Mechanical transmission system fault trend prediction system based on dynamic feature recognition

Abstract

The application discloses a mechanical transmission system fault trend prediction system based on dynamic feature recognition and relates to the technical field of mechanical state monitoring. The system comprises a signal acquisition module which synchronously acquires a load torque signal, a lubrication state parameter signal and vibration acceleration signals of multiple measuring points of a transmission system; a working condition decoupling feature generation module which performs time-frequency analysis on the vibration signals, and according to the load torque signal, calls a pre-stored load disturbance spectrum template to perform adaptive difference processing to eliminate load fluctuation interference, and according to the lubrication state parameter signal, calls a correction rule set to perform morphological reorganization on the signals to compensate for the influence of the lubrication state, and finally outputs working condition decoupling features representing the health state of mechanical components; and a trend prediction module which calculates fault development trends and residual life estimation data through a pre-trained fault prediction model. The application effectively extracts dynamic features representing the essential degradation of components, and improves the accuracy and reliability of mechanical transmission system fault trend prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical state monitoring, in particular to a mechanical transmission system fault trend prediction system based on dynamic feature recognition. BACKGROUND

[0002] Mechanical transmission system fault prediction and health management is a core technology in the field of industrial intelligent operation and maintenance, aiming to identify fault signs and predict remaining life in advance through real-time monitoring and data analysis of equipment operating state, so as to realize predictive maintenance and ensure production safety and efficiency. In recent years, with the progress of vibration analysis technology and artificial intelligence algorithm, fault prediction methods are developing towards high precision and self-adaptation.

[0003] In the prior art, most schemes collect the vibration signals of the transmission system, and combine basic working condition parameters such as load and speed to perform feature extraction and trend analysis. These methods can realize a certain degree of early warning for common faults under stable working conditions by establishing a statistical relationship between vibration features and equipment degradation, and have considerable practical value. However, in actual operation, the dynamic response of the mechanical transmission system is a complex product of the mutual coupling of load torque, lubrication state and other working condition factors and component health status. The existing methods often regard load fluctuation as the main interference for simple filtering or working condition segmentation processing, but generally ignore the nonlinear modulation effect of dynamic changes in lubrication state on the form of vibration features. For example, poor lubrication can change the contact stiffness and damping characteristics of gears and bearings, causing changes in the energy distribution of the vibration frequency spectrum, and this change is highly overlapped with the fault features caused by component wear in the frequency spectrum.

[0004] Therefore, in the existing technology, in the industrial scene where the lubrication condition fluctuates greatly, the prediction model is prone to false positives and false negatives, and its generalization ability and prediction accuracy face a bottleneck. SUMMARY

[0005] In view of the deficiencies in the prior art, the present application provides a mechanical transmission system fault trend prediction system based on dynamic feature recognition.

[0006] In order to achieve the above-mentioned purpose, the technical scheme of the present application is as follows:

[0007] The present application discloses a mechanical transmission system fault trend prediction system based on dynamic feature recognition, comprising:

[0008] A signal acquisition module is used to synchronously acquire the load torque signal and the lubrication state parameter signal of the mechanical transmission system and the vibration acceleration signals of multiple measuring points on the transmission chain;

[0009] A working condition decoupling feature generation module is used to perform the following operations:

[0010] generate time-frequency distribution data based on the vibration acceleration signal;

[0011] According to the load torque signal, a corresponding load disturbance spectrum template is called from a pre-stored load disturbance spectrum mapping relationship, and adaptive differential processing is performed on the time-frequency distribution data to generate processed time-frequency distribution data.

[0012] According to the lubrication state parameter signal, a corresponding correction rule set is called to perform morphological reorganization on the processed time-frequency distribution data, and working condition decoupling features representing the health state of the mechanical component are output.

[0013] A trend prediction module is configured to calculate the fault development trend data and the remaining life estimation data of the mechanical transmission system based on the working condition decoupling features through a pre-trained fault prediction model.

[0014] Compared with the prior art, the present application has the following advantages:

[0015] 1. The present application can selectively separate the linear disturbance component related to the load fluctuation from the complex vibration response. This processing makes the time-frequency data relied on subsequent analysis no longer change dramatically in energy characteristics with the load fluctuation, laying a foundation for extracting stable fault features under variable load conditions.

[0016] 2. The present application can compensate or suppress the nonlinear modulation effect of the lubrication condition change on the vibration signal spectrum form, so that the final output working condition decoupling features can more purely reflect the essential state change of the mechanical component caused by wear, fatigue and other degradation processes, effectively avoiding the feature confusion and misjudgment caused by lubrication fluctuation.

[0017] 3. The present application can more accurately track the fault development trend of the mechanical transmission system and provide reliable remaining life estimation under the complex industrial scene of changing load and lubrication state, improving the accuracy and practicality of the predictive maintenance strategy. BRIEF DESCRIPTION OF DRAWINGS

[0018] The disclosure of the present application will be described with reference to the accompanying drawings. It should be understood that the drawings are only for illustrative purposes, and are not intended to limit the scope of protection of the present application. In the drawings, the same reference numerals are used to refer to the same parts. Among them:

[0019] Figure 1 is a system module connection diagram of the present application;

[0020] Figure 2 is a system module flowchart of the present application;

[0021] Figure 3 is a construction process flowchart of the lubrication feature form mapping relationship of the present application;

[0022] Figure 4 A step flow chart for the cycle drift identification module of the present application;

[0023] Figure 5 A step flow chart for the conduction delay difference extraction module of the present application. DETAILED DESCRIPTION

[0024] It is easy to understand that, according to the technical solution of the present application, a person skilled in the art can propose a plurality of structure modes and implementation modes which can be replaced with each other without changing the essential spirit of the present application. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present application, and should not be regarded as the whole or as the limitation or restriction of the technical solution of the present application.

[0025] In the prior art, most solutions can combine basic working condition parameters such as load, but generally lack a fine compensation mechanism for dynamic changes in lubrication state. When the load fluctuates or the lubrication state changes, the vibration frequency spectrum characteristics of the traditional method are easily disturbed, resulting in inaccurate fault feature extraction. The existing solution lacks a fine compensation mechanism for dynamic changes in working conditions, especially under variable load and variable lubrication conditions. A single signal processing model will have systematic deviations, making it difficult to meet the demand for high-precision predictive maintenance.

[0026] In order to solve the above problems, it is found that there is a stable mapping relationship between the load torque and the disturbance mode of the vibration frequency spectrum, and the lubrication state has a nonlinear modulation effect on the frequency spectrum form. By establishing a load disturbance spectrum and a lubrication correction rule set, the working condition interference can be peeled off layer by layer. It is further found that adaptive difference based on the load interval can effectively suppress linear disturbances, and the form reorganization of the lubrication state can compensate for nonlinear distortion. Therefore, the idea of generating working condition decoupling features through two-level processing of "difference-reorganization" is proposed. Further, through online monitoring and model updating, the load disturbance spectrum mapping relationship and the lubrication feature form mapping relationship are incorporated into a dynamic optimization mechanism to form a continuously adaptive feature extraction system.

[0027] After introducing the basic concept of the present application, the embodiments of the present application will be specifically introduced with reference to the drawings.

[0028] Embodiment:

[0029] As shown in Figure 1 The mechanical transmission system fault trend prediction system based on dynamic feature identification comprises:

[0030] A signal acquisition module is configured to synchronously acquire a load torque signal and a lubrication state parameter signal of the mechanical transmission system, and vibration acceleration signals of a plurality of measuring points on the transmission chain.

[0031] The working condition decoupling feature generation module is configured to perform the following operations:

[0032] Generate time-frequency distribution data based on the vibration acceleration signal;

[0033] According to the load torque signal, a corresponding load disturbance spectrum template is called from a pre-stored load disturbance spectrum mapping relationship, and adaptive differential processing is performed on the time-frequency distribution data to generate processed time-frequency distribution data;

[0034] According to the lubrication state parameter signal, a corresponding correction rule set is called to perform morphological reorganization on the processed time-frequency distribution data, and working condition decoupling features representing the health state of the mechanical component are output.

[0035] The trend prediction module is configured to calculate the fault development trend data and the remaining life estimation data of the mechanical transmission system based on the working condition decoupling features through a pre-trained fault prediction model.

[0036] The training of the fault prediction model is completed through historical full life cycle data. Complete data sequences from healthy state operation to complete failure are collected from no less than 50 mechanical transmission systems of the same type. Each data sequence includes working condition decoupling features, auxiliary feature vectors, and corresponding remaining life labels. Before training, the feature data is standardized to eliminate the dimension effect. The model uses a deep fully connected network structure, including an input layer, three hidden layers, and an output layer. The number of hidden layer neurons is 256, 128, and 64, respectively, and the ReLU activation function is used. During training, the Adam optimizer is used, the initial learning rate is set to 0.001, the batch size is 32, and the number of training rounds is 200. The loss function uses a weighted combination of mean square error and cross-entropy loss, corresponding to the remaining life regression and fault classification tasks. Overfitting is prevented by early stopping, and training is terminated when the validation set loss does not decrease for 10 consecutive rounds. The final model can output fault development trend scores and remaining life estimates.

[0037] As shown in Figure 2 The working principle of the present application is as follows: The signal acquisition module is composed of a group of physical sensors arranged at key nodes of the mechanical transmission chain. Specifically, it includes torque sensors installed on the input / output shafts for real-time measurement of load torque signals; oil condition sensors integrated in the gearboxes or bearing seats for detecting the dielectric constant or viscosity parameters of the lubricating oil; and three-axis ICP type acceleration sensors distributed on the bearing seats along the transmission path for collecting vibration acceleration signals. All sensors are time-synchronized to the microsecond level through a synchronous data acquisition card to ensure the timing alignment of multi-source signals. The collected raw signals are transmitted to the edge computing node via industrial Ethernet for subsequent processing.

[0038] The working condition decoupling feature generation module is deployed in an industrial computer with parallel computing capability, and its data processing flow is as follows:

[0039] 1. Time-frequency distribution generation: the short-time Fourier transform algorithm is used for time-frequency analysis of the vibration acceleration signal to generate time-frequency distribution data containing time-frequency-energy three-dimensional information. This process intercepts the signal segment through a sliding window, performs FFT transformation after adding a Hanning window, and finally forms a time-frequency spectrum matrix.

[0040] 2. Adaptive differential processing: the system calls the reference template corresponding to the torque interval from the pre-stored load disturbance spectrum database according to the real-time load torque signal. The load disturbance spectrum database is stored in the form of a lookup table and interacts with the processing unit at high speed through the PCIe interface. The differential processing uses weighted spectral subtraction, dynamically calculates the attenuation factor according to the energy ratio of the current spectrum and the template spectrum in the feature band, and subtracts the scaled template component from the original time-frequency distribution to effectively suppress the spectral disturbance caused by load fluctuations.

[0041] 3. Morphology reorganization processing: in view of the influence of lubrication state, the system calls the correction rule set established based on the lubrication state parameter signal-spectrum response relationship model. The correction rule set is stored in the form of a configuration file and contains frequency band gain coefficients corresponding to different lubrication states. The processing unit determines the specific gain vector through the interpolation algorithm according to the real-time lubrication state parameter signal, and performs element-by-element multiplication operation with the time-frequency data after differential processing to realize energy compensation or suppression of the frequency band affected by lubrication, and finally outputs the working condition decoupling feature that purely represents the health state of the component.

[0042] The trend prediction module runs on a server platform, and its core is a fault prediction model based on a deep learning framework. After receiving the working condition decoupling feature, the trend prediction module first performs feature standardization processing to eliminate the dimension influence. Then the processed feature vector and the auxiliary feature vector obtained from other analysis paths are fused at the feature level to form a comprehensive feature description. The fused features are input into a pre-trained deep neural network model, which has fault pattern recognition ability through historical fault data training, and finally outputs fault development trend data representing the health degree change trend of the system and remaining life estimation data in time units in parallel.

[0043] Through the above technical solutions, the application combines the technical means of multi-source sensor data synchronous acquisition, working condition interference step-by-step decoupling and intelligent prediction model, so that the system can effectively extract feature information strongly related to the essential degradation of mechanical components in a variable working condition operation environment, thereby providing accurate state evaluation and life prediction basis for predictive maintenance of mechanical transmission systems.

[0044] The application further proposes that in the step of calling the corresponding load disturbance spectrum template from the pre-stored load disturbance spectrum mapping relationship, the construction and updating process of the load disturbance spectrum mapping relationship comprises:

[0045] In the application, the construction and updating of the load disturbance spectrum mapping relationship is realized by establishing a benchmark reference in a healthy state of the system and continuously optimizing in the running process. In specific implementation, after the mechanical transmission system is completed installation and debugging, before it is put into formal operation, the mechanical transmission system is determined to be in a stable running phase of a healthy state, the system collects vibration acceleration signals under different load torque signals through torque sensors and vibration acceleration sensors in the signal acquisition module, including setting sampling points at intervals of 5%-10% rated load in the range of 20%-100% rated load, and continuously collecting at least 300 seconds of stable running data at each load point. During the collection process, it is necessary to ensure that the lubrication system is working normally and the parameters are within the rated range, which serves as the basis for determining that the equipment is in a healthy state.

[0046] After obtaining the basic data, the data processing unit performs time-frequency analysis on the vibration acceleration signals of each load interval. The time-frequency distribution data is converted from the time domain signal by using the same short-time Fourier transform parameters (such as window length of 1024 points and overlap rate of 75%) as in the real-time processing stage. The time-frequency distribution data of the vibration acceleration signal is extracted, the arithmetic mean value of all time-frequency distribution data in each load interval is calculated along the time dimension, and is stored as a load disturbance spectrum template. The load disturbance spectrum template is a benchmark spectrum established in a healthy state, and is stored in the memory in the form of an array.

[0047] The load disturbance spectrum template The calculation formula is:

[0048]

[0049] Wherein, f is the frequency variable, corresponding to the frequency resolution of the time-frequency distribution data;

[0050] is the number of sampling frames in the load interval;

[0051] represents the time-frequency distribution data of the i-th frame.

[0052] The average time-frequency distribution data corresponding to each load interval calculated is stored in the non-volatile memory in the form of a key-value pair of

[0053] In the online monitoring process, in order to maintain the accuracy of the mapping relationship, the system introduces a dynamic updating mechanism. When the system is in a quasi-steady state condition with a load fluctuation rate of less than 15% and a total vibration energy fluctuation rate of less than 20%, it is determined that the current health state data can be used for updating. The health state data satisfies:

[0054] 1. Load stability: the fluctuation rate of the load torque signal in the current time window (e.g. 60 seconds) is less than the preset threshold (e.g. 15%);

[0055] 2. Vibration stability: the total energy (or characteristic frequency band energy) fluctuation rate of the vibration acceleration signal in the same time window is less than the preset threshold (e.g. 20%);

[0056] 3. Normal lubrication state: the lubrication state parameters (such as viscosity, dielectric constant) are within the normal working range specified by the equipment manufacturer;

[0057] 4. No alarm history: the system has not triggered any fault alarms in the current time window and a certain period of time (e.g. 10 minutes) in the past.

[0058] When and only when the data collected at M (e.g. M=5) time points are all determined to be health state data, the load disturbance spectrum mapping relationship is dynamically updated by a dynamic updating algorithm. The dynamic updating algorithm uses an exponential weighted moving average algorithm, and the specific formula is:

[0059]

[0060] wherein, is the updated load disturbance spectrum template;

[0061] is the original load disturbance spectrum template;

[0062] is the average frequency spectrum in the current period of time;

[0063] γ is the forgetting factor, which is usually taken as a value between 0.85 and 0.95 according to the stability requirements of the application scenario.

[0064] The exponential weighted moving average algorithm is executed by a real-time calculation engine on the embedded system, ensuring that the mapping relationship can gradually adapt to the slow changes of the equipment over time, while avoiding distortion of the template caused by single abnormal data.

[0065] Through the aforementioned construction and updating methods, this application enables the load disturbance spectrum mapping relationship to provide an accurate benchmark reference during the initial stage of system commissioning, while also adapting to the slow changes in equipment status during long-term operation. This dynamic maintenance mechanism ensures that adaptive differential processing can always effectively separate load disturbances from fault characteristics, thereby maintaining the accuracy of fault prediction throughout the entire system lifecycle.

[0066] This application further proposes that, based on the load torque signal, the specific steps for calling the corresponding load disturbance spectrum template from the pre-stored load disturbance spectrum mapping relationship and performing adaptive differential processing on the time-frequency distribution data include:

[0067] The system reads the load torque signal from the torque sensor in real time, maps the load torque signal to multiple preset load ranges, and calls the load disturbance spectrum template corresponding to each load range from the load disturbance spectrum mapping relationship. ,in The frequency is indicated. These load ranges are pre-defined based on the actual operating characteristics of the transmission system. For example, 0%-30% of the rated load can be classified as a light load range, 30%-70% as a medium load range, and 70%-100% as a heavy load range. The specific range boundary values ​​can be adjusted according to the system's rated parameters. The mapping process uses a nearest neighbor matching algorithm to assign the current load torque value to the nearest load range.

[0068] Subsequently, the system enters the core adaptive scaling factor calculation phase. This phase requires calculating two key energy parameters: one is the current time-frequency distribution data. In the characteristic frequency band Total energy within Secondly, the template energy of the load perturbation spectrum template P(f) within the same characteristic frequency band. .

[0069] The selection of the characteristic frequency band is determined based on the equipment characteristics. For gear transmission systems, the band containing the meshing frequency and its harmonics is typically selected, while for bearings, the band near the fault characteristic frequency is selected. Energy calculation employs a numerical integration method, specifically:

[0070] Total Energy Calculated by numerical integration:

[0071]

[0072] Template energy The same calculation method is used:

[0073]

[0074] in, , respectively, are the upper and lower limit frequencies of the characteristic frequency band.

[0075] Based on the above energy value, the system calculates an adaptive scaling factor , whose calculation formula is:

[0076]

[0077] wherein, is a preset gain parameter, whose value range is usually 0.5-1.2, and the specific value of the coefficient is determined through experimental calibration. Under the system health state, a series of step load changes covering the entire working range are applied, and the energy fluctuation rates of the working condition decoupling features processed by different values are calculated respectively. The value that minimizes the fluctuation rate is selected as the final calibration value. The purpose of this process is to adjust the strength of the difference processing, avoiding overcompensation (overly large leading to feature distortion) or insufficient compensation (overly small leading to residual load disturbance).

[0078] After completing the scaling factor calculation, the system performs adaptive difference operation on the time-frequency distribution data to generate processed time-frequency distribution data , whose calculation formula is:

[0079]

[0080] The adaptive difference operation is performed in parallel at each frequency point in the time-frequency domain, and the calculation process is accelerated through the vector operation unit in the embedded processor. For the problem of negative values that may appear in the processing result, the system adopts zeroing processing, i.e., all negative values are set to zero, to ensure the physical reasonableness of the time-frequency distribution data.

[0081] Through the above adaptive difference processing method, the present application can dynamically adjust the load disturbance compensation strength according to the current vibration energy level, avoiding the inadaptation problem of fixed coefficient compensation under variable working conditions. This processing method not only retains the reference provided by the load disturbance spectrum template, but also realizes accurate control of the compensation amount through real-time energy comparison, so that the vibration features related to component degradation can be effectively separated under varying load conditions.

[0082] As shown in Figure 3 , the present application further proposes to call the corresponding correction rule set according to the lubrication state parameter signal, and the correction rule set includes a lubrication feature morphology mapping relationship, aiming to quantify the nonlinear modulation effect of the lubrication state on the vibration spectrum morphology, and provide quantitative correction basis for subsequent morphology reorganization processing. The mapping relationship is constructed by collecting the corresponding data of lubrication state and vibration response in the actual running process and based on statistical learning method. ​​​

[0083] The establishment process of the lubrication feature pattern mapping relationship includes:

[0084] In the mapping relationship establishment stage, the system continuously monitors and collects lubrication state parameter signals and synchronous vibration acceleration signals during the operation of the mechanical transmission system; the lubrication state parameters are obtained through online oil sensors, including but not limited to oil viscosity (unit: cSt), oil dielectric constant (dimensionless), and oil temperature (unit: °C).

[0085] To ensure that the collected samples can accurately reflect the essential correlation between the lubrication state and the vibration features, the system sets strict sample screening conditions: when the fluctuation rate of the load torque signal in the preset third time window is lower than the first preset threshold and the fluctuation rate of the total energy of the vibration acceleration signal in the same time window is lower than the second preset threshold, it is determined that the device is in a stable and healthy operating state, and the data collected at this time can be used as valid samples.

[0086] The first preset threshold and the second preset threshold are empirical values determined based on the inherent fluctuation characteristics of the transmission system under steady-state conditions. The first preset threshold ranges from 10% to 20%, and preferably, the first preset threshold is 15%. The second preset threshold ranges from 10% to 30%, and preferably, the second preset threshold is 15%. The preset third time window is usually set to 10-60 seconds to cover several complete device working cycles. Through this screening mechanism, abnormal data caused by load mutations or external shocks are effectively excluded, ensuring the quality of the training samples.

[0087] For each valid sample point, the system records the current multi-dimensional lubrication state parameter vector L and simultaneously extracts the energy distribution vector E of the time-frequency distribution data generated based on the synchronous vibration acceleration signal in different frequency bands. As a sample pair, it is recorded and stored. The frequency band division is determined according to the characteristic frequency of the transmission system, usually containing 1-3 times the frequency of the meshing frequency and its sideband range, thereby forming a feature vector describing the spectral pattern. The lubrication state parameter vector L and the energy distribution vector E together constitute a sample pair (L, E) and are stored in the training sample database.

[0088] Using the accumulated sample pairs, a quantitative prediction model from lubrication state to spectral energy distribution is established through a multivariate linear regression algorithm. This quantitative prediction model represents the expected change in spectral pattern caused by changes in lubrication conditions under ideal healthy conditions. The mathematical expression of the quantitative prediction model is:

[0089]

[0090] wherein, for the predicted energy distribution vector, W is a regression coefficient matrix, and b is a bias vector. The model parameters of the quantitative prediction model are solved by minimizing the mean square error between the predicted value and the actual observation value:

[0091]

[0092] where N is the number of training samples, and respectively represent the energy distribution vector and the lubrication state parameter vector of the i-th sample.

[0093] The parameters (W, b) of the quantitative prediction model are stored and updated as the lubrication feature morphology mapping relationship; wherein the correction rule set is the gain coefficient predicted by inputting the current lubrication state parameter signal into the quantitative prediction model .

[0094] In the correction rule set actually applied, the current real-time collected lubrication state parameter signal is input into the quantitative prediction model, the predicted energy distribution is calculated, and then the gain coefficient for each frequency point is derived through comparison with the original reference energy distribution, which is used for compensation and correction of the vibration frequency spectrum morphology.

[0095] Through the above technical solution, the application can quantitatively evaluate the modulation effect of lubrication state change on the vibration frequency spectrum, effectively reduce the interference of lubrication condition fluctuation on fault feature extraction, and improve the representation ability of working condition decoupling features to the real health state of the component, thereby providing a technical basis for realizing accurate fault trend prediction under variable lubrication conditions.

[0096] The application further proposes that the corresponding correction rule set is called according to the lubrication state parameter signal, and the morphology of the processed time-frequency distribution data is reorganized, which is a key link for compensating the frequency spectrum distortion caused by the lubrication state after completing the load disturbance suppression. This processing is based on the pre-established lubrication feature morphology mapping relationship, and realizes fine correction of the vibration frequency spectrum through frequency domain gain adjustment. The specific steps include:

[0097] The system first classifies the real-time collected lubrication state parameter signal. These parameters include oil viscosity, dielectric constant and oil temperature. The lubrication state parameter signal is compared with the boundary values of the preset multiple lubrication state intervals to determine the current lubrication state category. The lubrication state interval is divided according to the technical specifications of the equipment manufacturer and practical experience, for example, it can be divided into three categories of “excellent lubrication”, “normal lubrication” and “poor lubrication”, each category corresponds to a group of explicit parameter threshold values.

[0098] In a specific implementation, taking the viscosity of the oil as an example, when the viscosity value is higher than 80% of the standard value of new oil, it is "good lubrication", between 60%-80% is "normal lubrication", and lower than 60% is "poor lubrication". The system adopts a multi-parameter joint criterion, and when the state categories indicated by each parameter are inconsistent, the final state category is determined according to the worst principle.

[0099] According to the lubrication state category, the gain coefficient for different frequency bands is called from the stored lubrication feature morphology mapping relationship . The specific value of the gain coefficient reflects the compensation degree required by each frequency band under different lubrication states. In the frequency band that is significantly affected by lubrication (usually the low frequency band of 200-800 Hz), the gain value may be greater than 1, while in the relatively stable high frequency band, it is close to 1.

[0100] After the gain coefficient is called, the system performs a spectrum morphology adjustment operation on the time-frequency distribution data after load difference processing , multiplies the frequency points according to the gain coefficient , and outputs the working condition decoupling feature

[0101] ;

[0102] Among them, represents the Hadamard product, which is used to compensate or suppress the energy of the frequency band that is significantly affected by the lubrication state.

[0103] This operation is performed independently in the frequency dimension, and the corresponding gain adjustment is applied to the amplitude at each frequency point f, so as to realize the accurate remodeling of the spectrum morphology.

[0104] Through the above morphology reorganization processing, the present application can effectively compensate for the spectrum distortion caused by the change of the lubrication state, so that the working condition decoupling feature obtained finally no longer contains the feature variation caused by the lubrication condition. This targeted spectrum correction combined with the load disturbance suppression described above forms a complete working condition decoupling processing chain, which provides a feature input that can purely reflect the health state of the mechanical parts for subsequent fault trend prediction, thereby maintaining the accuracy of the fault prediction model under varying lubrication conditions.

[0105] The present application further proposes that the working condition decoupling feature generation module is further used for:

[0106] extracting low-energy residual components from the time-frequency distribution data after morphology reorganization, further mining subtle features with low energy but fault warning value, and enhancing the feature representation ability through multi-measurement point information fusion.

[0107] In the implementation process, the system performs the process of low-energy residual component extraction by setting a dynamic energy threshold, which is 5%-10% of the global energy of the current time-frequency matrix. This proportion is determined based on the energy level statistics of early fault features in a large number of fault cases. The extraction method uses soft threshold filtering technology to retain weak components below the threshold but above the noise reference, forming independent low-energy residual time-frequency distributions for each measuring point where i represents the measuring point number.

[0108] After obtaining the low-energy residual components of each measuring point, the low-energy residual components extracted from different measuring points on the transmission chain are time-aligned and energy-normalized. Time alignment is based on the physical layout of the transmission chain and the signal propagation speed. By calculating the fixed time delay between measuring points and compensating, it ensures that the signals of the same vibration event at different measuring points are synchronized in time. Energy normalization uses the min-max normalization method to map the residual component energy of each measuring point to a unified range of [0, 1], eliminating the inconsistency of the basic energy level caused by differences in sensor sensitivity or installation location.

[0109] When generating the focused residual feature vector, the system introduces an adaptive weight allocation mechanism based on historical fault sensitivity. Based on the historical fault sensitivity of each measuring point, an adaptive weight is assigned to the corresponding low-energy residual component and weighted superimposed to generate the focused residual feature vector.

[0110] The weight of each measuring point is calculated through the early warning capability of the measuring point for a specific fault type in historical data. Specifically:

[0111]

[0112] where, represents the early detection probability of the i-th measuring point in historical fault cases;

[0113] represents the false alarm probability of the measuring point;

[0114] is a small positive number set to prevent division by zero (usually 0.01).

[0115] The weight calculation relies on the fault case library accumulated during system operation and is re-evaluated and updated every six months. The weighted superposition process is achieved by the following formula:

[0116]

[0117] where N is the total number of measuring points, is the focused residual feature vector generated.

[0118] The trend prediction module, based on the decoupled features of operating conditions and the focused residual feature vectors, calculates fault development trend data and remaining life estimation data through the fault prediction model. This fusion approach allows the prediction model to utilize both the obvious fault information in the main features and the early, subtle signs in the residual features.

[0119] Through the above technical solution, this application can effectively enhance the detection sensitivity of early faults, especially for weak fault features that are easily masked by the main vibration components in conventional vibration analysis. The multi-measurement point weighted fusion strategy makes full use of monitoring information at different locations on the transmission chain. By assigning higher weights to fault-sensitive areas, it optimizes the feature representation efficiency, thereby improving the early warning capability and prediction accuracy of the fault prediction system for potential faults.

[0120] like Figure 4 As shown, in one embodiment, this application further proposes that the system also includes a periodic drift identification module, which aims to obtain dynamic features reflecting changes in the state of mechanical components such as clearance and wear by analyzing the temporal stability of periodic events in the transmission system. This periodic drift identification module operates in parallel with the decoupled feature generation for operating conditions, together forming a multi-dimensional feature extraction system.

[0121] In practical implementation, the periodic drift identification module is used for:

[0122] This study identifies event time sequence of periodic meshing or impact events from vibration acceleration signals. A peak detection algorithm based on wavelet transform is employed. A Morlet wavelet basis function is designed centered on the inherent meshing frequency of the transmission system. By identifying the modulus maxima of the wavelet coefficients, the event occurrence time is accurately located, forming the original event time sequence. , where i=1,2,...,M represents the event sequence number.

[0123] Synchronously, the system acquires the rotational speed signal of the mechanical transmission system through an encoder mounted on the input shaft. This rotational speed signal records the rotor angle change in the form of a pulse sequence. By mapping the event time point sequence to the same rotor cycle reference system through the rotational speed signal, a standardized event time point sequence is obtained.

[0124] Specifically, a unified time base is constructed using rotational speed signals, mapping the original event time sequence to the angle domain:

[0125]

[0126] in, This represents the standardized perspective of the event. The reference start time for an analysis window. for the rotor rotation period. This mapping process eliminates the influence of speed fluctuation on the event timing analysis, and obtains a sequence of normalized event time points .

[0127] According to the sequence of normalized event time points, a sequence of event time intervals of adjacent periods is calculated , and the calculation formula is as follows:

[0128]

[0129] wherein j and j+1 represent the current event time point and the next event time point respectively;

[0130] The sequence of event time intervals reflects the stability of the periodic event in the angle domain.

[0131] The change rate in the preset sliding window of the sequence of event time intervals is calculated as a sequence of drift rates. The system uses a sliding window with a length of 20-30 periods to calculate the change rate of the interval sequence, i.e. the drift rate:

[0132]

[0133] wherein represents the drift rate of the kth window, is the length of the sliding window, which is usually set to 20-30 event periods. The determination of this length is based on the following principles: it should be much larger than the inherent fluctuation period of the transmission system speed (usually <5 periods) to ensure that random fluctuations can be smoothed; at the same time, it should be much smaller than the typical time scale of the period drift caused by the development of the fault (usually >50 periods) to ensure the sensitivity of the early fault detection.

[0134] Based on the calculated sequence of drift rates , the system constructs a period drift feature vector as an auxiliary feature vector. This vector contains statistical features of the drift rate: mean, standard deviation, skewness and proportion exceeding the threshold value (usually set to 0.5%), forming a four-dimensional feature vector . These features describe the stability characteristics of the periodic event from different angles, wherein the mean reflects the average drift trend, the standard deviation represents the fluctuation degree, the skewness indicates the asymmetry of the distribution, and the proportion exceeding the threshold value reflects the degree of abnormality.

[0135] The trend prediction module calculates the fault development trend data and the remaining life estimation data based on the working condition decoupling features and the period drift feature vector through the fault prediction model. This fusion makes the prediction system not only focus on the change of vibration energy, but also consider the evolution of the dynamic timing characteristics of the mechanical system.

[0136] By introducing the periodic drift feature analysis, the application can capture the subtle timing variations caused by the increase of gear backlash, the change of bearing clearance and other mechanical state changes. This feature in the time dimension complements the vibration energy feature, providing additional evidence for fault prediction, especially helping to identify early progressive failures that mainly show changes in fit rather than increases in vibration energy, thereby improving the comprehensiveness and reliability of the prediction system.

[0137] As Figure 5 shown, in one embodiment, the application further proposes that the system further comprises a conduction delay difference extraction module, which aims to capture the dynamic characteristic changes caused by structural stiffness degradation, connection loosening or component wear by analyzing the changes in the propagation characteristics of the vibration signal in the transmission chain, and to provide supplementary features reflecting wave propagation characteristics for fault prediction.

[0138] The conduction delay difference extraction module is used to:

[0139] Perform short-time cross-correlation analysis on the synchronous vibration acceleration signals of different measuring points on the transmission chain to obtain the conduction delay difference sequence between the signals.

[0140] In specific implementation, the conduction delay difference extraction module first performs short-time cross-correlation analysis on the synchronous vibration acceleration signals of different measuring points on the transmission chain. A sliding time window processing method is adopted, the window length is set to the duration of 2-3 main impact events (usually 100-200 milliseconds), and the window overlap rate is set to 50% to ensure the continuity of the analysis. For any two adjacent measuring points i and j, the cross-correlation function in the time window [ , ] is calculated as follows:

[0141] Where τ is the time delay variable.

[0142] The conduction delay difference is the τ value corresponding to the maximum value of the cross-correlation function:

[0143]

[0144] By repeating this process for consecutive time windows, the conduction delay difference sequence

[0145] is obtained.

[0146] Considering the influence of load changes on wave propagation speed, the conduction delay difference sequence is scaled and corrected based on the load torque signal to eliminate the influence of load changes on conduction speed. The relationship between load and conduction speed is established through the early calibration experiment, and the correction formula is:

[0147] ​​

[0148] wherein, is the corrected conduction delay difference, is the measured conduction delay difference, is the current load torque, is the rated load torque, k is the load influence coefficient (determined by experiment, typical value is 0.001-0.005 N· ). This correction eliminates the systematic influence of load fluctuations on the conduction delay difference, allowing subsequent analysis to focus on delay changes caused by equipment status.

[0149] Based on the corrected conduction delay difference sequence, a conduction delay difference feature vector is constructed as an auxiliary feature vector. This conduction delay difference feature vector contains multiple statistical features such as mean, standard deviation, coefficient of variation, and proportion of delay differences exceeding a threshold value (usually set to 15% of the reference value). In addition, the autocorrelation function decay time of the delay difference sequence is calculated to assess the persistence of delay changes. These features together form a multi-dimensional feature vector that describes the stability of wave propagation characteristics.

[0150] The trend prediction module calculates the fault development trend data and remaining life estimation data based on the working condition decoupling features and the conduction delay difference feature vector through the fault prediction model.

[0151] Through the above conduction delay difference feature extraction process, the application can effectively monitor the characteristic changes of the wave propagation path in the mechanical transmission system, which are often closely related to the structural connection state, component gap, and material stiffness characteristics. The conduction delay difference feature vector is used as an auxiliary feature to fuse with the main working condition decoupling features, providing unique information reflecting wave propagation characteristics for the fault prediction model, enhancing the early identification ability of faults such as structural looseness, bearing wear, and gear gap abnormalities, and providing supplementary diagnostic evidence that traditional vibration analysis cannot obtain, especially in predicting fault types related to dynamic force transmission paths.

[0152] The application further proposes that the fault development trend data of the mechanical transmission system is calculated through a pre-trained fault prediction model, specifically including the following steps:

[0153] Cache the working condition decoupling features of the last N time points in chronological order to generate a working condition decoupling feature time sequence matrix. In the specific implementation process, the system first establishes a time sequence analysis framework for working condition decoupling features. Cache the working condition decoupling features of the last 128 sampling time points in chronological order. The determination of this value is based on the time scale of typical fault development of the transmission system, which can cover several hours to several days of operation data. Arrange these features in chronological order to form a working condition decoupling feature time sequence matrix where d is the feature dimension.

[0154] The working condition decoupling feature time series matrix is subjected to principal component analysis along the time dimension to extract the first principal component score sequence containing the maximum variance information , which reflects the most important change mode of the equipment state. Linear fitting is performed on the first principal component score sequence:

[0155]

[0156] where a is the intercept, b is the slope obtained by fitting, and is the residual term. The slope b obtained by fitting is taken as the global trend intensity factor, and its positive value indicates that the state develops in the direction of degradation, and its negative value indicates that the state recovers, and the absolute value size reflects the change rate.

[0157] The working condition decoupling feature at the current time is input into the pre-trained multi-classification neural network. The training process of the pre-trained multi-classification neural network and the fault prediction model is completed based on historical full life cycle data. The training data set is derived from at least 30 mechanical transmission systems of the same type from healthy state operation to complete data records of failure, containing more than 500 sample sequences covering different loads and lubrication conditions. After time-frequency transformation of the original vibration signal in the data preprocessing stage, the Z-score standardization method is used to eliminate the dimension influence. The multi-classification neural network uses a deep fully connected architecture, contains 3 hidden layers, each layer has 256 neurons, uses ReLU activation function, and is trained by historical running to failure full life cycle data. During training, the Adam optimizer is used, the learning rate is set to 0.001, and the batch size is 32.

[0158] Output the probability distribution of the mechanical transmission system jumping from the current healthy state to multiple predefined degradation states within a preset first time window (for example, a 24-hour time window) in the future where m is the number of state categories (usually set to 4-6 ordered degradation states). Based on the probability distribution, the state deterioration tendency index is calculated:

[0159]

[0160] where, is the weight coefficient of the i-th state, which is set to increase according to the state severity.

[0161] ​Finally, the global trend intensity factor and the state deterioration tendency index are weighted and fused to generate a comprehensive fault development trend score, which is the main component of the fault development trend data. The weight distribution is based on the historical prediction performance of each indicator, and the trend intensity factor is usually set to 0.6, and the state deterioration tendency index is usually set to 0.4, to form a comprehensive fault development trend score:

[0162]

[0163] wherein, and are normalization coefficients to ensure that the two indicators have the same dimension.

[0164] Through the above fault development trend score calculation method, the present application can simultaneously consider the long-term slow change trend of the device state and the short-term state transition risk, reflecting both the gradual degradation of the state and the possibility of state mutation. This dual perspective evaluation mechanism provides a more comprehensive basis for predictive maintenance decisions, allowing maintenance personnel to adjust maintenance strategies in a timely manner based on score changes, ensuring safe operation of the equipment while optimizing maintenance resource allocation.

[0165] The present application further proposes that the specific steps of calculating the fault development trend data and the remaining life estimation data of the mechanical transmission system by the pre-trained fault prediction model include:

[0166] The pre-stored fault feature template is called to calculate the feature degradation distance D between the current working condition decoupling feature and the fault feature template. The fault feature template is pre-existing in the device database and is derived from the feature state when the historical operation is running to complete failure. It is established by collecting the average value of the features of a plurality of same type devices in the last 8 hours before failure.

[0167] The feature degradation distance D is calculated using Mahalanobis distance to consider the correlation between the feature dimensions:

[0168]

[0169] wherein D is the feature degradation distance, represents the working condition decoupling feature vector at the current time, is the fault feature template vector, and Σ is the covariance matrix of the feature distribution, which is estimated by historical normal operation data.

[0170] At the same time, the system calculates the change trend slope K of the working condition decoupling feature in a preset second time window (usually set to 8-24 hours, preferably 24 hours). This calculation uses a weighted least squares method to linearly fit the feature sequence of the last 24 sampling points, and the recent data points are given higher weights to more sensitively reflect the recent state change trend.

[0171] The characteristic degradation distance D and the change trend slope K are input into a pre-trained nonlinear function to calculate the remaining useful life estimate; the nonlinear function is trained by collecting no less than 50 groups of full life cycle data from running to failure, and the Levenberg-Marquardt optimization algorithm is used to solve the model parameters.

[0172] The nonlinear function is trained by historical full life cycle data from running to failure, and its expression is:

[0173]

[0174] Wherein, RUL is the remaining useful life estimate (unit: hour), D is the characteristic degradation distance, K is the change trend slope, A represents the initial remaining useful life reference value of the equipment in the healthy state, B controls the influence strength of the characteristic degradation distance on the life estimate, and C adjusts the weight of the state change trend. The model parameters A, B and C are determined through the training process, and the typical value intervals are A∈[200, 500], B∈[0.1, 0.5] and C∈[50, 150], respectively. The specific values are different for different types of equipment and operating conditions.

[0175] In the training process, the objective function is set to minimize the root mean square error between the predicted remaining useful life and the actual remaining useful life, and the training data needs to cover the complete degradation process from the healthy state to the complete failure, including diversified degradation paths under different operating conditions and load conditions.

[0176] Through the above remaining useful life estimation method, the system can dynamically estimate the remaining useful life of the equipment based on the relative position of the current state and the failure state and the state change rate. Both the absolute degradation degree of the state and the trend information of the state change are considered, so that the life estimation result can timely reflect the accelerated degradation or recovery of the equipment state, providing a more accurate and adaptive time reference for the development of maintenance plan, especially showing better prediction adaptability in the face of nonlinear degradation process.

[0177] The following is a specific embodiment of a mechanical transmission system fault trend prediction system based on dynamic feature recognition:

[0178] A large-scale open-pit mine main conveyor belt gear transmission system (model ZLYJ250, rated power 160kW) has been running in a working condition with dense dust and frequent load fluctuations for a long time, the rated load switching range covers 50% to 100%, and the lubrication system is easily contaminated by particulate matter, resulting in dynamic changes in state. Traditional fault prediction methods often misreport faults due to sudden load increases, or miss hidden dangers due to lubrication contamination masking wear characteristics. After applying the fault trend prediction system based on dynamic feature recognition, accurate monitoring and early warning under complex working conditions are realized.

[0179] System deployment multi-source sensor array to achieve synchronous data acquisition: input shaft installed HBM T40B torque sensor (sampling rate 100 Hz) real-time acquisition of load torque signal; gear box lubricating oil integrated viscosity sensor (measurement range 2-20 cSt) and dielectric constant sensor (sampling rate 50 Hz) to monitor the lubrication state; transmission chain bearing seat distribution 3 sets of PCB 352C65 three-axis acceleration sensor (sampling rate 2048 Hz) to collect vibration acceleration signal. All sensors realize hardware level synchronization through IEEE 1588 PTP protocol, and the time error is controlled within 1 ms.

[0180] The vibration signal is transformed into time-frequency distribution data S(f, t) by short-time Fourier transform (window length 1024 points, Hanning window, overlap rate 75%). When the real-time load torque is 80% of the rated load, the system calls the pre-stored load disturbance spectrum template P(f) in the medium load interval; the total energy of the current time-frequency distribution in the gear meshing frequency band (250 Hz ± 50 Hz) is calculated (relative energy unit), template energy (relative energy unit), preset gain parameter =0.8, adaptive scaling factor ; through adaptive differential processing , effectively eliminate the interference of load fluctuation. For the lubrication state (current viscosity 18 cSt, dielectric constant 1.8), call the lubrication characteristic form mapping relationship to generate gain coefficient G(f) - based on multivariate linear regression model to predict energy distribution =[1.36,2.39], compared with the reference energy distribution =1.13, =1.195; form reorganization is realized by Hadamard product , compensate for the spectral distortion caused by lubrication pollution, output working condition decoupling characteristics.

[0181] The period drift identification module identifies the gear meshing event time point sequence from the vibration signal, maps to the rotor period reference frame combined with the speed encoder signal (1024 pulses / revolution), calculates the interval between adjacent period events , sliding window drift rate , construct period drift feature vector; conduction delay difference extraction module performs short-time cross-correlation analysis on vibration signals at different measuring points, and obtains conduction delay difference sequence after load scaling correction, extracts statistical features to form auxiliary feature vector.

[0182] ​The fault prediction model receives decoupling features and auxiliary features of operating conditions, caches the decoupling features of the most recent 128 time points to generate a time series matrix, extracts the first principal component score sequence through principal component analysis and obtains a global trend intensity factor b=0.03 through linear fitting; the multi-class neural network outputs the probability distribution of degradation state in the next 24 hours. =[0.6,0.3,0.1], calculate the deterioration tendency index. =0.6 + 10.3 + 20.1 = 0.5; Overall fault development trend score =100.03 + 0.50.5 = 0.55, indicating a slow degradation trend in the system. The remaining lifetime is estimated by calculating the characteristic degradation distance D = 1.2 (Mahanobis distance) and the trend slope K = 0.04, and then substituting these values ​​into a nonlinear function. =300 exp(-0.41.2)+500.04≈187 hours, providing a precise basis for the maintenance plan.

[0183] Traditional monitoring systems often trigger false alarms under these mine conditions due to sudden load increases (such as increased vibration amplitude during heavy-load startup being misjudged as a fault) or miss alarms due to lubrication contamination masking early wear characteristics. After applying this system, in a case where the load jumped from 60% to 90%, adaptive differential processing successfully isolated the load disturbance and did not generate false alarms. When the dielectric constant of the lubricating oil increased abnormally (indicating contamination), morphological recombination compensation effectively captured the early wear characteristics of the gear tooth surface, issuing an early warning. Maintenance personnel replaced the lubricating oil and worn gears in time, avoiding mine production interruptions caused by conveyor belt shutdowns and improving the reliability and economy of equipment operation and maintenance.

[0184] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A mechanical transmission system failure trend prediction system based on dynamic signature recognition, characterized by: The method comprises the following steps: a signal acquisition module for synchronously acquiring load torque signals and lubrication state parameter signals of a mechanical transmission system and vibration acceleration signals of multiple measuring points on a transmission chain, and sending the acquired multi-source signals to a working condition decoupling feature generation module; a working condition decoupling feature generation module connected with the signal acquisition module, for receiving the multi-source signals and performing the following operations: generating time-frequency distribution data based on the vibration acceleration signals; calling corresponding load disturbance spectrum templates from a pre-stored load disturbance spectrum mapping relationship according to the load torque signals, and performing adaptive differential processing on the time-frequency distribution data to generate processed time-frequency distribution data; calling corresponding correction rule sets according to the lubrication state parameter signals, performing morphological reorganization on the processed time-frequency distribution data, outputting working condition decoupling features representing the health state of the mechanical component, and sending the working condition decoupling features to a trend prediction module; a trend prediction module connected with the working condition decoupling feature generation module, for calculating fault development trend data and remaining life estimation data of the mechanical transmission system based on the working condition decoupling features through a pre-trained fault prediction model; wherein the fault development trend data of the mechanical transmission system calculated through the pre-trained fault prediction model comprises the following steps: buffering the working condition decoupling features of the last N time points in time sequence to generate a working condition decoupling feature time sequence matrix; performing principal component analysis on the working condition decoupling feature time sequence matrix along the time dimension to extract a first principal component score sequence and perform linear fitting, and taking the slope obtained by fitting as a global trend intensity factor; inputting the working condition decoupling features at the current time into a pre-trained multi-classification neural network to output the probability distribution of the mechanical transmission system jumping from the current health state to multiple predefined degradation states within a preset first time window in the future, and calculating a state deterioration tendency index; weighting and fusing the global trend intensity factor and the state deterioration tendency index to generate a comprehensive fault development trend score as a component of the fault development trend data.

2. The dynamic signature based mechanical drive system failure trend prediction system of claim 1, wherein: In the step of calling corresponding load disturbance spectrum templates from a pre-stored load disturbance spectrum mapping relationship, the construction and updating process of the load disturbance spectrum mapping relationship comprises: acquiring vibration acceleration signals under different load torque signals during the stable operation stage of the mechanical transmission system determined to be in a healthy state; extracting time-frequency distribution data of the vibration acceleration signals, calculating average time-frequency distribution data corresponding to each load torque signal, and storing the average time-frequency distribution data as the load disturbance spectrum templates; during online monitoring, dynamically updating the load disturbance spectrum mapping relationship based on newly acquired health state data through a dynamic updating algorithm.

3. The dynamic signature based mechanical drive system failure trend prediction system of claim 1, wherein: The specific steps of calling corresponding load disturbance spectrum templates from a pre-stored load disturbance spectrum mapping relationship according to the load torque signals, and performing adaptive differential processing on the time-frequency distribution data comprise: mapping the load torque signal to a preset plurality of load intervals, and calling a load disturbance spectrum template corresponding to each load interval from the load disturbance spectrum mapping relationship wherein denotes frequency; calculating current time-frequency distribution data total energy in the characteristic frequency band and the load disturbance spectrum template template energy in the same characteristic frequency band calculating an adaptive scaling factor : wherein is a preset gain parameter Adaptively differentiating the time-frequency distribution data to generate the processed time-frequency distribution data The calculation formula is: .

4. The dynamic signature based mechanical drive system failure trend prediction system of claim 3, wherein: According to the lubrication state parameter signal, a corresponding correction rule set is called; wherein the correction rule set includes a lubrication feature morphology mapping relationship, and the establishment process of the lubrication feature morphology mapping relationship includes: In the running process of the mechanical transmission system, the lubrication state parameter signal and the synchronous vibration acceleration signal are continuously monitored and collected; When the fluctuation rate of the load torque signal is lower than a first preset threshold value and the total energy fluctuation rate of the vibration acceleration signal is lower than a second preset threshold value, it is determined that the equipment is in a stable and healthy running state, and the energy distribution of the lubrication state parameter signal and the time-frequency distribution data generated based on the synchronous vibration acceleration signal in different frequency bands at the current time are recorded and stored as a sample pair; Using the accumulated sample pairs, a quantitative prediction model is fitted through a multivariate linear regression algorithm; The parameters of the quantitative prediction model are stored and updated as the lubrication feature morphology mapping relationship; wherein the correction rule set is the gain coefficient predicted after inputting the current lubrication state parameter signal into the quantitative prediction model .

5. The dynamic signature based mechanical drive system failure trend prediction system of claim 4, wherein: According to the lubrication state parameter signal, a corresponding correction rule set is called, and the specific steps of morphological reorganization of the processed time-frequency distribution data include: The lubrication state parameter signal is compared with the boundary values of a plurality of preset lubrication state intervals to determine the current belonging lubrication state category; According to the lubrication state category, the gain coefficient for different frequency bands is called from the lubrication feature morphology mapping relationship ; The processed time-frequency distribution data with the gain coefficient The frequency point is multiplied, the spectrum form is adjusted, and the working condition decoupling characteristics are output The calculation formula is: ； wherein, denotes a Hadamard product, used to compensate or suppress the energy of the frequency bands that are significantly affected by the lubrication state.

6. The dynamic signature based mechanical drive system failure trend prediction system of claim 1, wherein: The specific steps of calculating the fault development trend data and the remaining useful life estimation data of the mechanical transmission system through the pre-trained fault prediction model include: A pre-stored fault feature template is called, and the feature degradation distance between the working condition decoupling feature at the current time and the fault feature template is calculated; The feature degradation distance and the change trend slope of the working condition decoupling feature within a preset second time window are input into a pre-trained nonlinear function to calculate the remaining useful life estimation value; The nonlinear function is obtained by training full life cycle data from historical operation to failure, and an expression thereof is: Wherein, RUL is the remaining useful life estimation value, D is the feature degradation distance, K is the change trend slope, A, B, and C are model parameters determined through training.

Images