fusion method based on a graph neural network (GNN). The GNN-based feature fusion process first treats different types of features extracted from the gear soundprint signal, such as time-domain features (e.g., RMS value, peak value), frequency-domain features (e.g., frequency content, harmonic amplitude), and joint time-frequency features, as nodes in a graph. These nodes represent different aspects of the soundprint, and the correlations between them are represented by edges in the graph. For example, certain frequency components may be associated with specific time-domain peaks, and this correlation is reflected in the graph as an edge. By constructing such a feature graph, the GNN can conduct in-depth learning and processing. During the GNN operation, information is transferred and aggregated between nodes. Each node not only considers its own feature information but also receives information from neighboring nodes. Through a series of convolution and aggregation operations, the complex underlying relationships between features are discovered. For example, it may be found that changes in certain frequency components coincide with changes in specific time-domain features and are correlated with certain operating conditions or failure modes of the gear. After GNN processing, previously scattered features are integrated into a more representative feature vector. This feature vector more comprehensively and accurately reflects the essential characteristics of the gear soundprint signal, providing a higher-quality data foundation for subsequent performance evaluation. Furthermore, a dynamic feature update mechanism is introduced to adapt to changes in the gear's operating state. As a gear continues to operate, its soundprint characteristics change due to factors such as wear and load variations. This dynamic feature update mechanism regularly updates and optimizes the feature vector based on increased operating time and the accumulation of new data. Specifically, the system applies a specific algorithm to recalculate the importance score of each feature based on newly collected soundprint data. Features that show a higher correlation with the current gear operating state in the new data are weighted higher, while features that are no longer relevant or have a decreasing correlation are weighted lower or even removed. Furthermore, if valuable features that reflect new gear conditions or new failure modes emerge in the newly collected data, they are added to the feature vector. In this way, the feature vector accurately reflects changes in the gear's operating state in real time, ensuring that gear performance evaluation based on soundprint analysis remains highly accurate and timely, providing a more reliable basis for gear maintenance and management.

[0034] In this invention, the evaluation model building step utilizes model enhancement technology based on a generative adversarial network (GAN). A GAN consists of two core components: a generator and a discriminator. In the context of gear soundprint analysis, the generator's task is to generate simulated soundprint feature vectors and corresponding performance metrics. By learning the distribution patterns of real soundprint data and leveraging a complex neural network structure, it attempts to generate samples similar to real data. For example, it generates simulated data based on the frequency, amplitude, and phase characteristics of soundprints under different gear conditions, such as normal operation, wear, and failure. The discriminator, acting as a "data authentication expert," is responsible for distinguishing between the simulated data generated by the generator and real, collected soundprint data. By learning from large amounts of real and simulated data, it continuously improves its discrimination capabilities. During the training process, the generator and the discriminator engage in a fierce "competition." The generator strives to generate more realistic data to deceive the discriminator, while the discriminator continuously improves its discrimination capabilities and accurately identifies simulated data. As this adversarial training continues, the generator gradually becomes capable of generating highly realistic simulated data, which effectively expands the training dataset. Training the evaluation model with the expanded dataset enables it to learn a more diverse range of relationships between voiceprint features and performance indicators, significantly improving its generalization and robustness, enabling it to more accurately evaluate gear performance under diverse operating conditions and noise interference. Furthermore, a meta-learning mechanism is introduced to further optimize the evaluation model. Meta-learning aims to enable the model to learn how to learn. In gear performance evaluation, by training on multiple different tasks (i.e., different gear types or operating conditions), the model learns common initialization parameters and optimization strategies. For example, for gears of different materials and modules, as well as operating conditions with varying loads and speeds, the meta-learning process enables the model to understand the commonalities and differences between these tasks. When encountering new gear types or operating conditions, the model can quickly adjust its parameters based on the learned initialization parameters and optimization strategies, achieving rapid convergence and adapting to the new task. This allows it to accurately evaluate gear performance under these new conditions. This significantly improves the flexibility and practicality of the evaluation model, providing stronger technical support for accurate gear performance evaluation based on voiceprint analysis.

[0035] In this invention, the gear performance evaluation step utilizes a multi-dimensional assessment strategy. This approach breaks away from the traditional focus on overall performance and remaining life, delving deeper into multiple key aspects of gear operation. Regarding tooth contact strength, the distribution of contact stress between tooth surfaces during operation is considered. Sub-assessment models are established, utilizing techniques such as finite element analysis to simulate the tooth contact process, analyzing the magnitude and variation of contact stress under different operating conditions. This allows the tooth surface's ability to resist fatigue wear during long-term operation to be assessed. Regarding tooth root bending strength, a corresponding model is developed based on the principles of material mechanics, combining gear geometry and load conditions to calculate the bending stress at the tooth root. The effects of factors such as the tooth root fillet radius and stress concentration factor on bending strength are also considered to accurately assess the risk of tooth root fracture under alternating loads. Lubrication condition assessment focuses on lubricant viscosity, oil film thickness, and lubrication method. Leveraging oil analysis techniques and tribological theory, it is determined whether the lubricant effectively reduces friction and wear between tooth surfaces and maintains a favorable lubrication environment. For each sub-assessment model, a large amount of relevant data is collected for training and validation. For example, data such as tooth surface temperature and wear under different operating conditions are collected to inform the tooth contact strength model; stress and strain data at the tooth root are obtained to optimize the tooth root bending strength model; and the physical and chemical properties of the lubricant and lubrication system parameters are monitored to refine the lubrication condition assessment model. Subsequently, a comprehensive analysis algorithm is used to integrate the results of these sub-assessment models. Through weight allocation and data fusion, a more comprehensive and detailed gear performance assessment report is generated, clearly presenting the gear's performance status in all aspects. Furthermore, trend analysis and forecasting techniques are introduced. Time series analysis methods, such as the ARIMA model, are used to deeply mine historical assessment data. This historical data contains information on various gear performance indicators and operating conditions over different time periods. Through stationarity testing of the data series and parameter estimation, a suitable forecasting model is established. Combined with current assessment results, this model can accurately predict future trends in gear performance. For example, it can predict the growth rate of tooth surface wear and the decay of tooth root bending strength. This provides a reliable basis for equipment maintenance personnel to formulate scientific and reasonable maintenance plans in advance, avoiding equipment downtime due to gear failures and ensuring production continuity and stability.

[0036] In this invention, the result verification and feedback steps establish a blockchain-based verification and feedback system. Blockchain, with its unique distributed ledger technology, provides a solid guarantee for data security and reliability during the evaluation process. Each time a gear undergoes voiceprint analysis and evaluation, the voiceprint data, extracted feature vectors, final evaluation results, and actual test results are accurately recorded on the blockchain. Blockchain's immutability stems from its encryption algorithm and chain-like data structure. Each data block contains the hash value of the previous block. Once data is recorded, any attempt to modify its contents will result in a change in the hash value, which is detected by the entire blockchain network, ensuring the authenticity and reliability of the data. Furthermore, data traceability allows users to review the evaluation history at any time and understand every detail of the evaluation process. Smart contracts play a crucial role in this system. They are essentially self-executing code deployed on the blockchain. When the evaluation results are compared with the actual test results, if any discrepancies are detected, the smart contract automatically triggers optimization of the evaluation model based on pre-defined rules. For example, it may invoke relevant algorithms to adjust the evaluation model parameters or retrain the model. This process requires no human intervention, significantly improving the timeliness and accuracy of feedback, enabling the evaluation model to continuously adapt to real-world conditions and optimize performance. Furthermore, the blockchain network provides an open platform for diverse users and organizations to share and exchange evaluation data and experience. Through a consensus mechanism, each node can securely exchange information without the need to trust a third party. This sharing of data and experience breaks down information barriers and fosters industry-wide exchange and collaboration on gear performance evaluation technology. Different organizations can learn from each other's success stories and jointly explore solutions to challenges encountered during the evaluation process, driving technological advancement across the industry and elevating the overall level of accurate gear performance evaluation methods based on voiceprint analysis.

[0037] In the present invention, the voiceprint data collection step utilizes a distributed voiceprint collection network. In a gear system, given the differences in voiceprint signals generated by different gear parts during operation, multiple acoustic sensors are deployed at various locations within the gearbox, such as near the gear mesh and near the bearing seat. These sensors possess high sensitivity and a wide frequency response range, enabling them to accurately capture the various voiceprint signals generated during gear operation. The sensors are interconnected via wireless communication technology and establish a communication link with the central processing unit. For example, low-power wide area network (LPWAN) technology is employed to ensure stable and reliable data transmission in complex industrial environments. Each sensor transmits the collected voiceprint signals to the central processing unit in real time, enabling comprehensive and multi-angle collection of gear voiceprint signals. This is like establishing a 360-degree "sound monitoring network" for the gear's operating status, capturing every subtle acoustic change. Synchronous acquisition technology plays a key role in this process. To ensure temporal consistency in the data collected by each sensor, the system utilizes a high-precision clock synchronization mechanism. The clocks of all sensors are precisely calibrated using technologies such as the Global Positioning System (GPS) or the Precision Time Protocol (PTP). This ensures that the timestamp accuracy and consistency are guaranteed regardless of where the voiceprint signal is collected within the gear system. This provides a solid foundation for subsequent time-series-based voiceprint signal analysis and significantly improves the accuracy of the analysis results. During data transmission, great attention is paid to data security and privacy. Advanced encryption algorithms, such as the Advanced Encryption Standard (AES) or Elliptic Curve Cryptography (ECC), are used to encrypt the collected voiceprint data. Data is converted into ciphertext before transmission and can only be restored to its original form after decryption by the central processing unit. This encryption effectively prevents data theft or tampering during transmission, ensuring the security and privacy of voiceprint data and providing a strong foundation for the reliable application of accurate gear performance assessment methods based on voiceprint analysis.

[0038] In this invention, the voiceprint feature extraction step utilizes an adaptive spectral clustering feature classification method. Numerous features, such as frequency and amplitude, are extracted from the voiceprint signals generated by gear operation. The adaptive spectral clustering algorithm acts as an intelligent "classifier," breaking the limitations of traditional clustering algorithms, which require pre-setting the number of clusters and centers. The algorithm constructs a similarity graph based on similarity metrics between these voiceprint feature data points. In this graph, the connection weights between data points reflect their degree of similarity. By analyzing the graph's Laplacian matrix and utilizing its eigenvalues and eigenvectors, the optimal number of clusters and cluster centers are automatically determined. During the clustering process, the clustering strategy is dynamically adjusted based on the actual distribution characteristics of the voiceprint features. For example, the distribution of voiceprint features changes as gears undergo different wear stages or load conditions. The adaptive spectral clustering algorithm can sensitively capture these changes and accurately classify the voiceprint features into different categories. Each category corresponds to a specific gear operating state or fault mode, such as normal operation, mild wear, or severe wear. By deeply analyzing the commonalities and differences in features within each category, the current gear operating state and fault type can be more accurately identified. Furthermore, a fuzzy feature extraction method is incorporated, introducing a fuzzy membership function. Voiceprint features inherently exhibit a degree of uncertainty and ambiguity. For example, during the transition from a normal to a faulty gear state, the voiceprint features do not exhibit clear boundaries. The fuzzy membership function maps voiceprint feature values to different fuzzy sets. A given voiceprint feature value may belong to both the "normal" and "minor wear" fuzzy sets with a certain degree of membership. The fuzzy feature vector derived in this way provides a more flexible and comprehensive description of the voiceprint feature's characteristics, effectively improving the ability to identify complex gear operating conditions and enabling more accurate judgments even when feature boundaries are unclear. The evaluation model is trained using a federated learning approach. In a distributed data environment, each participant can train a local model locally. The federated learning algorithm securely and efficiently aggregates the parameters of each local model to form a global evaluation model. To further enhance the model's practicality and deployment efficiency, model compression technology is also introduced. For example, methods such as pruning and quantization can be used to remove redundant parameters and connections in the model, thereby reducing the model's storage space and computational overhead without significantly affecting the model's performance, allowing the evaluation model to be more quickly and conveniently applied to actual gear performance evaluation scenarios.

[0039] In the present invention, when the equipment is faced with the situation of changing the gear supplier, voiceprint analysis plays a key role in determining whether the new gear can perfectly replace the original gear. After completing the traditional quality inspections such as hardness and size, the equipment operation voiceprint data after the new gear is installed is first collected. With the help of high-precision acoustic sensors, it is ensured that the most subtle sound changes during equipment operation can be captured. Using advanced signal processing technology, analysis is performed from multiple dimensions such as time domain, frequency domain, time-frequency domain and phase space reconstruction. In the time domain analysis, attention is paid to the amplitude, period and other time-varying characteristics of the voiceprint signal; the frequency domain analysis focuses on the energy distribution of the signal at different frequencies; the time-frequency domain analysis combines the time and frequency dimensions, and can more comprehensively reveal the frequency characteristics of the signal at different times; the phase space reconstruction processes the voiceprint signal to explore its deep dynamic characteristics. Through these analyses, a feature vector containing multi-dimensional information is obtained. The cosine distance formula is used Calculate the new gear eigenvector Compared with the original gear characteristic vector This formula is based on the vector dot product and vector modulus, and scientifically measures the angular relationship between two eigenvectors in space. The smaller the angle, the higher the similarity. If the calculated similarity is higher than the set threshold (such as 0.9), it is necessary to further evaluate the performance status, remaining life and other indicators of the new gear with the help of the evaluation model. Only when the difference between these indicators and the original gear is less than 5% can it be determined that the new gear can effectively replace the original gear, providing strong support for the supplier switching decision. The entire process uses multi-dimensional technical means to ensure the scientific nature and accuracy of the gear replacement decision.

[0040] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for accurately evaluating gear performance based on voiceprint analysis, characterized in that: The following steps are involved: Voiceprint data collection steps: An intelligent sensor layout algorithm is used to determine the optimal position of the acoustic sensor for different gears. An adaptive sampling strategy is introduced into the collection process to adjust the sampling frequency and duration in real time according to the gear operating status; Voiceprint data preprocessing steps: Use convolutional neural networks (CNN) combined with recurrent neural networks (RNN) for denoising. The network is trained on noisy and noise-free data to adaptively identify and remove noise. An improved normalization method is proposed, using a quantile normalization formula to process voiceprint data. Voiceprint feature extraction steps: Extract time domain, frequency domain, time-frequency domain, and phase space reconstruction features, reconstruct the voiceprint time series through the delay embedding method, extract features to capture early fault information, introduce multi-scale entropy features, and decompose the voiceprint signal at multiple scales to calculate the entropy value; Steps for constructing feature vectors: adopt a feature importance ranking method, calculate the correlation between features and gear performance indicators, determine the score based on mutual information, select high-scoring features to form feature vectors, and remove redundant and irrelevant features; Steps to establish the evaluation model: Build an evaluation model based on deep forest, combining decision trees and deep learning ideas, use an adaptive forest growth strategy to train the model, introduce a model fusion mechanism to integrate deep forest and traditional machine learning models, and use a cross-validation optimization algorithm to dynamically adjust weights; Gear performance evaluation steps: Before inputting the feature vector into the model, the feature vector is compared with the historical feature vector to calculate the similarity. The evaluation model outputs the gear performance status, remaining life and confidence interval. The uncertainty of the evaluation results is quantified through Monte Carlo simulation and Bayesian inference methods. Result verification and feedback steps: Introduce virtual verification technology, compare simulated and actual voiceprint features, establish an adaptive optimization mechanism based on feedback control theory, and adjust the evaluation model parameters and structure according to the deviation.

2. The gear performance accurate evaluation method based on voiceprint analysis according to claim 1 is characterized in that: The voiceprint data collection step adopts acoustic metamaterial-assisted collection technology. By optimizing the acoustic metamaterial structure and parameters to match the target gear voiceprint characteristic frequency, the data is fused and processed by combining the sound, vibration, temperature and stress sensor fusion technology. The information entropy fusion algorithm is used to determine the weight according to the information entropy of the sensor data. The fusion formula is: , S fused is the fused signal, S i is the ith sensor signal, , H i is the information entropy of the i-th sensor data.

3. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1 is characterized in that: The voiceprint feature extraction step adopts a feature extraction method based on sparse representation. By solving the sparse optimization problem, the signal sparse representation coefficient is obtained. The higher-order cumulant feature is combined to calculate the higher-order cumulant. The high-order cumulant calculation formula is: , E is the mathematical expectation, and k is the order of the cumulant.

4. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1 is characterized in that: The step of constructing feature vectors adopts the feature fusion method based on graph neural network (GNN), uses GNN learning to process feature graphs, aggregates node information, explores potential relationships between features, introduces a dynamic feature update mechanism, regularly updates and optimizes feature vectors, calculates feature importance scores based on collected data, adds valuable features, and removes irrelevant features.

5. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1 is characterized in that: The steps of establishing the evaluation model adopt the generative adversarial network (GAN) model enhancement technology. GAN consists of a generator and a discriminator. The generator generates simulated voiceprint feature vectors and performance indicators, and the discriminator distinguishes between real data and generated data. The training data set is expanded by training the generator and discriminator, and the evaluation model is trained using the data set. The meta-learning mechanism is introduced to learn the optimization strategy through training.

6. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1 is characterized in that: The gear performance evaluation step adopts a multi-dimensional evaluation strategy to evaluate the overall performance status, remaining life, tooth surface contact strength, tooth root bending strength, and lubrication status of the gear. A corresponding sub-evaluation model is established for each dimension, and the evaluation results of the sub-evaluation model are comprehensively analyzed. Trend analysis and prediction technology are introduced. Based on historical evaluation data and current evaluation results, a time series analysis method is used to predict the trend of gear performance changes.

7. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1 is characterized in that: The result verification and feedback steps adopt a blockchain-based verification and feedback system, recording voiceprint data, feature vectors, evaluation results and actual test results on the blockchain, and using the blockchain smart contract function to verify and feedback the evaluation results. If there is a deviation between the evaluation results and the actual test results, the smart contract automatically triggers the evaluation model optimization process, and shares and exchanges evaluation data and experience through the blockchain network.

8. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1 is characterized in that: The sampling frequency adjustment formula in the voiceprint data collection step is: , f s (t) is the sampling frequency at time t, f s0 is the initial sampling frequency, k is the adjustment coefficient, Δv is the change in speed or load, and v0 is the initial speed or load; A distributed voiceprint collection network is adopted, and multiple acoustic sensors are arranged in different parts of the gear system to form a distributed collection network. The data collected by each sensor is transmitted to the central processing unit through wireless communication technology. Synchronous collection technology is used to collect data simultaneously, and encryption algorithms are used to encrypt data during the data transmission process.

9. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1 is characterized in that: The voiceprint feature extraction step adopts the adaptive spectral clustering feature classification method, divides the data points through the adaptive spectral clustering algorithm, determines the number of clusters and cluster centers, and dynamically adjusts according to the feature distribution. The gear operating status and fault type are identified by analyzing the category features. Combined with the fuzzy feature extraction method, the fuzzy membership function is introduced to map the voiceprint feature values to different fuzzy sets to obtain the fuzzy feature vector; the evaluation model establishment step adopts the federated learning method to train the model, aggregates the local model parameters through the federated learning algorithm to obtain the global evaluation model, and introduces the model compression technology to compress the evaluation model.

10. The method for accurately evaluating gear performance based on voiceprint analysis according to claim 1, characterized in that: When the equipment changes the gear supplier, the voiceprint data of the new gear after installation is collected through voiceprint analysis, and the feature vectors of time domain, frequency domain, time-frequency domain and phase space reconstruction are extracted, and the similarity is matched with the original gear feature vector in the historical database. The similarity calculation uses the cosine distance formula , is the new gear eigenvector, is the original gear eigenvector; If the similarity is higher than the set threshold and the performance status and remaining life indicators output by the evaluation model are not significantly different from those of the original gear, the new gear is determined to replace the original gear, supporting the supplier switching decision.

Images