A method and device for constructing a hydraulic pump life prediction model

Abstract

The present application provides a method and device for constructing a hydraulic pump life prediction model. The method provided in the present application includes: obtaining segmented life test data and continuous life test data of the hydraulic pump; performing degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; wherein the first fusion feature is a feature corresponding to the segmented life test data, and the second fusion feature is a feature corresponding to the continuous life test data; constructing a first life prediction model according to the first fusion feature; constructing a second life prediction model according to the second fusion feature; fusing the first life prediction model and the second life prediction model to obtain a life prediction model. The method and device for constructing a hydraulic pump life prediction model provided in the present application can solve the problem of incomplete degradation data of the hydraulic pump throughout its life cycle and difficulty in constructing a hydraulic pump life prediction model.

Description

Technical Field

[0001] The present application relates to the field of aircraft electromechanical technology, and in particular to a method and device for constructing a hydraulic pump life prediction model. Background Art

[0002] The electromechanical system of an aircraft is vital to the aircraft. In order to ensure the safety of the aircraft during flight, the electromechanical system needs to be inspected and maintained frequently.

[0003] As one of the important components in the electromechanical system, the hydraulic pump needs to be regularly inspected and maintained. However, the electromechanical system has a complex structure, many components, and obvious dynamic characteristics. When inspecting the hydraulic pump, it is necessary to go through a tedious disassembly and assembly process. In addition, traditional inspection and maintenance requires manual troubleshooting, which is inefficient and complicated. With the development of technology, it is possible to establish a life prediction model to perform real-time fault prediction and health status assessment on the electromechanical structure, and realize an active intelligent maintenance mode.

[0004] However, the current airborne hydraulic pumps are still mainly mechanical structures, lack the ability to autonomously perceive their status, and lack full life cycle data on airborne hydraulic pumps for health assessment and testing needs. It is impossible to determine the full life cycle degradation characteristics of the hydraulic pumps, which makes it difficult to establish a life prediction model for the hydraulic pumps. Summary of the invention

[0005] In view of this, the present application provides a method, device and equipment for constructing a hydraulic pump life prediction model to solve the problem of incomplete degradation data of the hydraulic pump throughout its life cycle and difficulty in constructing a hydraulic pump life prediction model.

[0006] Specifically, the present application is implemented through the following technical solutions:

[0007] The first aspect of the present application provides a method for constructing a hydraulic pump life prediction model, the method comprising:

[0008] Obtain segmented life test data and continuous life test data of hydraulic pumps;

[0009] Performing degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; wherein the first fusion feature is a feature corresponding to the segmented life test data, and the second fusion feature is a feature corresponding to the continuous life test data;

[0010] constructing a first lifespan prediction model according to the first fusion feature;

[0011] constructing a second lifespan prediction model according to the second fusion feature;

[0012] The first life prediction model and the second life prediction model are integrated to obtain a life prediction model; wherein, the real-time health status of the hydraulic pump is identified based on the first life prediction model, and the remaining life of the hydraulic pump in the real-time health status is predicted based on the second life prediction model.

[0013] A second aspect of the present application provides a device for constructing a hydraulic pump life prediction model, the device comprising an acquisition module, an analysis module and a construction module; wherein:

[0014] The acquisition module is used to acquire segmented life test data and continuous life test data of the hydraulic pump;

[0015] The analysis module is used to perform degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; wherein the first fusion feature is a feature corresponding to the segmented life test data, and the second fusion feature is a feature corresponding to the continuous life test data;

[0016] The construction module is used to construct a first life prediction model according to the first fusion feature;

[0017] The construction module is further used to construct a second life prediction model according to the second fusion feature;

[0018] The construction module is also used to fuse the first life prediction model and the second life prediction model to obtain a life prediction model; wherein the real-time health status of the hydraulic pump is identified based on the first life prediction model, and the remaining life of the hydraulic pump in the real-time health status is predicted based on the second life prediction model.

[0019] The present application provides a method and device for constructing a hydraulic pump life prediction model. On the one hand, by obtaining the continuous life test data and segmented life test data of the hydraulic pump, the performance changes and degradation process of the hydraulic pump can be understood from different angles and time scales. Combining the two types of data can more comprehensively reflect the actual working conditions of the hydraulic pump and reduce the deviations and limitations that may be caused by a single data source. On the second hand, the first life prediction model and the second life prediction model are constructed based on the first fusion feature and the second fusion feature. Multiple indicators and factors can be integrated to more accurately describe the health status and degradation trend of the hydraulic pump, thereby improving the accuracy and reliability of the model's prediction of the hydraulic pump life. On the third hand, by comprehensively using the model established by continuous life test data and segmented life test data, the health of the hydraulic pump can be comprehensively evaluated. The method provided by the present invention takes into account that the remaining life is related to the changing states that need to be experienced, as well as the duration of each state, that is, including transient state information and long-term state information. A continuous life test and a segmented life test are designed to obtain data respectively, and models are established based on the two tests respectively to predict the transient state and long-term change information respectively. At the same time, the change of the transient state will affect the duration of each state. That is to say, while predicting the state and duration respectively, the duration is adjusted and predicted based on the state to achieve a more accurate prediction. Finally, the combined effect of the two models is combined to predict the working state and remaining life of the product in advance, so that maintenance time and resources can be reasonably arranged to support predictive maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flowchart of a method for constructing a hydraulic pump life prediction model according to an embodiment of the present invention;

[0021] Figure 2 A flowchart of a second embodiment of the method for constructing a hydraulic pump life prediction model provided in the present application;

[0022] Figure 3 A flowchart of a third embodiment of the method for constructing a hydraulic pump life prediction model provided in the present application;

[0023] Figure 4 A flowchart of a fourth embodiment of the method for constructing a hydraulic pump life prediction model provided in the present application;

[0024] Figure 5 This is a structural schematic diagram of embodiment 1 of the device for constructing a hydraulic pump life prediction model provided in the present application. DETAILED DESCRIPTION

[0025] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0026] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to and includes any or all possible combinations of one or more associated listed items.

[0027] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0028] Specific embodiments are given below to introduce the technical solution of the present application in detail.

[0029] Figure 1 This is a flowchart of the first embodiment of the method for constructing a hydraulic pump life prediction model provided by this application. Figure 1 , the method provided in this embodiment may include:

[0030] S101. Obtain segmented life test data and continuous life test data of a hydraulic pump.

[0031] Specifically, the segmented life test of the hydraulic pump refers to dividing the entire life cycle of the hydraulic pump into several stages, allowing the hydraulic pump in each stage to enter a specified time test under rated operating conditions, and recording the operating data of the hydraulic pump in each stage in real time. The collection of the operating data of the hydraulic pump in these several stages is the segmented life test data of the hydraulic pump. The segmented life test can obtain the short-term degradation characteristics of the hydraulic pump in different stages over time. It can be understood that in the segmented life test, the test operation time of the hydraulic pump in each stage is the same. It should be noted that the specified operation time of the hydraulic pump in the segmented life test is set according to actual needs, and in this embodiment, it is not limited. For example, in one embodiment, the specified time is 50 hours.

[0032] Furthermore, the continuous life test of the hydraulic pump refers to the process of making the hydraulic pump continue to operate under the rated operating conditions until the hydraulic pump fails. The hydraulic pump data from the start of the test to the occurrence of the failure is the continuous life test data. The continuous life test can describe the degradation state of the hydraulic pump over time through the complete working process of the hydraulic pump.

[0033] Furthermore, when the hydraulic pump is subjected to continuous life tests and segmented life tests, various data of the hydraulic pump during the test are obtained in real time through sensors pre-arranged on the hydraulic pump. The sensors pre-arranged on the hydraulic pump may be pressure sensors, temperature sensors, flow sensors, vibration sensors, speed sensors, etc., so as to obtain pressure pulsation, temperature, flow, vibration, speed and other signals of the hydraulic pump during operation through sensors.

[0034] S102. Perform degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; wherein the first fusion feature is a feature corresponding to the segmented life test data, and the second fusion feature is a feature corresponding to the continuous life test data.

[0035] Specifically, degradation characteristics refer to characteristic parameters exhibited by the hydraulic pump during use that can reflect the performance decline, aging, or approaching failure state of the hydraulic pump over time.

[0036] Furthermore, the specific implementation steps of performing degradation feature analysis on the segmented life test data to obtain the first fusion feature are:

[0037] (1) Extracting features from the segmented life test data to obtain first parameter features.

[0038] Specifically, in this step, feature extraction can be performed on the segmented life test data through time domain analysis, frequency domain analysis, time-frequency domain analysis, and statistical analysis to extract useful degradation features. For example, effective value and peak-to-peak value features of the time domain can be extracted through time domain analysis for high-frequency information such as pressure pulsation and vibration, and fault features and frequency amplitude features of the frequency domain can be extracted through frequency domain analysis; features such as return oil temperature rise, outlet temperature rise, and amplitude trend can be extracted for low-frequency information such as temperature and flow.

[0039] Furthermore, in this step, a large model can be constructed using methods such as machine learning and deep learning. By learning a large amount of segmented life test data, degradation features related to hydraulic pump degradation can be automatically identified and extracted to obtain the first parameter features.

[0040] (2) Screening the first parameter feature to obtain a first sensitive feature.

[0041] It can be understood that, among the first parameter characteristics, different first parameter characteristics have different importance for the prediction of the health status of the hydraulic pump. In order to better build the model, it is necessary to screen the first parameter characteristics and select the characteristics of the first parameter characteristics that are sensitive to the health status of the hydraulic pump. For the convenience of explanation, the characteristics of the first parameter characteristics that are sensitive to the health status of the hydraulic pump are referred to as the first sensitive characteristics.

[0042] In specific implementation, the change trend of each first parameter feature over time is analyzed. If the feature becomes larger or smaller with time, the feature is a sensitive feature; on the contrary, if the feature has no change trend with time, the feature is an insensitive feature. Thus, the first parameter features are screened to obtain the first sensitive feature.

[0043] (3) Using principal component analysis to perform dimensionality reduction processing on the first sensitive feature to obtain a first dimensionality reduction feature.

[0044] Specifically, although the degradation feature extraction from multiple angles can fully express the health degradation state, the first sensitive feature inevitably contains some redundant, useless or even ambiguous features, which leads to heavy computational burden of the prediction algorithm, poor robustness of prediction performance, and increased complexity of subsequent analysis. Therefore, it is necessary to perform dimensionality reduction and compression processing on the first sensitive feature.

[0045] Specifically, principal component analysis is mainly used in fields such as data dimensionality reduction, feature extraction, and data visualization. It converts the original variables into a set of new unrelated variables, namely principal components, by performing a linear transformation on the original variables. These principal components are linear combinations of the original variables, which can retain as much information as possible from the original data while reducing the dimension of the data. For the specific implementation process of using principal component analysis to reduce the dimensionality of the first sensitive feature to obtain the first reduced dimensionality feature, please refer to the description in the relevant technology, which will not be repeated here.

[0046] (4) Fusing the first dimensionality reduction features to obtain a first fused feature.

[0047] Specifically, fusion features refer to integrating multiple different sensitive features through specific methods to form a comprehensive feature that can more comprehensively and accurately reflect the health status of the product. In the health assessment of hydraulic pumps, there may be multiple sensitive features, such as pressure pulsation, flow change, vibration amplitude, temperature change, etc. These features may only provide partial information when viewed individually, but through fusion, their advantages can be combined to more effectively describe the overall health status of the hydraulic pump.

[0048] In specific implementation, the first dimensionality reduction feature can be fused by weighted averaging, feature concatenation or other methods to obtain the first fused feature.

[0049] Furthermore, the degradation feature analysis is performed on the continuous life test data to obtain the second fusion feature in the following specific implementation steps:

[0050] (1) Extracting features from the continuous life test data to obtain second parameter features.

[0051] Specifically, continuous life test data usually contains a lot of information about the performance changes of hydraulic pumps over time. Feature extraction of continuous life test data is to find parameter characteristics that can represent the essential characteristics of hydraulic pump performance from complex and large data, providing a basis for subsequent analysis.

[0052] Furthermore, in this step, the continuous life test data can be feature extracted through time domain analysis, frequency domain analysis, time-frequency domain analysis and statistical analysis to extract useful degradation features. For example, for high-frequency information such as pressure pulsation and vibration, the effective value and peak-to-peak value characteristics of the time domain can be extracted through time domain analysis, and the fault characteristics and frequency amplitude characteristics of the frequency domain can be extracted through frequency domain analysis; for low-frequency information such as temperature and flow, the return oil temperature rise, outlet temperature rise and amplitude trend characteristics can be extracted.

[0053] Furthermore, in this step, a large model can be constructed using machine learning, deep learning and other methods. By learning a large amount of continuous life test data, degradation features related to hydraulic pump degradation can be automatically identified and extracted to obtain the second parameter features.

[0054] (2) Screening the second parameter feature to obtain a second sensitive feature.

[0055] (3) Performing dimensionality reduction processing on the second sensitive feature using a principal component analysis method to obtain a second dimensionality reduction feature.

[0056] (4) Fusing the second dimensionality reduction features to obtain second fused features, and mapping the health status according to the second fused features.

[0057] Specifically, for the implementation process of obtaining the second fusion feature in steps (2) to (4), please refer to the above description, which will not be repeated here.

[0058] Specifically, when mapping the health status based on the second fusion feature, it is first necessary to determine the health status level. It should be noted that the health status level of the hydraulic pump is set according to actual needs and is not limited in this embodiment. For example, in a possible implementation, the range of the health status H is 0-1. If H>0.8, the hydraulic pump is in good condition without performance degradation or insignificant performance degradation, and no maintenance is required at this time; if 0.4<H<0.8, it means that the hydraulic pump is in a general state and a task assessment is needed to determine the remaining service life and maintenance plan of the hydraulic pump; if H<0.4, it indicates that the hydraulic pump is in a dangerous state and immediate maintenance and replacement are required.

[0059] Furthermore, in this step, the mapping relationship between the second fusion feature and the health status can be constructed by methods such as mathematical analysis and machine learning. For example, in an embodiment, by performing data analysis on the second fusion feature and the pre-set health status, the relationship between the two is found, and a mathematical model between the second fusion feature and the health status is established. During use, when the second fusion feature is input into the constructed mathematical model, the mathematical model will output the health status corresponding to the second fusion feature.

[0060] S103. Construct a first life prediction model according to the first fusion feature.

[0061] The specific implementation steps include:

[0062] (1) Determine the number of health status classification clusters according to the feature differences under different health status levels of the hydraulic pump.

[0063] Specifically, combining the above description, the features under different health status levels of the hydraulic pump are the features corresponding to each health status in the first fusion feature. The number of classification clusters is determined through clustering evaluation indicators, which can be the Silhouette Coefficient, Calinski-Harabasz index, Davies-Bouldin index, etc. The clustering evaluation indicators are selected according to actual needs and are not limited in this embodiment.

[0064] (2) Cluster the first fusion feature according to the determined number of health status classification clusters.

[0065] Specifically, after determining the number of health status classification clusters, the clustering algorithm can be used to cluster the first fusion feature. The first fusion feature is obtained by extracting and fusing the hydraulic pump monitoring data. It integrates information from multiple dimensions and can more comprehensively reflect the health status of the hydraulic pump. The first fusion feature can be clustered using algorithms such as K-Means clustering and hierarchical clustering. The features are divided into different clusters based on the similarity between the features, so that the features within the same cluster have high similarity, while the features between different clusters have large differences.

[0066] (3) The cluster with the largest number of features is the target health state corresponding to the first fusion feature.

[0067] Specifically, after clustering is completed, the number of features in each cluster is counted. It can be understood that the cluster with the largest number of features represents the most common health state of the hydraulic pump, so the cluster with the lowest number of features is determined as the target health state corresponding to the first fusion feature.

[0068] (4) Adjusting the number of clusters and cluster center features based on the difference between the target health state and the labeled health state corresponding to the first fusion feature to obtain a first life expectancy prediction model.

[0069] Specifically, the marked health status corresponding to the first fusion feature can be obtained by manual labeling, comparison of historical data, or machine learning assisted labeling. If there is a large difference between the target health status and the marked health status, the number of clusters and the cluster center features can be adjusted. For example, if it is found that the features in a cluster do not match the marked health status, the number of clusters can be readjusted or the center features of the cluster can be adjusted to improve the accuracy of clustering.

[0070] Furthermore, after obtaining the first fused feature data, its corresponding health status is known, that is, it constitutes sample data carrying health status labels. The clustering model is trained based on the sample data to determine the number of health status classification clusters and the central feature data of each cluster, and the first life prediction model is obtained.

[0071] Furthermore, distance evaluation, regression analysis and other methods may be used to construct a first life prediction model based on the first fusion feature. The input of the first life prediction model is the first fusion feature, and the output of the first life prediction model is the health status of the hydraulic pump.

[0072] S104: Construct a second life expectancy prediction model according to the second fusion feature.

[0073] The specific implementation steps include:

[0074] (1) Obtaining the real-time health status of the sample hydraulic pump identified by the first life prediction model.

[0075] Specifically, the sample hydraulic pump is identified using the first life prediction model to obtain the real-time health status of the sample hydraulic pump output by the first life prediction model. The real-time health status reflects the operating status of the sample hydraulic pump at the current moment.

[0076] (2) The real-time health status of the sample hydraulic pump and the second fusion feature are input into the BiLSTM, and the BiLSTM is trained based on the remaining life label of the sample hydraulic pump to obtain a second life prediction model.

[0077] Specifically, BiLSTM is Bidirectional Long Short-Term Memory, a bidirectional long short-term memory network. The BiLSTM model consists of a forward LSTM model (Long Short-Term Memory) and a backward LSTM model. The forward LSTM model achieves diagnosis, prediction and other tasks by learning from past information, while the backward LSTM model retains future information. In the backward running process, it can capture the features of the subsequent part of the sequence. By combining the forward LSTM model and the backward LSTM model to establish a BiLSTM model, the BiLSTM model can save past and future information at any time point, thereby more comprehensively understanding the contextual relationship of the data.

[0078] Furthermore, the remaining life label is the actual remaining life value of the known sample hydraulic pump. The remaining life label can be obtained through actual monitoring and experiments or determined based on experience and expert knowledge. The real-time health status, the second fusion feature and the remaining life label of the sample hydraulic pump are input into BiLSTM. By adjusting the parameters of the model, BiLSTM can learn the mapping relationship between the input feature and the remaining life. During the training process, the loss function is used to measure the difference between the predicted value of the model and the actual remaining life label, and the difference is continuously reduced through the optimization algorithm. After multiple iterative training, when the performance of the model meets the preset requirements, the second life prediction model can be obtained. The input of the second life prediction model is the second fusion feature value, and the output of the second life prediction model is the remaining life of the hydraulic pump.

[0079] S105. Fuse the first life prediction model and the second life prediction model to obtain a life prediction model; wherein, the real-time health status of the hydraulic pump is identified based on the first life prediction model, and the remaining life of the hydraulic pump in the real-time health status is predicted based on the second life prediction model.

[0080] Specifically, the input of the first life prediction model is the current state parameters of the hydraulic pump, and the output of the first life prediction model is the prediction result of the current health status of the hydraulic pump. The first life prediction model is used to identify the monitoring data of the hydraulic pump and determine the current health status of the hydraulic pump; the input of the second life prediction model is the state parameters of the hydraulic pump in the current health state predicted by the first life prediction model, and the output of the second life prediction model is the remaining life of the hydraulic pump in the current health state. The second life prediction model is used to predict the remaining life of the hydraulic pump based on the health status of the hydraulic pump predicted by the first life prediction model.

[0081] Furthermore, the first life prediction model and the second life prediction model may be fused by hierarchical fusion, parallel fusion, feedback fusion and other methods, which are not limited in this embodiment.

[0082] For example, in one possible implementation, a hierarchical fusion method is used to fuse the first life prediction model and the second life prediction model. First, the first life prediction model is used to evaluate the current health status of the hydraulic pump, that is, the first life prediction model outputs the current health status of the hydraulic pump. Then, the continuous test data of the current health status of the hydraulic pump output by the first life prediction model in the continuous life test data is input into the second life prediction model. Thus, the second life prediction model predicts the remaining life of the hydraulic pump and outputs the remaining life of the hydraulic pump.

[0083] The method for constructing a hydraulic pump life prediction model provided in this embodiment, on the one hand, by obtaining the continuous life test data and segmented life test data of the hydraulic pump, the performance changes and degradation process of the hydraulic pump can be understood from different angles and time scales. Combining the two types of data can more comprehensively reflect the actual working conditions of the hydraulic pump and reduce the deviations and limitations that may be caused by a single data source. On the second hand, by constructing the first life prediction model and the second life prediction model based on the first fusion feature and the second fusion feature, multiple indicators and factors can be integrated to more accurately describe the health status and degradation trend of the hydraulic pump, thereby improving the accuracy and reliability of the model's prediction of the hydraulic pump life. On the third hand, by comprehensively using the model established by continuous life test data and segmented life test data, the health of the hydraulic pump can be comprehensively evaluated. By predicting the working status and remaining life of the product in advance, maintenance time and resources can be reasonably arranged to support predictive maintenance.

[0084] Figure 2 This is a flowchart of the second embodiment of the method for constructing a hydraulic pump life prediction model provided by this application. Figure 2 , based on the above embodiment, the method of obtaining continuous life test data of the hydraulic pump comprises:

[0085] S201. Perform a durability life test and an accelerated life test on the hydraulic pump.

[0086] Specifically, the steps of performing a durability life test on the hydraulic pump include:

[0087] (1) Determine the rated operating conditions of the hydraulic pump and build a durability test bench according to the rated operating conditions.

[0088] Specifically, the rated operating conditions of a hydraulic pump refer to a set of specific working conditions under which the hydraulic pump is expected to operate stably and reliably during design and normal use, which usually include parameters such as pressure, flow, and speed. When conducting an endurance life test on a hydraulic pump, these parameters should be as close as possible to the working conditions of the hydraulic pump in actual operation.

[0089] Furthermore, a durability test bench is built based on the determined rated operating conditions.

[0090] (2) The durability test bench constructs the rated operating condition working environment, so that the hydraulic pump continues to operate under the rated operating condition until the hydraulic pump fails.

[0091] Specifically, the hydraulic pump is made to work continuously on the test bench according to the determined rated operating conditions (pressure, flow, speed). This process simulates the long-term use of the hydraulic pump in actual applications to test its durability and reliability. Specific fault judgment criteria are formulated according to actual conditions, and the status of the hydraulic pump is continuously monitored and evaluated during the test. Once the fault judgment criteria are met, the test is stopped immediately.

[0092] For example, in a possible implementation, the fault judgment criterion is:

[0093] 1. Serious performance degradation, such as significant reduction in flow rate, inability to maintain pressure at rated value, etc.;

[0094] 2. Abnormal noise and increased vibration may indicate damage or excessive wear of internal parts;

[0095] 3. Leakage occurs, affecting the normal operation of the hydraulic system.

[0096] Furthermore, the step of performing an accelerated life test on the hydraulic pump includes:

[0097] (1) Determine the operating conditions of the hydraulic pump and build an accelerated test bench according to the operating conditions;

[0098] Specifically, in the accelerated life test, the aging and failure process of the hydraulic pump should be accelerated by increasing the stress conditions so as to evaluate its life characteristics in a shorter time. The operating conditions of the hydraulic pump can be determined by increasing the speed, pressure, displacement, etc. of the hydraulic pump, or by increasing the working environment temperature and increasing the environmental pollution, etc., which are not limited in this embodiment.

[0099] Furthermore, as in the endurance life test, an accelerated test bench is built according to the determined operating conditions.

[0100] (2) The accelerated test bench constructs the operating condition working environment, allows the hydraulic pump to continuously operate under the operating condition, and adjusts the operating condition until the hydraulic pump fails.

[0101] Specifically, the hydraulic pump is made to work continuously on the accelerated test bench according to the determined operating conditions, a specific fault judgment standard is formulated according to the actual situation, and the state of the hydraulic pump is continuously monitored and evaluated during the test. Once the fault judgment standard is met, the test is stopped immediately. For example, in one embodiment, the fault judgment standard is:

[0102] 1. Serious performance degradation, such as significant reduction in flow rate, inability to maintain pressure at rated value, etc.;

[0103] 2. Abnormal noise and increased vibration may indicate damage or excessive wear of internal parts;

[0104] 3. Leakage occurs, affecting the normal operation of the hydraulic system.

[0105] S202. Recording in real time the state parameters of the hydraulic pump under the endurance life test and the accelerated life test.

[0106] Specifically, various state parameters of the hydraulic pump under the endurance life test and the accelerated life test are monitored and recorded in real time through sensors installed on the hydraulic pump.

[0107] S203, combining the state parameters of the endurance life test and the state parameters of the accelerated life test to obtain the continuous life test data.

[0108] Specifically, determine the start and end time of the endurance life test and the accelerated life test, as well as the state parameters corresponding to each time point during the test. Take the test duration of the endurance life test as the reference time period, and use the acceleration factor to convert the test duration of the accelerated life test so that the test duration of the accelerated life test is the same as the test duration of the endurance life test. For example, in one possible implementation, the acceleration factor is 2, and the test duration of the accelerated life test is 10 hours, which is equivalent to 20 hours in the endurance life test.

[0109] Furthermore, the order of parameters of the merged data is determined according to the characteristics to be analyzed. For example, the order can be time, pressure, flow, temperature, vibration, etc. The endurance life test data and the adjusted accelerated life test data are merged line by line. For the data at the same time point, if there are records in both tests, you can choose to keep one of them or take the average; if there is only one test record, keep the record. In this way, continuous life test data is obtained.

[0110] The method provided in this embodiment performs a durability life test and an accelerated life test on the hydraulic pump respectively, and combines the state parameters of the durability life test and the state parameters of the accelerated life test to determine continuous life test data. In this way, the performance of the hydraulic pump under different working conditions can be fully covered, providing a richer data basis for life prediction, and ensuring the comprehensiveness of the hydraulic pump test data. In addition, a single durability life test may take a long time to obtain sufficient data, and in actual applications, the equipment may encounter various unexpected situations, resulting in uncertainty in the test results; although the accelerated life test can obtain results in a shorter time, due to the difference between the acceleration conditions and the actual use conditions, the results may have certain deviations. The data of the two tests can be combined to complement and verify each other, thereby improving the reliability and accuracy of life prediction.

[0111] Figure 3 This is a flowchart of the third embodiment of the method for constructing a hydraulic pump life prediction model provided by this application. Figure 3 Based on the above embodiment, the step of obtaining segmented life test data of the hydraulic pump comprises:

[0112] S301. Classify the health status of the hydraulic pump according to the wear fault of the hydraulic pump.

[0113] Specifically, the health status of the hydraulic pump is divided into the following types according to the wear and tear of the hydraulic pump:

[0114] (1) Good condition, indicating that the hydraulic pump has no wear and tear and is a newly put into use product;

[0115] (2) Slight wear state, indicating that the output power of the hydraulic pump has decreased to a certain extent, and the function has decreased to a certain extent, but it can work normally and can be downgraded for use;

[0116] (3) Moderate wear indicates that the hydraulic pump function has declined and cannot meet the performance requirements, and needs to be disassembled and repaired;

[0117] (4) Severe wear indicates that the hydraulic pump has been scrapped and cannot be used and needs to be replaced.

[0118] S302, selecting hydraulic pumps in different health states to continuously operate for a preset time under the same operating conditions.

[0119] Specifically, for each hydraulic pump in a healthy state, it is made to continuously operate for a preset time under the same operating conditions. The preset time is set according to actual needs and is not limited in this embodiment.

[0120] Furthermore, the specific steps of the experiment are described above and will not be repeated here.

[0121] S303, recording in real time the test data of the hydraulic pump in different health states during the test process to obtain the segmented life test data.

[0122] Specifically, by installing sensors on the hydraulic pump, the test data of the hydraulic pump in different health states during the test is monitored and recorded in real time, and the test data corresponding to the different health states are integrated together to obtain segmented life test data.

[0123] The method provided in this embodiment divides the health status of the hydraulic pump according to the wear fault of the hydraulic pump; selects hydraulic pumps in different health status to continuously run for a preset time under the same operating conditions; and records the test data of the hydraulic pumps in different health status during the test in real time to obtain the segmented life test data. The performance of the hydraulic pump at different wear stages can be understood more accurately. Different wear faults correspond to different health levels, which can provide more targeted information for subsequent analysis. In addition, by monitoring wear faults, potential problems can be discovered in time before the hydraulic pump has a serious fault. Classification according to different health states can identify early signs of faults such as minor wear earlier, providing sufficient time for taking preventive maintenance measures.

[0124] Figure 4 This is a flowchart of the fourth embodiment of the method for constructing a hydraulic pump life prediction model provided by this application. Figure 4 , after obtaining the life prediction model; further comprising:

[0125] S401, detecting real-time status data of the hydraulic pump to be evaluated.

[0126] Specifically, this step collects real-time data of the hydraulic pump to be evaluated during operation through various sensors and monitoring equipment. The real-time status data of the hydraulic pump includes the current pressure pulsation, vibration information, temperature, flow rate and other parameters of the hydraulic pump.

[0127] S402: Identify the status category to which the real-time health status of the to-be-evaluated hydraulic pump belongs based on the real-time status data and the first life prediction model.

[0128] Specifically, in combination with the above description, the first life prediction model is used to identify the health status category of the hydraulic pump, which classifies the current health status of the hydraulic pump into different status categories according to the input real-time status data of the hydraulic pump. It should be noted that the health status of the hydraulic pump is pre-set. For example, in a possible implementation, the health status of the hydraulic pump includes: normal state, light wear state, medium wear state, heavy wear state and fault state.

[0129] S403: The second life prediction model predicts the state retention time of the hydraulic pump to be evaluated based on the state category.

[0130] Specifically, the input of the second life prediction model is the state category of the hydraulic pump output by the first life prediction model. Based on the state category and combined with the historical data and equipment characteristics of the hydraulic pump in the data set under this state category, the state retention time of the hydraulic pump to be evaluated is predicted. It can be understood that the state retention time of the hydraulic pump refers to the length of time the hydraulic pump is expected to continue to operate normally under the current state category.

[0131] S404: Taking the state category as a starting point and the state holding time as an interval, determine the elapsed time of all states to obtain the remaining life of the hydraulic pump to be evaluated.

[0132] Specifically, the state category output by the first life prediction model is used as the starting point. Then, the state retention time output by the second life prediction model is used as the interval to gradually calculate the time the hydraulic pump has been in different states. By accumulating the time spent in each state, the total time the hydraulic pump has experienced from the initial state to the current state can be obtained. The remaining life of the hydraulic pump to be evaluated can be obtained by subtracting the experienced life time from the rated life of the hydraulic pump.

[0133] The method provided in this embodiment obtains the real-time status data of the hydraulic pump and classifies the health status of the hydraulic pump using the first life prediction model, which helps maintenance personnel quickly understand the current status of the equipment and determine the urgency and priority of maintenance. It can also issue an early warning before a serious failure of the hydraulic pump by identifying the status category, and then provide a clear time reference by predicting the status retention time based on the status category to which it belongs through the second life prediction model, to ensure that the equipment is repaired or replaced before it fails, and accurate status retention time prediction can help enterprises better plan production tasks. Avoid production interruptions caused by sudden equipment failures and improve production efficiency. In summary, through real-time status data detection, the first life prediction model is used to predict status category identification, and the second life prediction model is used to predict the status retention time based on the status category, which provides comprehensive and accurate information for the maintenance and management of the hydraulic pump, helping to improve the reliability and production efficiency of the equipment and reduce maintenance costs.

[0134] Corresponding to the aforementioned embodiment of a method for constructing a hydraulic pump life prediction model, the present application also provides an embodiment of a device for constructing a hydraulic pump life prediction model.

[0135] Figure 5 This is a schematic diagram of the structure of the first embodiment of the device for constructing a hydraulic pump life prediction model provided by this application. Figure 5 The device provided in this embodiment includes an acquisition module 510, an analysis module 520 and a construction module 530; wherein:

[0136] The acquisition module 510 is used to acquire segmented life test data and continuous life test data of the hydraulic pump;

[0137] The analysis module 520 is used to perform degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; wherein the first fusion feature is a feature corresponding to the segmented life test data, and the second fusion feature is a feature corresponding to the continuous life test data;

[0138] The construction module 530 is used to construct a first life prediction model according to the first fusion feature;

[0139] The construction module 530 is further used to construct a second life prediction model according to the second fusion feature;

[0140] The construction module 530 is also used to fuse the first life prediction model and the second life prediction model to obtain a life prediction model; wherein the real-time health status of the hydraulic pump is identified based on the first life prediction model, and the remaining life of the hydraulic pump in the real-time health status is predicted based on the second life prediction model.

[0141] The device of this embodiment can be used to perform Figure 1 The steps, specific implementation principles and implementation processes of the method embodiment shown are similar and will not be repeated here.

[0142] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0143] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0144] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for constructing a hydraulic pump life prediction model, characterized in that: The method comprises: Obtain segmented life test data and continuous life test data of hydraulic pumps; Performing degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; wherein the first fusion feature is a feature corresponding to the segmented life test data, and the second fusion feature is a feature corresponding to the continuous life test data; Constructing a first life prediction model according to the first fusion feature; comprising: Determine the number of health status classification clusters according to the characteristic differences of the hydraulic pump at different health status levels; Clustering the first fusion feature according to the determined number of health status classification clusters; The cluster with the largest number of features is the target health state corresponding to the first fusion feature; adjusting the number of clusters and cluster center features based on the difference between the target health state and the labeled health state corresponding to the first fusion feature to obtain a first life prediction model; Constructing a second lifespan prediction model according to the second fusion feature; comprising: Acquire the real-time health status of the sample hydraulic pump identified by the first life prediction model; Inputting the real-time health status of the sample hydraulic pump and the second fusion feature into the BiLSTM, training the BiLSTM based on the remaining life label of the sample hydraulic pump, and obtaining a second life prediction model; The first life prediction model and the second life prediction model are integrated to obtain a life prediction model; wherein, the real-time health status of the hydraulic pump is identified based on the first life prediction model, and the remaining life of the hydraulic pump in the real-time health status is predicted based on the second life prediction model.

2. The method according to claim 1, characterized in that: The method of obtaining continuous life test data of the hydraulic pump comprises: Conducting a durability life test and an accelerated life test on the hydraulic pump respectively; Recording in real time the state parameters of the hydraulic pump under the endurance life test and the accelerated life test; The state parameters of the endurance life test and the state parameters of the accelerated life test are combined to obtain the continuous life test data.

3. The method according to claim 2, characterized in that The endurance life test of the hydraulic pump comprises: Determine the rated operating conditions of the hydraulic pump, and build a durability test bench according to the rated operating conditions; The endurance test bench constructs the rated operating condition working environment, allowing the hydraulic pump to continue to operate under the rated operating condition until the hydraulic pump fails.

4. The method according to claim 2, characterized in that: The accelerated life test of the hydraulic pump comprises: Determine the operating conditions of the hydraulic pump and build an accelerated test bench according to the operating conditions; The accelerated test bench constructs the operating condition working environment, allows the hydraulic pump to continue to operate under the operating condition, and adjusts the operating condition until the hydraulic pump fails.

5. The method according to claim 1, characterized in that The step of obtaining segmented life test data of the hydraulic pump comprises: Classifying the health status of the hydraulic pump according to the wear fault of the hydraulic pump; Select hydraulic pumps in different health states to run continuously for a preset time under the same operating conditions; The test data of the hydraulic pump in different health states during the test process are recorded in real time to obtain the segmented life test data.

6. The method according to claim 1, characterized in that Performing degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; comprising: Extracting features from the segmented life test data to obtain first parameter features; Screening the first parameter feature to obtain a first sensitive feature; Performing dimensionality reduction processing on the first sensitive feature by using a principal component analysis method to obtain a first dimensionality reduction feature; Fusing the first dimensionality reduction features to obtain a first fused feature; Extracting features from the continuous life test data to obtain second parameter features; Screening the second parameter feature to obtain a second sensitive feature; Performing dimensionality reduction processing on the second sensitive feature by using a principal component analysis method to obtain a second dimensionality reduction feature; The second dimensionality reduction features are fused to obtain second fused features, and the health status is mapped according to the second fused features.

7. The method according to claim 1, characterized in that After obtaining the life prediction model, the method further includes: Detect the real-time status data of the hydraulic pump to be evaluated; Identify the state category to which the real-time health state of the to-be-evaluated hydraulic pump belongs based on the real-time state data and the first life prediction model; The second life prediction model predicts the state retention time of the hydraulic pump to be evaluated based on the state category; Taking the state category as a starting point and the state retention time as an interval, the elapsed time of all states is determined to obtain the remaining life of the hydraulic pump to be evaluated.

8. A device for constructing a hydraulic pump life prediction model, characterized in that: The device comprises an acquisition module, an analysis module and a construction module; wherein, The acquisition module is used to acquire segmented life test data and continuous life test data of the hydraulic pump; The analysis module is used to perform degradation feature analysis on the segmented life test data and the continuous life test data respectively to obtain a first fusion feature and a second fusion feature; wherein the first fusion feature is a feature corresponding to the segmented life test data, and the second fusion feature is a feature corresponding to the continuous life test data; The construction module is used to construct a first life prediction model according to the first fusion feature; comprising: Determine the number of health status classification clusters according to the characteristic differences of the hydraulic pump at different health status levels; Clustering the first fusion feature according to the determined number of health status classification clusters; The cluster with the largest number of features is the target health state corresponding to the first fusion feature; adjusting the number of clusters and cluster center features based on the difference between the target health state and the labeled health state corresponding to the first fusion feature to obtain a first life prediction model; The construction module is further used to construct a second life prediction model according to the second fusion feature; comprising: Acquire the real-time health status of the sample hydraulic pump identified by the first life prediction model; Inputting the real-time health status of the sample hydraulic pump and the second fusion feature into the BiLSTM, training the BiLSTM based on the remaining life label of the sample hydraulic pump, and obtaining a second life prediction model; The construction module is also used to fuse the first life prediction model and the second life prediction model to obtain a life prediction model; wherein the real-time health status of the hydraulic pump is identified based on the first life prediction model, and the remaining life of the hydraulic pump in the real-time health status is predicted based on the second life prediction model.

Images