A method and system for predicting hydraulic pump life

Abstract

The present invention discloses a method and system for predicting the life of a hydraulic pump, which relates to the technical field of hydraulic pump life prediction, including collecting vibration signals, pressure signals and temperature signals of the hydraulic pump under different working conditions, and forming original monitoring data; performing wavelet decomposition on the original monitoring data, extracting high-frequency coefficients and low-frequency coefficients of the original monitoring data, and constructing a feature vector based on the high-frequency coefficients and low-frequency coefficients; inputting the feature vector into a pre-trained health estimation model to obtain a health index of the hydraulic pump; establishing a Weibull distribution model based on the health index, and calculating the scale parameter and shape parameter of the Weibull distribution model by the maximum likelihood estimation method; and calculating the remaining service life of the hydraulic pump according to the scale parameter and shape parameter. The present invention combines the health index with the Weibull distribution model, and uses the maximum likelihood estimation method to accurately calculate the scale parameter and shape parameter, thereby predicting the remaining service life of the hydraulic pump.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment health management, and in particular to a method and system for predicting the life of a hydraulic pump. Background Art

[0002] As the core component of the hydraulic system, the performance and life of the hydraulic pump have a vital impact on the reliability and working efficiency of the entire hydraulic system. In industrial equipment, engineering machinery, and aerospace fields, hydraulic pumps are widely used in power transmission and energy conversion. In recent years, with the rapid development of intelligent manufacturing and industrial automation, the frequency of use and load complexity of hydraulic pumps have increased significantly, and higher requirements have been placed on their life prediction technology. Hydraulic pump life prediction is an important research direction in the field of equipment health management. Its core goal is to provide a basis for equipment maintenance decisions, reduce unplanned downtime, and reduce maintenance costs by monitoring the operating status data of the hydraulic pump, analyzing its degradation trend, and predicting its remaining service life.

[0003] At present, the life prediction technology of hydraulic pumps is mainly based on methods such as vibration analysis, oil monitoring, and multi-physics field data modeling. Among them, vibration signal analysis is widely used in the field of fault diagnosis and life prediction of hydraulic pumps. Through feature extraction in the time domain, frequency domain, and time-frequency domain, it can effectively identify the degradation mode of the internal components of the hydraulic pump. However, traditional methods often rely on the manual extraction of specific signal features, and have limited ability to process multidimensional signals under complex working conditions. In addition, hydraulic pumps are affected by many factors (such as pressure fluctuations and temperature changes) during long-term operation, and their performance degradation is nonlinear and random. A single feature extraction or life prediction model is difficult to fully reflect the true state of the equipment. At the same time, due to the high dimensionality and complexity of hydraulic pump operation data, existing research often faces problems such as high computational cost and insufficient prediction accuracy in feature extraction and model construction. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention proposes a hydraulic pump life prediction method and system.

[0005] Therefore, the present invention provides a method for predicting the life of a hydraulic pump, which can solve the problems mentioned in the background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method for predicting the life of a hydraulic pump, which comprises:

[0008] Collecting vibration signals, pressure signals and temperature signals of the hydraulic pump under different working conditions, and combining the vibration signals, pressure signals and temperature signals into original monitoring data;

[0009] Performing wavelet decomposition on the original monitoring data, extracting high-frequency coefficients and low-frequency coefficients of the original monitoring data, and constructing a feature vector based on the high-frequency coefficients and low-frequency coefficients;

[0010] Inputting the feature vector into a pre-trained health estimation model to obtain a health index of the hydraulic pump;

[0011] Establishing a Weibull distribution model based on the health index, and calculating the scale parameter and shape parameter of the Weibull distribution model by maximum likelihood estimation method;

[0012] The remaining service life of the hydraulic pump is calculated based on the dimension parameters and the shape parameters.

[0013] As a preferred solution of the hydraulic pump life prediction method of the present invention, the construction of the characteristic vector includes:

[0014] Calculating an adaptive weight coefficient based on the high frequency coefficient and the low frequency coefficient;

[0015] Calculating characteristic parameters for the high-frequency coefficient and the low-frequency coefficient respectively; the characteristic parameters include weighted mean, combined standard deviation, composite kurtosis, enhanced skewness, modified root mean square value and waveform factor;

[0016] The characteristic parameters form an initial characteristic matrix, and feature dimension reduction is performed through a principal component analysis method to obtain the characteristic vector.

[0017] As a preferred solution of the hydraulic pump life prediction method of the present invention, the calculation of the adaptive weight coefficient is shown in the following formula:

[0018] ,

[0019] in, and is the regulating factor, For the The standard deviation of the layer wavelet coefficients, For the The energy value of the layer, is the total energy value, For the Wavelet detail coefficients.

[0020] As a preferred solution of the hydraulic pump life prediction method of the present invention, the weighted mean is calculated as follows:

[0021] ,

[0022] The calculation of the combined standard deviation is shown in the following formula:

[0023] ,

[0024] The calculation of the composite kurtosis is shown in the following formula:

[0025] ,

[0026] The calculation of the enhanced skewness is shown below:

[0027] ,

[0028] The calculation of the modified RMS value is shown in the following formula:

[0029] ,

[0030] The calculation of the shape factor is shown below:

[0031] ,

[0032] in, is the signal strength weight, is the number of sampling points, is the Hilbert transform, and is the correction factor, is the fast Fourier transform, is the signal-to-noise ratio, and is the gain coefficient, is the weighted mean, is the combined standard deviation, is the composite kurtosis, is the enhanced skewness, is the corrected RMS value, is the shape factor.

[0033] As a preferred solution of the hydraulic pump life prediction method of the present invention, wherein: the health estimation model includes an input layer, a multi-scale feature extraction layer, a bidirectional LSTM layer, an attention layer and an output layer;

[0034] The multi-scale feature extraction layer is expressed as follows:

[0035] ,

[0036] in, Indicates the scale level, For the time step The input feature vector, is the scale weight matrix, is the bias vector; is the scale activation function, and its specific formula is:

[0037] ,

[0038] in, is the adaptive coefficient, is the hyperbolic tangent function, is the sigmoid function, is the linear combination output of the current neuron.

[0039] As a preferred solution of the hydraulic pump life prediction method of the present invention, wherein: the bidirectional LSTM layer includes a forward propagation LSTM and a backward propagation LSTM;

[0040] The state update of the bidirectional LSTM layer is shown as follows:

[0041] ,

[0042] ,

[0043] ,

[0044] ,

[0045] ,

[0046] in, , , They are input gate, forget gate and output gate respectively. is the memory unit state, represents the Hadamard product, σ is the Sigmoid activation function, is the input gate weight matrix, is the hidden state at the previous moment, is the input at the current moment, is the recurrent weight matrix of the input gate, is the memory unit state at the previous moment, is the input gate bias term, is the Sigmoid activation function, is the forget gate weight matrix, is the recurrent weight matrix of the forget gate, is the forget gate bias term, is the candidate information weight matrix, is the candidate information bias term, is the output gate weight matrix, is the recurrent weight matrix of the output gate, is the output gate bias term, For LSTM units at time steps The hidden state output of .

[0047] As a preferred solution of the hydraulic pump life prediction method of the present invention, the calculation of the remaining service life of the hydraulic pump includes:

[0048] The health index obtained by the health estimation model is used as a basic data source to perform modeling through a probability density function of a Weibull distribution model;

[0049] According to the modeled Weibull distribution model, construct the likelihood function and derive it;

[0050] The derived likelihood function is approximated to the optimal parameters by numerical iteration.

[0051] Substituting the optimal parameters back into the derived likelihood function to obtain the scale parameter and shape parameter;

[0052] The remaining service life of the hydraulic pump is calculated according to the scale parameter, the shape parameter and the health index.

[0053] As a preferred solution of the hydraulic pump life prediction method of the present invention, the remaining service life of the hydraulic pump is calculated as shown in the following formula:

[0054] ,

[0055] in, is the scale parameter, is the probability correction factor, Represents the elapsed time. is the probability decay function of the health index, is the dynamic adjustment coefficient, is the degradation rate correction factor, is the remaining useful life, is the shape parameter.

[0056] In a second aspect, an embodiment of the present invention provides a hydraulic pump life prediction system, which includes:

[0057] A data acquisition module is used to collect vibration signals, pressure signals and temperature signals of the hydraulic pump under different working conditions, and to form original monitoring data from the vibration signals, pressure signals and temperature signals;

[0058] A feature vector construction module, used to perform wavelet decomposition on the original monitoring data, extract high-frequency coefficients and low-frequency coefficients of the original monitoring data, and construct a feature vector based on the high-frequency coefficients and low-frequency coefficients;

[0059] A model building module, used for inputting the feature vector into a pre-trained health estimation model to obtain a health index of the hydraulic pump;

[0060] A parameter solving module, used for establishing a Weibull distribution model based on the health index, and calculating the scale parameter and shape parameter of the Weibull distribution model by maximum likelihood estimation method;

[0061] The life calculation module is used to calculate the remaining service life of the hydraulic pump according to the scale parameter and the shape parameter.

[0062] The beneficial effects of the present invention are that by collecting multi-source data (vibration signals, pressure signals and temperature signals) of the hydraulic pump under different working conditions, and extracting high-frequency and low-frequency features using the wavelet decomposition method, the transient impact characteristics and steady-state operation characteristics of the hydraulic pump during operation are fully captured; combined with improved characteristic parameter calculation and dimensionality reduction technology, optimized feature vectors are generated, which greatly improves the efficiency of data processing and the accuracy of feature expression; through the combination of a bidirectional multi-scale long short-term memory neural network model and an attention mechanism, the health index of the hydraulic pump is accurately evaluated, and the ability to extract key degradation information in time series is significantly enhanced; based on the health index, a Weibull distribution model is established, and parameter calculation is optimized by the maximum likelihood estimation method to achieve high-precision prediction of the remaining life of the hydraulic pump. This method effectively integrates multi-source data analysis and deep learning technology, has both robustness and practicality, and can be widely used in the fields of equipment health management and intelligent maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0064] Figure 1 The figure is a flow chart of a method for predicting the life of a hydraulic pump. DETAILED DESCRIPTION

[0065] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0066] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0067] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0068] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0069] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0070] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0071] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides a hydraulic pump life prediction method, comprising:

[0072] S1: collecting vibration signals, pressure signals and temperature signals of the hydraulic pump under different working conditions, and forming the vibration signals, pressure signals and temperature signals into original monitoring data;

[0073] A sensor array is installed at the oil inlet, oil outlet and pump body surface of the hydraulic pump, respectively, wherein the sensor array includes a vibration sensor, a pressure sensor and a temperature sensor; the radial vibration signal and the axial vibration signal of the hydraulic pump are collected by the vibration sensor when the hydraulic pump is working, and the sampling frequency of the vibration signal is 20kHz; the pressure signal of the oil inlet and outlet of the hydraulic pump is collected by the pressure sensor, and the sampling frequency of the pressure signal is 10kHz; the temperature signal of each monitoring point of the hydraulic pump is collected by the temperature sensor, and the sampling frequency of the temperature signal is 1kHz; the vibration signal, pressure signal and temperature signal are synchronized according to the timestamp to form original monitoring data including vibration characteristics, pressure characteristics and temperature characteristics; wherein the working conditions of the hydraulic pump include 25% load, 50% load, 75% load and 100% load at rated speed.

[0074] S2: performing wavelet decomposition on the original monitoring data, extracting high-frequency coefficients and low-frequency coefficients of the original monitoring data, and constructing a feature vector based on the high-frequency coefficients and low-frequency coefficients;

[0075] The original monitoring data is subjected to a 5-layer wavelet decomposition using the db4 wavelet basis to obtain 5-scale wavelet detail coefficients d1, d2, d3, d4, d5 and an approximate coefficient a5, wherein d1 and d2 are high-frequency coefficients that characterize the transient impact characteristics of the hydraulic pump, and d3, d4, d5 and a5 are low-frequency coefficients that characterize the steady-state operation characteristics of the hydraulic pump; based on the high-frequency coefficients and the low-frequency coefficients, the adaptive weight coefficient λi is calculated, and the calculation formula of λi is:

[0076] ,

[0077] in, and is the regulating factor, is the standard deviation of the wavelet coefficients at the i-th layer, For the The energy value of the layer, is the total energy value, For the Wavelet detail coefficients;

[0078] Calculate improved characteristic parameters for the high frequency coefficients and the low frequency coefficients, including weighted mean , combined standard deviation , composite kurtosis , enhanced skewness , Corrected RMS and improved form factor , and its calculation formula is:

[0079] ,

[0080] ,

[0081] in, is the signal strength weight, is the number of sampling points;

[0082] ,

[0083] ,

[0084] in, is the Hilbert transform, and is the correction factor;

[0085] ,

[0086] ,

[0087] in, is the fast Fourier transform, is the signal-to-noise ratio, and is the gain factor.

[0088] Furthermore, the characteristic parameters are combined into an initial characteristic matrix F, and the characteristic dimension reduction is performed by the principal component analysis method:

[0089] ,

[0090] in, is a weighted matrix, which is determined by the contribution of each characteristic parameter. is the eigenvector matrix; finally, a 16-dimensional optimized eigenvector is obtained.

[0091] S3: inputting the feature vector into a pre-trained long short-term memory neural network model to obtain a health index of the hydraulic pump, wherein the health index represents a degree of performance degradation of the hydraulic pump;

[0092] A bidirectional multi-scale long short-term memory neural network model is constructed, wherein the neural network model includes an input layer, a multi-scale feature extraction layer, a bidirectional LSTM layer, an attention layer and an output layer; the feature vector is divided into data windows of length T according to the time series, where T=128.

[0093] More specifically, the calculation formula of the multi-scale feature extraction layer is:

[0094] ,

[0095] in, represents the scale level (l=1,2,3), For the time step The input feature vector, is the scale weight matrix, is the bias vector; is the scale activation function, and its specific formula is:

[0096] ,

[0097] in, is the adaptive coefficient, is the hyperbolic tangent function, is the sigmoid function, is the linear combination output of the current neuron;

[0098] The bidirectional LSTM layer further enhances the model’s ability to capture time dependencies. This layer consists of two LSTM networks, the forward and reverse, and can simultaneously learn the forward and reverse patterns of the time series. This gating mechanism enables the model to selectively memorize or forget historical information, thereby better understanding the evolution of the hydraulic pump state. The state update formula is:

[0099] ,

[0100] ,

[0101] ,

[0102] ,

[0103] ,

[0104] in, , , They are input gate, forget gate and output gate respectively. is the memory unit state, represents the Hadamard product, σ is the Sigmoid activation function, is the input gate weight matrix, is the hidden state at the previous moment, is the input at the current moment, is the recurrent weight matrix of the input gate, is the memory unit state at the previous moment, is the input gate bias term, is the Sigmoid activation function, is the forget gate weight matrix, is the hidden state at the previous moment, is the recurrent weight matrix of the forget gate, is the memory unit state at the previous moment, is the forget gate bias term, is the candidate information weight matrix, is the candidate information bias term, is the output gate weight matrix, is the recurrent weight matrix of the output gate, is the output gate bias term, For LSTM units at time steps The hidden state output of .

[0105] The attention layer introduces a multi-head self-attention mechanism to further enhance the model's ability to extract key features. Different attention weights are assigned to different time steps by calculating the query matrix Q, key matrix K, and value matrix V. This mechanism enables the model to adaptively focus on the features that have the greatest impact on the health status of the hydraulic pump, greatly improving the accuracy of the prediction.

[0106] At the output layer, the final health index is calculated by the Mish activation function, as shown in the following formula:

[0107] ,

[0108] in, Mish activation function:

[0109] ,

[0110] ,

[0111] in, is the health index weight matrix, is the attention weight of the forward LSTM, is the attention weight of the reverse LSTM, is the health index bias term.

[0112] The health index has a value range of [0,1], where 0 indicates that the hydraulic pump has completely failed, and 1 indicates that the hydraulic pump is in good condition.

[0113] S4: establishing a Weibull distribution model based on the health index, and calculating the scale parameter and shape parameter of the Weibull distribution model by maximum likelihood estimation method;

[0114] First, the health index time series obtained by the long short-term memory neural network model is used as the basic data source to represent the dynamic evolution of the equipment from the initial state to performance degradation. The probability density function of the Weibull distribution model is further used to accurately describe the failure characteristics of mechanical equipment, as shown in the following formula:

[0115] ,

[0116] in, (Size parameter) reflects the characteristic life of the equipment, (Shape parameters) reveal the nature of the failure mode.

[0117] Constructing the likelihood function by maximum likelihood estimation Then the log-likelihood function Derivative, the complex nonlinear parameter estimation problem is transformed into solving the log-likelihood function, which is simplified to By using the numerical iteration method, the optimal parameters are approached through repeated iterations, and strict convergence criteria are introduced (such as the error is less than 10^-6), and finally the estimated values ​​of η and β with high precision are obtained.

[0118] S5: Calculating the remaining service life of the hydraulic pump according to the scale parameter and the shape parameter.

[0119] The scale parameter η (eta) and shape parameter β (beta) obtained by maximum likelihood estimation in S4 are used. The scale parameter and shape parameter respectively characterize the concentration degree and failure mechanism of the equipment failure time; the remaining service life of the hydraulic pump is calculated by determining the current state in combination with the health index obtained according to S3, as shown in the following formula:

[0120] ,

[0121] in, is the scale parameter, is the probability correction factor, Represents the elapsed time. is the probability decay function of the health index, is the dynamic adjustment coefficient, is the degradation rate correction factor, is the remaining useful life, is the shape parameter.

[0122] Furthermore, this embodiment also provides a hydraulic pump life prediction system, comprising:

[0123] A data acquisition module is used to collect vibration signals, pressure signals and temperature signals of the hydraulic pump under different working conditions, and to form original monitoring data from the vibration signals, pressure signals and temperature signals;

[0124] A feature vector construction module, used to perform wavelet decomposition on the original monitoring data, extract high-frequency coefficients and low-frequency coefficients of the original monitoring data, and construct a feature vector based on the high-frequency coefficients and low-frequency coefficients;

[0125] A model building module, used for inputting the feature vector into a pre-trained health estimation model to obtain a health index of the hydraulic pump;

[0126] A parameter solving module, used for establishing a Weibull distribution model based on the health index, and calculating the scale parameter and shape parameter of the Weibull distribution model by maximum likelihood estimation method;

[0127] The life calculation module is used to calculate the remaining service life of the hydraulic pump according to the scale parameter and the shape parameter.

[0128] This embodiment also provides a computer device, which is suitable for the hydraulic pump life prediction method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the hydraulic pump life prediction method proposed in the above embodiment.

[0129] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0130] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for predicting the life of a hydraulic pump as proposed in the above embodiment is implemented.

[0131] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0132] Embodiment 2 is the second embodiment of the present invention. This embodiment provides a method for predicting the life of a hydraulic pump. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0133] This experiment selected a certain type of hydraulic pump for construction machinery as the research object, aiming to verify the effectiveness of the hydraulic pump health status assessment and life prediction method based on multi-signal fusion and deep learning. The experiment was carried out on a professional test bench in a hydraulic equipment research center, which was equipped with a high-precision data acquisition system and a multi-dimensional sensor array.

[0134] Before the experiment, the hydraulic pump was fully tested and calibrated. The engineering hydraulic pump model HD-P650A was selected, with a rated power of 120kW and a rated speed of 2500r / min. The sensor array was precisely installed on the oil inlet, oil outlet and pump body surface of the hydraulic pump. The sensor array consists of three types of high-sensitivity sensors: (1) piezoelectric vibration sensor with a measurement range of ±50m / s² and a frequency response of 0-20kHz; (2) pressure sensor with a measurement range of 0-40MPa and an accuracy level of 0.5; (3) thermocouple temperature sensor with a measurement range of -50-300℃ and an accuracy of ±0.5℃.

[0135] The data collection process is strictly carried out according to the preset working conditions. Through the load adjustment system of the hydraulic pump test bench, steady-state working condition tests are carried out at 25%, 50%, 75% and 100% load respectively. Under each working condition, monitoring data is continuously collected for 300 seconds. The sampling frequency of vibration signal is 20kHz, the sampling frequency of pressure signal is 10kHz, and the sampling frequency of temperature signal is 1kHz. During the data collection process, the time synchronization between sensors is ensured, and high-precision timestamp technology is used to control the error in microseconds.

[0136] In the data preprocessing stage, the db4 wavelet base was used to perform 5-layer wavelet decomposition to extract high-frequency coefficients (d1, d2) and low-frequency coefficients (d3, d4, d5, a5). By calculating the adaptive weight coefficient, the improved feature parameters were extracted, including weighted mean, combined standard deviation, composite kurtosis, etc. Then, the principal component analysis method was used to reduce the dimension of the feature vector to 16 dimensions, providing high-quality input for the subsequent deep learning model.

[0137] At the end of the experiment, the data shown in Table 1 are obtained:

[0138] Table 1 Hydraulic pump health status monitoring experimental data

[0139] Working conditions Vibration intensity (m / s²) Pressure fluctuation(MPa) Temperature change (℃) Health Index Failure probability (%) Estimated remaining life (hours) 25% load 2.3 12.5 38.6 0.92 3.5 8760 50% load 3.7 22.1 45.2 0.85 6.8 6540 75% load 5.6 32.4 52.7 0.73 12.5 4320 100% load 8.2 39.6 61.3 0.58 24.6 2160 Overload operation 12.5 45.8 68.9 0.35 48.7 720 Restoration after repair 1.8 10.2 36.4 0.95 2.1 10920

[0140] Through in-depth analysis of the experimental data, the performance change characteristics of the hydraulic pump under different working conditions can be clearly observed. As the load increases, the vibration intensity, pressure fluctuations and temperature changes show a significant increasing trend, which reflects the performance degradation law of the equipment under high-intensity working conditions.

[0141] The health index dropped rapidly from 0.92 at 25% load to 0.58 at 100% load, a drop of 37%, which fully demonstrates the sensitivity and accuracy of the method of the present invention to the equipment status. Compared with the traditional empirical judgment and single parameter monitoring method, this method has the following significant advantages:

[0142] By simultaneously collecting vibration, pressure and temperature signals, a comprehensive equipment health profile is constructed, avoiding the limitations of single sensor monitoring.

[0143] Wavelet decomposition and principal component analysis techniques are used to effectively remove noise interference, extract key features, and significantly improve the accuracy of health status assessment. Based on the Weibull distribution model, the remaining service life of the equipment can be accurately predicted. Experimental data show that under 100% load conditions, the estimated remaining service life is only 2160 hours, while under 25% load conditions, it can reach 8760 hours.

[0144] The experimental results verify the high reliability and practicality of this method in engineering practice. Compared with the traditional regular maintenance mode, this innovative method can significantly reduce maintenance costs, improve equipment utilization, and provide strong support for intelligent manufacturing and Industry 4.0.

[0145] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting the life of a hydraulic pump, characterized in that: include: Collecting vibration signals, pressure signals and temperature signals of the hydraulic pump under different working conditions, and combining the vibration signals, pressure signals and temperature signals into original monitoring data; Performing wavelet decomposition on the original monitoring data, extracting high-frequency coefficients and low-frequency coefficients of the original monitoring data, and constructing a feature vector based on the high-frequency coefficients and low-frequency coefficients; Inputting the feature vector into a pre-trained health estimation model to obtain a health index of the hydraulic pump; Establishing a Weibull distribution model based on the health index, and calculating the scale parameter and shape parameter of the Weibull distribution model by maximum likelihood estimation method; Calculating the remaining service life of the hydraulic pump according to the dimensional parameters and the shape parameters; The health estimation model includes an input layer, a multi-scale feature extraction layer, a bidirectional LSTM layer, an attention layer and an output layer; The multi-scale feature extraction layer is expressed as follows: ， in, Indicates the scale level, For the time step The input feature vector, is the scale weight matrix, is the bias vector; is the scale activation function, and its specific formula is: ， in, is the adaptive coefficient, is the hyperbolic tangent function, is the sigmoid function, is the linear combination output of the current neuron; The bidirectional LSTM layer includes a forward propagation LSTM and a backward propagation LSTM; The state update of the bidirectional LSTM layer is shown as follows: ， ， ， ， ， in, , , They are input gate, forget gate and output gate respectively. is the memory unit state, represents the Hadamard product, σ is the Sigmoid activation function, is the input gate weight matrix, is the hidden state at the previous moment, is the input at the current moment, is the recurrent weight matrix of the input gate, is the memory unit state at the previous moment, is the input gate bias term, is the Sigmoid activation function, is the forget gate weight matrix, is the recurrent weight matrix of the forget gate, is the forget gate bias term, is the candidate information weight matrix, is the candidate information bias term, is the output gate weight matrix, is the recurrent weight matrix of the output gate, is the output gate bias term, For LSTM units at time steps The hidden state output of .

2. The hydraulic pump life prediction method according to claim 1, characterized in that: The construction of the feature vector includes: Calculating an adaptive weight coefficient based on the high frequency coefficient and the low frequency coefficient; Calculating characteristic parameters for the high-frequency coefficient and the low-frequency coefficient respectively; the characteristic parameters include weighted mean, combined standard deviation, composite kurtosis, enhanced skewness, modified root mean square value and waveform factor; The characteristic parameters form an initial characteristic matrix, and feature dimension reduction is performed through a principal component analysis method to obtain the characteristic vector.

3. The hydraulic pump life prediction method according to claim 2, characterized in that: The calculation of the adaptive weight coefficient is shown in the following formula: ， in, and is the regulating factor, For the The standard deviation of the layer wavelet coefficients, For the The energy value of the layer, is the total energy value, For the wavelet detail coefficients.

4. The hydraulic pump life prediction method according to claim 3, characterized in that: The weighted mean is calculated as follows: ， The calculation of the combined standard deviation is shown in the following formula: ， The calculation of the composite kurtosis is shown in the following formula: ， The calculation of the enhanced skewness is shown below: ， The calculation of the modified RMS value is shown in the following formula: ， The calculation of the shape factor is shown below: ， in, is the signal strength weight, is the number of sampling points, is the Hilbert transform, and is the correction factor, is the fast Fourier transform, is the signal-to-noise ratio, and is the gain coefficient, is the weighted mean, is the combined standard deviation, is the composite kurtosis, is the enhanced skewness, is the corrected RMS value, is the shape factor.

5. The hydraulic pump life prediction method according to claim 4, characterized in that: Calculating the remaining useful life of the hydraulic pump includes: The health index obtained by the health estimation model is used as a basic data source to perform modeling through a probability density function of a Weibull distribution model; According to the modeled Weibull distribution model, construct the likelihood function and derive it; The derived likelihood function is approximated to the optimal parameters by numerical iteration. Substituting the optimal parameters back into the derived likelihood function to obtain the scale parameter and shape parameter; The remaining service life of the hydraulic pump is calculated according to the scale parameter, the shape parameter and the health index.

6. The hydraulic pump life prediction method according to claim 5, characterized in that: The remaining service life of the hydraulic pump is calculated as follows: ， in, is the scale parameter, is the probability correction factor, Represents the elapsed time. is the probability decay function of the health index, is the dynamic adjustment coefficient, is the degradation rate correction factor, is the remaining useful life, is the shape parameter.

7. A hydraulic pump life prediction system, based on the hydraulic pump life prediction method according to any one of claims 1 to 6, characterized in that: include: A data acquisition module is used to collect vibration signals, pressure signals and temperature signals of the hydraulic pump under different working conditions, and to form original monitoring data from the vibration signals, pressure signals and temperature signals; A feature vector construction module, used to perform wavelet decomposition on the original monitoring data, extract high-frequency coefficients and low-frequency coefficients of the original monitoring data, and construct a feature vector based on the high-frequency coefficients and low-frequency coefficients; A model building module, used for inputting the feature vector into a pre-trained health estimation model to obtain a health index of the hydraulic pump; A parameter solving module, used for establishing a Weibull distribution model based on the health index, and calculating the scale parameter and shape parameter of the Weibull distribution model by maximum likelihood estimation method; The life calculation module is used to calculate the remaining service life of the hydraulic pump according to the scale parameter and the shape parameter.

Images