Feature matrix-based gearbox life prediction method and system

Abstract

The invention provides a gear box service life prediction method and system based on a characteristic matrix, relates to the technical field of gear box service life prediction, and aims to solve the technical problem that the gear box service life prediction method is relatively poor in related technologies. The method comprises the following steps: acquiring multi-source operation data of a gearbox in a preset operation period; dividing the preset operation cycle into a plurality of time slices, and dividing the operation condition of the gearbox into a plurality of condition intervals; extracting a vibration signal in the multi-source operation data, and obtaining an energy dispersion index corresponding to the vibration signal; constructing a space-time state degradation matrix; filling the energy dispersion index obtained in each time slice into a matrix unit in which a working condition interval corresponding to the time slice and the main operation working condition of the time slice is located in a space-time state degradation matrix; identifying a row of which the index value presents the maximum increasing trend along with time change as a main degradation path; and predicting the remaining service life of the gearbox.

Description

Technical Field

[0001] This application relates to the field of gearbox life prediction technology, and in particular to a gearbox life prediction method and system based on a feature matrix. Background Technology

[0002] Gearboxes, as core components for transmitting power and motion in modern industrial equipment, are widely used in critical fields such as wind power generation, rail transportation, aerospace, and metallurgy and mining. Their reliability and safety directly affect the stability of the entire system and even production safety. Due to long-term operation under heavy loads, variable speeds, and complex environments, critical components such as gears and bearings within gearboxes inevitably experience various forms of failure, including wear, fatigue, pitting, galling, and even fracture. Failure to detect and predict the evolution of these failures in a timely manner can lead to catastrophic equipment downtime and significant economic losses. Therefore, effectively monitoring the health status of gearboxes and accurately predicting their Remaining Useful Life (RUL) is crucial for achieving predictive maintenance, improving equipment availability, and reducing operating costs.

[0003] In related technologies, gearbox life prediction technology mainly relies on vibration signal analysis and data-driven models. Traditional methods typically involve extracting various time-domain, frequency-domain, or time-frequency-domain feature indicators from the raw vibration signal, such as root mean square, kurtosis, margin, spectral kurtosis, and wavelet packet energy. Then, the evolution trend of these feature indicators is used to determine the degradation state. However, these single or combined feature indicators are often insensitive in the early stages of degradation and are easily affected by changes in operating conditions (such as speed and load), leading to unstable feature extraction and difficulty in establishing a clear correspondence between feature extraction and lifespan.

[0004] To overcome the aforementioned problems, intelligent prediction methods based on machine learning and deep learning, such as Support Vector Machines (SVM), Neural Networks (NN), and Long Short-Term Memory Networks (LSTM), can automatically learn the complex nonlinear mapping relationship between raw data and remaining lifespan, exhibiting high prediction accuracy on certain datasets. However, since deep learning models are often considered "black boxes," their internal decision-making logic is difficult to understand and verify. For industrial applications with high reliability requirements, unexplained predictions are difficult to trust and cannot provide in-depth physical insights for maintenance decisions. Most models are trained on a limited number of operating conditions. When new combinations of operating conditions not included in the training set appear in actual operation, the model's performance may drop sharply. How to effectively fuse degradation information under different operating conditions and identify the key operating conditions leading to lifespan loss is a major challenge for current technology. Furthermore, the training process of many intelligent models is randomized, resulting in different models each time they are trained, and the prediction results may have slight differences. This uncertainty is detrimental to lifespan prediction tasks that require high accuracy and high reliability. Summary of the Invention

[0005] This application provides a gearbox life prediction method and system based on feature matrix, which improves upon at least one of the following technical problems in related technologies: gearbox life prediction methods rely on black-box models, have poor interpretability, lack robustness to changes in operating conditions, and lack determinism in the prediction process.

[0006] To achieve the above objectives, the embodiments of this application adopt the following technical solutions:

[0007] In a first aspect, this application provides a gearbox life prediction method based on a feature matrix, comprising: acquiring multi-source operating data of a gearbox within a preset operating cycle; dividing the preset operating cycle into multiple time segments and the operating conditions of the gearbox into multiple operating condition intervals based on the operating condition parameters in the multi-source operating data; extracting vibration signals from the multi-source operating data for each time segment and obtaining the energy dispersion index corresponding to the vibration signals; constructing a spatiotemporal state degradation matrix, wherein the rows of the spatiotemporal state degradation matrix correspond to the operating condition intervals and the columns correspond to the time segments; filling the matrix cells in the spatiotemporal state degradation matrix containing the energy dispersion index obtained in each time segment, which correspond to the operating condition interval of the time segment and the main operating condition of the time segment; identifying the row in the spatiotemporal state degradation matrix where the index value shows the maximum growth trend over time as the main degradation path; and predicting the remaining service life of the gearbox based on the energy dispersion index sequence on the main degradation path and a preset failure threshold.

[0008] In one possible implementation of the first aspect, acquiring the multi-source operating data of the gearbox within a preset operating cycle includes: acquiring vibration signal data, temperature data, rotational speed data, and load data of the gearbox.

[0009] In one possible implementation of the first aspect, dividing the operating conditions of the gearbox into multiple operating condition intervals includes: defining a two-dimensional operating condition grid based on the speed data and the load data; and defining each grid cell of the two-dimensional operating condition grid as an operating condition interval.

[0010] In one possible implementation of the first aspect, after the step of filling the spatiotemporal state degradation matrix with the energy dispersion index obtained in each time segment, the method further includes: for the unfilled matrix cells in the spatiotemporal state degradation matrix, performing state inertia filling based on the index value of the matrix cell corresponding to its adjacent previous time segment.

[0011] In one possible implementation of the first aspect, obtaining the energy dispersion index corresponding to the vibration signal includes: performing spectral analysis on the vibration signal within the time segment to obtain an energy spectrum; identifying the gear meshing frequency and the energy peak corresponding to its harmonics in the energy spectrum; and obtaining the energy dispersion index based on the ratio of the frequency band energy other than the energy peak in the energy spectrum to the total energy of the energy peak.

[0012] In one possible implementation of the first aspect, identifying the main degradation path includes: for each row of the spatiotemporal state degradation matrix, obtaining the gradient of its energy dispersion index value over time; and selecting the row with the largest positive gradient value as the main degradation path.

[0013] In one possible implementation of the first aspect, predicting the remaining service life of the gearbox based on the energy dispersion index sequence on the main degradation path includes: performing trend extrapolation on the energy dispersion index sequence on the main degradation path to obtain a degradation trend curve; determining the time point at which the degradation trend curve reaches the preset failure threshold; and obtaining the remaining service life of the gearbox based on the time point and the current time.

[0014] Secondly, this application provides a gearbox life prediction system based on a feature matrix, comprising: a data acquisition module for acquiring multi-source operating data of the gearbox within a preset operating cycle; a working condition division module for dividing the preset operating cycle into multiple time segments and dividing the operating conditions of the gearbox into multiple working condition intervals based on the operating condition parameters in the multi-source operating data; an index quantification module for extracting vibration signals from the multi-source operating data for each time segment and obtaining the energy dispersion index corresponding to the vibration signals; and a matrix construction module for constructing a spatiotemporal state degradation matrix. The rows of the spatiotemporal state degradation matrix correspond to the operating condition intervals, and the columns correspond to the time segments. The energy dispersion index obtained in each time segment is filled into the matrix cell of the spatiotemporal state degradation matrix that corresponds to the time segment and the main operating condition in which the time segment is located. The path identification module is used to identify the row in the spatiotemporal state degradation matrix where the index value shows the largest growth trend over time, as the main degradation path. The life prediction module is used to predict the remaining life of the gearbox based on the energy dispersion index sequence on the main degradation path and a preset failure threshold.

[0015] In one possible implementation of the second aspect, it further includes: a life correction module, used to: acquire the temperature data of the gearbox and construct a thermal stress matrix; and correct the remaining service life based on the thermal stress value corresponding to the main degradation path in the thermal stress matrix.

[0016] In one possible implementation of the second aspect, the index quantification module is further configured to: acquire the load data of the gearbox and acquire the load instability factor within each time segment; and adjust the energy dispersion index based on the load instability factor to reflect the impact of impact load on gearbox degradation. Attached Figure Description

[0017] Figure 1 A schematic flowchart of a gearbox life prediction system based on a feature matrix provided for some embodiments of this application;

[0018] Figure 2 A schematic diagram of the structure of a gearbox life prediction method based on a feature matrix provided for some embodiments of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0020] In the following description, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0021] Furthermore, in this application, directional terms such as "upper," "lower," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and may change accordingly depending on the orientation of the components in the accompanying drawings.

[0022] In this application, unless otherwise expressly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a direct connection or an indirect connection through an intermediate medium. Furthermore, the term "electrical connection" can refer to the manner in which an electrical connection is used to achieve signal transmission.

[0023] As used herein, “about,” “approximately,” or “approximately” includes the stated value and a reference value within an acceptable range of deviation from the given value, characterized in that the acceptable range of deviation is determined by a person skilled in the art taking into account the measurement under discussion and the error associated with the measurement of the given quantity (i.e., the limitations of the measurement method).

[0024] This embodiment provides a gearbox life prediction method based on a feature matrix. The method steps disclosed in this embodiment will be described in detail below. Figure 1 As shown, the method includes:

[0025] S100: Obtain multi-source operating data of the gearbox within a preset operating cycle.

[0026] To accurately understand the operating status and health of the gearbox, it is necessary to collect signal data from multiple physical dimensions during its operation. In this embodiment, the multi-source operating data includes vibration signal data, temperature data, rotational speed data, and load data.

[0027] The vibration signal data is core information revealing the health status of internal mechanical components of the gearbox, such as gears and bearings. When early defects such as pitting, wear, and cracks appear on the gear surface, abnormal impact vibrations are generated during meshing. The variation patterns of these vibration signals contain degradation information. The data acquisition system acquires high-frequency vibration signals by deploying one or more accelerometers at key locations in the gearbox housing, such as near the bearing housing or gear meshing area. For example, an accelerometer with a sampling frequency of 20kHz can be used to continuously or intermittently acquire the vibration waveforms of the gearbox throughout its entire operating cycle.

[0028] The temperature data reflects the overall thermal load status of the gearbox. Abnormal friction and wear can lead to localized temperature increases, while sustained high temperatures accelerate lubricant aging and material degradation. Temperature data can be obtained by installing thermocouples or infrared sensors on the gearbox oil sump or the outer surface of the gearbox.

[0029] The speed and load data are key parameters defining the gearbox's operating conditions. The wear rate of the gearbox is closely related to the speed and load levels it experiences; under different operating conditions, its degradation mode and rate may vary significantly. This data can typically be read directly from the equipment's main control system, such as a programmable logic controller (PLC) or a distributed control system (DCS). For example, speed data can be obtained from a controller connected to the motor encoder, measured in revolutions per minute (RPM); load data can be indirectly obtained from the inverter's output torque or current signal and quantified as a percentage of the rated load.

[0030] The preset operating cycle can be a complete equipment lifespan or any monitoring period requiring health assessment and lifespan prediction, such as from the time the equipment is first put into use until the current moment. All collected data is accompanied by a precise timestamp to ensure consistency in the time dimension during subsequent analysis.

[0031] S200. Based on the operating condition parameters in the multi-source operating data, the preset operating cycle is divided into multiple time segments, and the operating condition of the gearbox is divided into multiple operating condition intervals.

[0032] To map a continuous physical process into a discrete representation that facilitates structured analysis, this method discretizes the gearbox's operation. This discretization includes dividing the operation into preset operating cycles and operating conditions.

[0033] The division of operating time involves dividing the entire preset operating cycle into a series of time segments T1, T2, ..., TN of equal or unequal length, based on chronological order, where N is the total number of time segments. The length of each time segment must balance the statistical significance of the signal characteristics with the ability to capture changes in operating conditions. For example, every 10 consecutive operating hours can be defined as one time segment. If the gearbox operates for a total of 1000 hours, then 100 time segments can be obtained, i.e., M = 100.

[0034] The division of operating conditions involves discretizing the continuous operating condition space defined by speed and load. To this end, this method constructs a two-dimensional operating condition grid. OpCond The grid has two dimensions: speed and load. Specifically, the entire operating speed range and load range of the gearbox are divided into several non-overlapping intervals.

[0035] For example, suppose a gearbox has a rated speed of 3000 RPM and an operating speed range of [0, 3000] RPM. This range can be divided into M... S A speed range, for example, divided into three ranges:

[0036] Speed ​​range S1: [0, 1000] RPM (defined as low speed)

[0037] Speed ​​range S2: (1000, 2000] RPM (defined as medium speed)

[0038] Speed ​​range S3: (2000, 3000] RPM (defined as high speed)

[0039] In this example, M S =3.

[0040] Similarly, assume its operating load range is [0, 100]% of the rated load. This range can be divided into M... L A load range, for example, divided into three ranges:

[0041] Load range L1: [0, 40]% (defined as light load)

[0042] Load range L2: (40, 70)% (defined as medium load)

[0043] Load range L3: (70, 100]% (defined as heavy load)

[0044] In this example, M L =3.

[0045] By this M S Each speed range and M L Combining the Cartesian products of the load intervals forms an M.S ×M L Two-dimensional working condition grid OpCond In this example, it is a 3×3 grid containing 9 grid cells. Each grid cell C i,j =(S i ,L j ), where i∈1,2,3 and j∈1,2,3, defines a specific, discrete operating condition interval. For example, operating condition interval C 3,3 This represents the operating state of the gearbox under high-speed, heavy-load conditions. These discrete operating condition intervals will serve as row indices for the subsequently constructed spatiotemporal state degradation matrix.

[0046] After discretizing the time and operating conditions, it is necessary to determine the primary operating condition for each time segment Tk. This can be achieved by analyzing the speed and load data within that time segment. One method is to calculate the primary operating condition for time segment Tk. This is done by analyzing the speed and load data within that time segment. Another method is to calculate the primary operating condition for time segment Tk. k Within, the gearbox operates in each operating condition range C. i,j The total duration is calculated, and the longest operating condition interval is selected as the time segment.

[0047] S300. For each time segment, extract the vibration signal from the multi-source operation data and obtain the energy dispersion index corresponding to the vibration signal.

[0048] The energy dispersion index is used to quantify the early degradation state of the gearbox. In a healthy gearbox, the vibrational energy is highly concentrated at a deterministic frequency determined by gear geometry and rotational speed, namely the gear meshing frequency (f). m ) and its harmonics (n·f m As early defects such as wear and pitting appear on the gear teeth, the smoothness of the meshing process is disrupted, leading to modulation phenomena. This causes energy to diffuse from the main harmonic frequency to the sidebands on both sides. Simultaneously, non-periodic impacts excite random vibrations across a wider frequency band, manifesting as a rise in the spectral baseline. Therefore, the degree to which vibration energy shifts from "concentration" to "dispersion" is directly related to the health of the gear.

[0049] Energy Dispersion Index I EDF The specific acquisition process is as follows:

[0050] For a given time segment, first extract the vibration signal data within that segment. Perform spectral analysis on this vibration signal, typically using a Fast Fourier Transform (FFT) to transform it from the time domain to the frequency domain, obtaining the energy spectrum P(f). Identify the key energy concentration regions within the energy spectrum P(f). This step requires combining the dominant operating conditions of the time segment Tk (especially rotational speed information) to calculate the theoretical gear meshing frequency fm and its first few (e.g., the first K) harmonic frequencies: f m ,2f m ,...,Kf m This portion of energy represents the "concentrated" energy associated with normal meshing. The total energy E across the entire analysis band is then calculated. total .

[0051] Energy Dispersion Index I EDF Defined as the ratio of non-concentrated energy to total energy, it reflects the relative dispersion of energy. In this embodiment, it can be expressed as I. EDF =(E total -E peak ) / E total A very healthy gear, its E peak Very close to E total Therefore I EDF The value is close to 0. As degradation occurs, energy diffuses towards the sidebands and the noise baseline, E peak The decrease in its proportion of total energy leads to I EDF The value gradually increases. This index is a relative ratio, therefore it has good robustness to changes in the overall energy amplitude caused by variations in load or speed.

[0052] By repeating the above process for each time segment T1, T2, ..., TN, an energy dispersion index sequence corresponding to the time series can be obtained: I EDF (T1),I EDF (T2),...,I EDF (T N ).

[0053] S400. Construct a spatiotemporal state degradation matrix and fill it with the energy dispersion index obtained in each time segment.

[0054] The spatio-temporal state degradation matrix, denoted as M, is a key feature of this matrix. STSD This matrix is ​​used to integrate one-dimensional time-series data and discrete operating condition information into a two-dimensional structured space. It aims to demonstrate the evolution of the degradation process across both time and operating condition dimensions.

[0055] The spatiotemporal state degradation matrix M STSD The structure is defined as follows: It is a two-dimensional matrix whose columns are divided into time segments T1, T2, ..., T in S200. N The index represents the passage of time; its rows are indexed by the operating condition intervals divided in S200. If there is M... S Each speed range and M L There are M load ranges, so there are a total of M total =M S ×M L There are several operating condition intervals. For ease of matrix representation, the two-dimensional operating condition interval C can be represented as a matrix interval. i,j Flatten into a one-dimensional row index k, where k = (i-1) × M L +j,k∈1,2,...,M total Therefore, M STSD It is an M total A matrix of size ×N. Each element M in the matrix. STSD (k,t) represents the degradation state measure of the gearbox in the k-th operating condition interval and the t-th time segment.

[0056] The matrix filling operation involves filling each time segment T in S300. t Calculated energy dispersion index I EDF (T t ), and place it into a matrix cell jointly determined by the time segment and its dominant operating condition.

[0057] The specific filling logic is as follows: for each time segment index t = 1, 2, ..., N:

[0058] Get the time segment T t The dominant operating condition range C identified in S200 dominant (T t ), and the matrix row index k corresponding to this working condition interval. dom .

[0059] Get the time segment T t Energy dispersion index I calculated in S300 EDF (T t ).

[0060] The index value I EDF (T t Fill the corresponding positions in the matrix: M STSD (k dom ,t)=I EDF (T t ).

[0061] After the above steps, for each time segment column, only one cell is filled with the newly calculated degradation index, resulting in a sparse matrix. To facilitate the analysis of continuous degradation trends, this embodiment can also introduce a state inertia filling mechanism. The principle of this mechanism is that even if a certain condition is not dominant in the current time segment, its degradation state should be inherited from the previous time segment. A feasible state inertia filling method is: for any condition row k≠k that is not dominant in time segment t. dom Its matrix element value M STSD (k,t) can be determined by adding a small fundamental degradation increment to its value at the previous time step, i.e., M. STSD (k,t)=M STSD (k,t-1)+ΔI base Where ΔI base It is a constant representing natural aging. For the first column with t=1, all non-dominant elements can be initialized to 0.

[0062] For example, suppose there are 9 operating condition intervals (rows) and 100 time segments (columns), forming a 9×100 matrix M. STSD In time segment T 50 When the dominant operating condition is calculated to be C 3,3 (For high-speed, heavy-load conditions, corresponding to line 9), the energy dispersion index is 0.15, so fill in the value 0.15 in M. STSD (9,50). For the other 8 rows (k = 1,...,8), their value in column 50 is determined by M. STSD (k,49)+ΔI base To calculate.

[0063] M constructed in this way STSD Each row of the matrix depicts the degradation history trajectory under a specific operating condition, and the entire matrix constitutes a "map" of the gearbox's health status evolving over time and operating conditions.

[0064] S500. Identify the main degradation path in the spatiotemporal state degradation matrix.

[0065] The overall lifespan of a gearbox is primarily determined by its degradation rate under the most severe or frequent operating conditions. This most critical degradation process occurs in M... STSD The matrix will show a trajectory where the indicator value increases the fastest and most significantly. This step is to identify this trajectory and define it as the "Primary Degradation Path" (PDP).

[0066] The primary degradation path is M STSDA given row in the matrix represents the operating condition that contributes the most to gearbox lifespan loss. Identifying this path is a deterministic search and comparison process. Specifically, the identification method is as follows: for M... STSD Each row of the matrix k (k = 1, ..., M) total The time series M of the energy dispersion index it represents was evaluated. STSD (k,1),M STSD (k,2),...,M STSD The growth trend of (k,N).

[0067] A direct and effective method is to calculate the overall gradient of the sequence. Linear regression can be performed on the data points of each row to obtain a fitted straight line y(t) = a. k ·t+b k Where the slope a k This represents the average growth rate of the degradation index under the k-th operating condition.

[0068] Therefore, the algorithm for identifying the main degradation path can be described as follows:

[0069] For M STSD Each row k∈1,...,M total :

[0070] a. Extracting time series data V k =[M STSD (k,1),...,M STSD (k,N)].

[0071] b. Obtain V k Gradient a that varies with time k .

[0072] In all the calculated gradient values In the gradient, select the one with the largest positive value and the corresponding row index k. pdp That is what is sought, that is

[0073] Index k pdp The line was identified as the "main degradation path." The corresponding operating condition range represents the condition that contributes the most to the lifespan loss of the gearbox. The energy dispersion index sequence M along this path... STSD (k pdp ,1),...,M STSD (k pdp The inputs (N) constitute the most critical inputs for subsequent lifetime prediction.

[0074] S600. Based on the energy dispersion index sequence on the main degradation path and combined with a preset failure threshold, predict the remaining service life of the gearbox.

[0075] After identifying the main degradation path, the data sequence that best represents the degradation trend is used for extrapolation to predict the time when it will reach the failure state, thereby obtaining the remaining useful life (RUL).

[0076] This process requires defining a failure threshold I. fail This threshold is the energy dispersion index I. EDF A critical value is defined as the value at which the actual performance of the gearbox reaches or exceeds, indicating that it has failed or requires immediate repair. fail The determination can be based on historical failure data of this gearbox model, expert experience, or industry standards.

[0077] For example, through full life cycle testing of similar gearboxes, it was found that before a serious failure (such as tooth breakage) occurred, its I... EDF The value will generally reach around 0.4, so I can be... fail =0.4 is set as the failure threshold. Once set, this value remains constant during the prediction process.

[0078] For the main degenerate path k pdp Energy dispersion index sequence V pdp =[M STSD (k pdp ,1),...,M STSD (k pdp Extrapolate the trend using [N]. This means that a functional model needs to be built to describe the evolution of the sequence and used to predict future values. A direct approach is to continue using the linear model obtained from S500. Assume the fitted line of the main degradation path is I... EDF (t)=a pdp ·t+b pdp , where t is the index of the time segment. The predictive metric reaches I. fail Time point t fail That is, solving equation I fail =a pdp ·t fail +b pdp Solving for t, we get t. fail =(I fail -b pdp ) / a pdp .

[0079] This t fail This refers to the predicted failure time, represented by the time segment number. If each time segment represents H hours of runtime, then the predicted Total Useful Life (TUL) is TUL = t fail• H hours. The current time segment index is N, and the running time is T. current = N·H. Therefore, the predicted remaining useful life RUL is RUL = TUL - T current =(t fail -N)·H.

[0080] In some cases, the degradation process of gears may exhibit nonlinear characteristics, such as slow initial degradation, accelerated degradation in the middle stage, and rapid degradation at the end. In such cases, other deterministic function models can be used for extrapolation, such as polynomial regression or exponential function fitting. If quadratic polynomial fitting is used, the model is I... EDF (t)=c2t 2 If +c1t+c0, then a quadratic equation needs to be solved. To get t fail Regardless of the model used, its selection and parameters are calculated deterministically based on existing data, ensuring the transparency and reproducibility of the entire prediction process.

[0081] This embodiment also provides a gearbox life prediction system based on a feature matrix, which is the hardware implementation of the method described in Embodiment 1. Through modular design, this system solidifies each step of the method into functionally defined hardware or software modules, enabling automated, online prediction of gearbox life.

[0082] like Figure 2 As shown, the system includes a data acquisition module, a working condition classification module, an index quantification module, a matrix construction module, a path identification module, and a lifespan prediction module.

[0083] The data acquisition module can be physically implemented as an integrated data acquisition box, containing signal conditioning circuitry, an analog-to-digital converter (ADC), and local storage. This module is physically connected to various sensors mounted on the gearbox, such as accelerometers, thermocouples, motor encoders, and current transformers. Its function is to perform the operation described in S100, namely, to continuously acquire vibration, temperature, rotational speed, and load data according to a preset sampling rate and accuracy, assign a uniform timestamp to the data, and send the data stream to subsequent processing modules via an internal bus or network interface.

[0084] The operating condition segmentation module can be implemented by a software program in an embedded processor or a host computer. This module receives speed and load data streams, as well as timestamp information, from the data acquisition module. Internally, it stores the operating condition grid parameters and time segment length definitions defined in S200. This module runs continuously, dividing the continuous time axis into time segments and determining the dominant operating condition interval for each time segment in real time, ultimately outputting a discretized time segment index and the corresponding dominant operating condition index.

[0085] The index quantization module can be implemented by software on the processor. It receives raw vibration signal data within a specific time segment. Internally, this module stores a signal processing algorithm that executes the energy dispersion index I defined in S300. EDF The calculation process involves receiving a vibration signal, performing spectral analysis, identifying the main harmonics, calculating the relevant energy ratios, and finally outputting a single I0 signal. EDF Scalar value.

[0086] The matrix construction module can be a memory management and data structure maintenance program. It maintains a large two-dimensional array, namely the spacetime state degradation matrix M, in system memory. STSD This module receives the time segment index and dominant operating condition index from the operating condition segmentation module, and the I from the index quantification module. EDF Value. According to the S400 rules, I... EDF The value should be filled precisely into M. STSD The specified position of the matrix, and optionally the state inertia filling algorithm is executed.

[0087] The path identification module will periodically, or upon receiving an instruction, update the entire M maintained by the matrix construction module. STSD The matrix undergoes a global analysis. Internally, it incorporates the path identification algorithm described in S500. For example, it iterates through each row of the matrix, calculates the linear regression slope, compares all slopes, finds the index of the row with the largest value, and outputs it as the identifier of the main degenerate path.

[0088] The lifetime prediction module receives the main degradation path identifier from the path identification module, and from M STSD Extract the complete data sequence corresponding to the path from the matrix. The module internally stores a failure threshold I. fail It also incorporates the trend extrapolation algorithm described in S600. It performs extrapolation calculations to determine the failure time t. fail Combined with the current time N, the final remaining useful life (RUL) value is calculated. This value can be displayed on the human-machine interface or sent as an alarm signal to the equipment maintenance system.

[0089] In some application scenarios, to further improve the accuracy of predictions, this system may also include a lifetime correction module. This module is designed to take into account the accelerating effect of temperature on material aging and lubrication failure. Specifically, the lifetime correction module will construct another module in parallel with M... STSD Thermal stress matrix M of identical size TS The filler values ​​of this matrix are no longer vibration indices, but rather cumulative thermal stress values ​​derived from temperature data.

[0090] For example, the thermal stress value can be obtained by multiplying the operating time within a certain operating condition range by a weighting factor related to the average temperature of that operating condition. After the life prediction module calculates the preliminary RUL, the life correction module will search for the main degradation path k. pdp In the thermal stress matrix M TS The corresponding cumulative thermal stress value. If this value exceeds a certain threshold, the initial RUL will be reduced according to a preset correction function to obtain a more conservative and realistic final predicted life.

[0091] In some embodiments, when analyzing data for each time segment, the index quantification module, in addition to calculating I... EDF In addition, the load data will be analyzed to calculate a load instability factor F. LI This factor can be defined as the ratio of the standard deviation of the load signal to its mean within that time segment, used to quantify the severity of load fluctuations. Then, when I... EDF Before filling in the matrix, adjust it, for example by I' EDF =I EDF ·(1+λ·F LI The gain coefficient λ is adjusted to a variable gain coefficient. This amplifies the degradation indicators corresponding to periods of severe load fluctuations, thus reflecting their harmfulness earlier in the matrix. This allows the identification of the main degradation path and the final lifetime prediction to more accurately reflect the additional damage caused by the impact load.

[0092] Those skilled in the art should understand that the embodiments described above are merely exemplary embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

[0093] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0094] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0095] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Depending on actual needs, some or all of the units can be selected to achieve the purpose of this embodiment.

[0096] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware.

[0097] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A gearbox life prediction method based on feature matrix, characterized in that, include: Acquire multi-source operating data of the gearbox within a preset operating cycle; Based on the operating condition parameters in the multi-source operating data, the preset operating cycle is divided into multiple time segments, and the operating condition of the gearbox is divided into multiple operating condition intervals. For each time segment, the vibration signal in the multi-source operation data is extracted, and the energy dispersion index corresponding to the vibration signal is obtained. Construct a spatiotemporal state degradation matrix, wherein the rows of the spatiotemporal state degradation matrix correspond to the operating condition intervals and the columns correspond to the time segments; The energy dispersion index obtained in each time segment is filled into the matrix cell of the spatiotemporal state degradation matrix that corresponds to the time segment and the main operating condition of the time segment. In the spatiotemporal state degradation matrix, the row where the index value shows the greatest growth trend over time is identified as the main degradation path; Based on the energy dispersion index sequence along the main degradation path and a preset failure threshold, the remaining service life of the gearbox is predicted.

2. The method according to claim 1, characterized in that, The acquisition of multi-source operating data of the gearbox within a preset operating cycle includes: The vibration signal data, temperature data, speed data, and load data of the gearbox are acquired.

3. The method according to claim 1, characterized in that, The process of dividing the operating conditions of the gearbox into multiple operating condition ranges includes: Based on the speed data and the load data, a one- or two-dimensional operating condition grid is defined; Each grid cell of the two-dimensional working condition grid is defined as a working condition interval.

4. The method according to claim 1 or 3, characterized in that, After the step of filling the spatiotemporal state degradation matrix with the energy dispersion index obtained in each time segment, the method further includes: For the unfilled matrix elements in the spatiotemporal state degradation matrix, state inertia filling is performed based on the index values ​​of the matrix elements corresponding to the adjacent previous time segment.

5. The method according to claim 1, characterized in that, The step of obtaining the energy dispersion index corresponding to the vibration signal includes: The vibration signal within the time segment is subjected to spectral analysis to obtain the energy spectrum; Identify the gear meshing frequency and the energy peak corresponding to its harmonics in the energy spectrum; The energy dispersion index is obtained based on the ratio of the frequency band energy other than the energy peak in the energy spectrum to the total energy of the energy peak.

6. The method according to claim 1, characterized in that, The identification of the primary degradation path includes: For each row of the spatiotemporal state degradation matrix, the gradient of its energy dispersion index value as a function of time is obtained; The row with the largest positive gradient value is selected as the main degradation path.

7. The method according to claim 1, characterized in that, The prediction of the remaining service life of the gearbox based on the energy dispersion index sequence along the main degradation path includes: By extrapolating the energy dispersion index sequence along the main degradation path, a degradation trend curve is obtained. Determine the time point at which the degradation trend curve reaches the preset failure threshold; Based on the aforementioned time point and the current time, the remaining service life of the gearbox is obtained.

8. A gearbox life prediction system based on a feature matrix, characterized in that, include: The data acquisition module is used to acquire multi-source operating data of the gearbox within a preset operating cycle; The operating condition division module is used to divide the preset operating cycle into multiple time segments and the operating condition of the gearbox into multiple operating condition intervals based on the operating condition parameters in the multi-source operating data. The index quantification module is used to extract vibration signals from the multi-source operation data for each time segment and obtain the energy dispersion index corresponding to the vibration signals. A matrix construction module is used to construct a spatiotemporal state degradation matrix. The rows of the spatiotemporal state degradation matrix correspond to the operating condition intervals, and the columns correspond to the time segments. The energy dispersion index obtained in each time segment is filled into the matrix cell of the spatiotemporal state degradation matrix that corresponds to the time segment and the operating condition interval of the main operating condition in which the time segment is located. The path identification module is used to identify the row in the spatiotemporal state degradation matrix where the index value shows the greatest growth trend over time, as the main degradation path; The life prediction module is used to predict the remaining service life of the gearbox based on the energy dispersion index sequence on the main degradation path and in combination with a preset failure threshold.

9. The system according to claim 8, characterized in that, Also includes: Lifetime correction module, used for: The temperature data of the gearbox is obtained, and a thermal stress matrix is ​​constructed. The remaining service life is corrected based on the thermal stress value corresponding to the main degradation path in the thermal stress matrix.

10. The system according to claim 8, characterized in that, The indicator quantification module is also used for: Obtain the load data of the gearbox and the load instability factor within each time segment; Based on the load instability factor, the energy dispersion index is adjusted to reflect the impact of impact load on gearbox degradation.

Images