Transmission system prediction maintenance modeling method based on multi-level characteristic parameters

Abstract

The invention develops a transmission system prediction maintenance modeling method based on multi-level characteristic parameters, and the method comprises the steps: building a multi-level characteristic parameter spectrum from a monitoring parameter to a state index and then to a health index, building a PHM model of a transmission system, developing a python algorithm to calculate the state index and the health index, and on the basis, carrying out the prediction maintenance modeling of the transmission system. And judging whether the index is suitable for fault diagnosis of the transmission system by comparing the recognition capability of the state index on the fault mode, judging whether the model is suitable for health monitoring of the transmission system by comparing the recognition capability of the health index on the health state, and determining the threshold value of the health index for health monitoring. According to the method, through PHM model development, the appropriate state index and health index of the PHM model and the threshold value of the state index and the health index are determined from the recognition capability.

Description

Technical Field

[0001] The present invention belongs to the field of health monitoring and fault diagnosis of drive systems, and particularly relates to a predictive maintenance modeling method for drive systems based on multi-level characteristic parameters. Background Art

[0002] Aviation maintenance has evolved from the traditional preventive maintenance concept of "safety first, prevention first" through years of development, continuous research, creation, and innovation, to the modern aviation maintenance concept that combines the MSG maintenance concept and "reliability-based methods" to control maintenance. In recent years, with the rise of technologies such as industrial Internet of Things, machine learning, and AR, aviation maintenance technologies have been continuously iterated and upgraded, and the maintenance concept has shifted to predictive maintenance based on data analysis and trend analysis. In the application of PHM models for civil aircraft, significant achievements have been made in the fields of air conditioning system health monitoring, environmental control system fault diagnosis, bleed air system fault diagnosis, auxiliary power unit (APU) health assessment and life prediction, engine life prediction and fault warning, engine digital twin modeling and performance degradation trend prediction, etc.

[0003] Generally speaking, the drive system mainly consists of five parts: the main reducer, the intermediate reducer, the tail reducer, the main drive shaft, and the tail drive shaft, commonly known as "three reducers and two shafts". The five parts form a non-redundant power transmission chain, and some aircraft may not have an intermediate reducer. Among them, each level of reducer undertakes the tasks of realizing power steering and speed change. Their main components are three types of components: shafts, gears, and bearings. Due to the defects in the manufacturing process and inevitable assembly errors of these components, faults will inevitably occur in a complex working environment, which will in turn affect aircraft safety.

[0004] According to the investigation of the current situation of domestic aircraft maintenance, it is found that whether military or civilian, aircraft still adopt preventive maintenance of regular inspection and repair. Although it can basically meet the maintenance needs of aircraft, due to regular inspections, there will still be unnecessary waste, resulting in high maintenance costs and long maintenance cycles. In addition, preventive maintenance is still a bit lacking for military aircraft in actual combat with high intensity and fast pace.

[0005] Aircraft predictive maintenance is a key enabling technology to improve aircraft safety, availability, and economy. Predictive maintenance utilizes advanced theoretical methods and models to conduct system predictive model adaptability analysis, focusing on formulating maintenance strategies and decision-making methods, so as to minimize maintenance costs and improve efficiency during the maintenance process. Generally speaking, the research on aircraft predictive maintenance involves theoretical methods, model research, adaptability analysis, decision-making models and methods research, as well as effectiveness evaluation, etc. Its comprehensive application will help improve aircraft availability and maintenance efficiency, and contribute to safety and reliability. Summary of the Invention

[0006] Object of the Invention: The technical problem to be solved by the present invention is to provide a predictive maintenance modeling method for the transmission system based on multi-level characteristic parameters using flight parameter data collected by advanced sensors, construct a multi-level characteristic parameter spectrum based on the flight parameter data, calculate the state parameters of the transmission system, calculate the health index through the established PHM model of the aircraft transmission system, and based on the operation data, judge whether the model is suitable for the predictive maintenance of the aircraft transmission system by comparing the recognition capabilities of the PHM model and the health index, and determine the threshold of the PHM model.

[0007] The present invention includes the following steps:

[0008] Step 1, construct a multi-level characteristic parameter spectrum for the transmission system, support the design of multi-level parameter spectra, and provide functions such as feature spectrum file management, preprocessing script management, and feature dependency; the multi-level includes component raw signals, mathematical statistics, complex mathematical statistics, and system health indicators;

[0009] Step 2, perform data import and data analysis, support uploading the original data from the transmission system through csv files, calculate the first three levels of multi-level characteristic parameters using the multi-level characteristic parameter spectrum, and be able to calculate the system-level health index using the model in Step 4;

[0010] Step 3, perform prediction and health management, establish a general algorithm library for realizing correlation analysis, trend prediction, and anomaly detection, model the PHM model, and develop different PHM (Prognostics Health Management, PHM) models according to different needs of customers (such as calculating the correlation degree of multi-level parameters for specific faults, predicting the values of multi-level parameters in the next period of time, etc.);

[0011] Step 4, perform training of the PHM model, conduct model training and optimization more than twice for the selected machine learning algorithm, and when the recognition accuracy of the model reaches the set requirements, release the PHM model for calculating the system-level health index;

[0012] Step 5: Visualize the health assessment of the drive system, display the multi-level characteristic parameters of the drive system according to different requirements, and quickly evaluate the health status of the drive system through visualization to achieve fault detection, so as to realize the predictive maintenance of the drive system.

[0013] In Step 1, the multi-level characteristic parameter spectrum includes four levels of parameters, namely the first-level parameter, the second-level parameter, the third-level parameter, and the fourth-level parameter. According to their functions, the second-level parameter and the third-level parameter are called state indicators, and the fourth-level parameter is called a health indicator.

[0014] The first-level parameter is the basis for constructing the characteristic parameter spectrum of the drive system and is the original signal directly collected from the aircraft, that is, the monitoring parameter. According to the structural composition characteristics of the drive system, the vibration signals of its "three devices and two shafts" are usually observed with emphasis, and there will be different adjustments according to different aircraft models.

[0015] The second-level parameter is a mathematical statistic obtained by mathematical transformation of the first-level parameter, including time-domain statistics and frequency-domain statistics such as peak-to-peak value, root mean square, kurtosis, and sideband index. Some of these parameters have dimensions and some are dimensionless. These parameters can characterize the state of components to a certain extent and are defined as state indicators.

[0016] The third-level parameter is a more complex third-level parameter obtained by further calculation based on the second-level parameter, including energy ratio, sideband level factor, and sixth-order central moment of the differential signal. These parameters are more complex to calculate than the second-level parameter and are more sensitive to the state of components, and are also defined as state indicators.

[0017] The fourth-level parameter is the output obtained by taking the second-level parameter and the third-level parameter as the input of the PHM model and is used to characterize the health status of the drive system.

[0018] The characteristic spectrum file management and preprocessing script management are to manage the files of the multi-level characteristic parameter spectrum and the preprocessing scripts involved.

[0019] The characteristic dependency relationship is to display the dependency relationship of the characteristics at each level through a visualization approach.

[0020] In Step 2, the data import needs to achieve data upload according to specific naming requirements, that is, in the format of "flight time + flight number + sampling frequency", which is convenient for management. At the same time, data analysis can calculate and store single features or two or more characteristic parameters for single-group or two or more groups of data.

[0021] In Step 3, based on the multi-level characteristic parameter spectrum, a general algorithm library is developed to build the PHM model, or the customer develops an algorithm by themselves and imports it to achieve the PHM model building.

[0022] In step 4, the training data is used to train the PHM model. During training, the PHM model arranges the four-level parameters from small to large to determine the discrimination boundary. Then, the health index at the boundary is defined as the threshold by comparing with the actual health status.

[0023] After completing the training of the PHM model, the test data is input. According to the results output by the PHM model, the PHM model is evaluated using the test standard in GB / T43555-2023. An excellent model (a model with a state recognition accuracy greater than 90% and an abnormal state false negative rate lower than 10% is an excellent model) is selected, or a suitable model is selected according to the customer-defined standard. The model is stored in the parameter spectrum file management.

[0024] In step 5, the visualization of the health assessment of the transmission system includes the visualization display of multi-level characteristic parameters under different aircraft models, different flights, and different working conditions. The four-level characteristic parameters displayed at the top, that is, the system health index, can help quickly evaluate the health status of the transmission system, and the first three-level characteristic parameters can be screened and displayed at the bottom as required to help quickly isolate the location of the transmission system failure.

[0025] The present invention also provides an electronic device, including a processor and a memory. The memory stores program code. When the program code is executed by the processor, the processor executes the steps of the method.

[0026] The present invention also provides a storage medium storing a computer program or instruction. When the computer program or instruction is run, the method is implemented.

[0027] The beneficial effect of the present invention is that a predictive maintenance modeling method for the transmission system based on multi-level characteristic parameters is developed. From the perspective of characteristic parameters, a complete multi-level characteristic parameter spectrum from monitoring parameters to state indicators and then to health indicators is constructed. By establishing a PHM model for the aircraft transmission system, model development applicable to different flight stages is carried out, and the health indicators characterizing the health status are calculated, providing indicators for the assessment of the aircraft health status in actual engineering applications and helping to achieve predictive maintenance. Description of the Drawings

[0028] The following further specifically describes the present invention in conjunction with the drawings and specific embodiments, and the above and / or other advantages of the present invention will become clearer.

[0029] Figure 1 It is a flowchart of the method of the present invention. Detailed Embodiments

[0030] Such as Figure 1As shown in the figure, an embodiment of the present invention provides a predictive maintenance modeling method for a transmission system based on multi-level characteristic parameters, including the following steps:

[0031] Step 1: Construct a multi-level characteristic parameter spectrum for the transmission system, supporting the parameter spectrum design of multi-level "component original signal - mathematical statistic - complex mathematical statistic - system health indicator", and also including functions such as characteristic spectrum file management, preprocessing script management, and characteristic dependency relationships;

[0032] Step 2: Import and analyze data, supporting the upload of original data from the transmission system through a csv file, capable of calculating the first three levels of multi-level characteristic parameters using the parameter spectrum constructed in Step 1, and capable of calculating the system-level health indicator using the model in Step 4;

[0033] Step 3: Build a Prognostics Health Management (PHM) model, providing a general algorithm library for implementing correlation analysis, trend prediction, and anomaly detection, etc., for developing different PHM models according to different requirements;

[0034] Step 4: Train the PHM model, perform multiple model trainings and optimizations for the selected machine learning algorithm, and when the recognition accuracy of the model reaches the set requirements, the model can be released for calculating the system-level health indicator;

[0035] Step 5: Visualize the health assessment of the transmission system, capable of displaying the multi-level characteristic parameters of the transmission system according to different requirements, and quickly assessing the health status of the transmission system through visual display to achieve fault detection, thereby realizing the predictive maintenance of the transmission system;

[0036] In Step 1, the constructed multi-level characteristic parameter spectrum of the transmission system includes four levels of parameters, namely the first-level parameter, the second-level parameter, the third-level parameter, and the fourth-level parameter. Among them, according to the function, the second-level parameter and the third-level parameter are also called state indicators, and the fourth-level parameter is also called the health indicator;

[0037] The first-level parameter is the basis for constructing the characteristic parameter spectrum of the transmission system, which is the original signal directly collected from the aircraft, that is, the monitoring parameter. According to the structural composition characteristics of the transmission system, usually the vibration signals of its "three devices and two shafts" will be key observed, and there will be different adjustments according to different aircraft models;

[0038] The second-level parameter is the mathematical statistic obtained by mathematical transformation of the first-level parameter, including time-domain statistics and frequency-domain statistics such as peak-to-peak value, root mean square, kurtosis, and sideband index. Some of these parameters have dimensions and some are dimensionless. These parameters can characterize the state of the component to a certain extent and are defined as state indicators;

[0039] The tertiary parameters are further calculated based on the secondary parameters to obtain more complex tertiary parameters, such as energy ratio, sideband level factor, and sixth-order central moment of the differential signal. These parameters are more complex to calculate than the secondary parameters and are more sensitive to the state of the component, and are also defined as state indicators.

[0040] The quaternary parameters are the outputs obtained by using the secondary parameters and the tertiary parameters as the inputs of the PHM model. It can characterize the health state of the transmission system and is defined as a health indicator.

[0041] The feature spectrum file management and preprocessing script management mentioned above are the program codes for managing the implementation of feature calculation.

[0042] The feature dependency relationship shows the dependency relationship of features at each level through a visual approach.

[0043] In step 2, the data import needs to implement data upload according to specific naming requirements, that is, in the format of "flight time + flight number + sampling frequency", which is convenient for management. At the same time, data analysis can calculate and store single features or more than two feature parameters for single-group or multi-group data.

[0044] In step 3, for the PHM model building, the provided general algorithm library for implementing correlation analysis, trend prediction, and anomaly detection is used to select corresponding algorithms according to the customer's customized requirements, and it also supports the customer to develop and import algorithms to achieve model building. The multi-level feature parameter spectrum constructed in step 1 is used for model building.

[0045] In step 4, the PHM model training is carried out using the imported training and test data. When training, the determination of the discrimination boundary of the quaternary parameter, that is, the health indicator, depends on arranging this indicator from small to large, and then comparing it with the actual health state and defining the health indicator at the boundary as the threshold. When testing the model, the test data is used as the input of the trained model, and the model is evaluated according to the test standard in GB / T 43555-2023 based on the output result. An excellent model is selected or a suitable model is selected according to the customer's customized standard, and the model is stored in the parameter spectrum file management.

[0046] In step 5, the visualization of the transmission system health assessment includes the visual display of multi-level feature parameters under different aircraft types, different flights, and different working conditions. The quaternary feature parameters shown at the top, that is, the system health indicators, can help quickly assess the health state of the transmission system, and the first three levels of feature parameters can be screened and displayed at the bottom according to requirements to help quickly isolate the location of the transmission system fault.

[0047] The present invention also provides a storage medium storing a computer program or instructions, which, when run, implement the predictive maintenance modeling method for a transmission system based on multi-level characteristic parameters.

[0048] A specific embodiment of the present invention provides a method for determining evaluation indexes and thresholds of an aircraft predictive maintenance model oriented to cost, and the flow chart is as Figure 1 shown, including:

[0049] Step 1: Construct a multi-level characteristic parameter spectrum of the transmission system. According to the splitting of the transmission system, it is divided into three types of key components: gears, shafts, and bearings. Then, based on the vibration signals collected by sensors as the original signals of the components, secondary parameters and tertiary parameters serving as state indexes are calculated accordingly, and preparations are made for the calculation of quaternary parameters. The calculation programs of the parameters are stored in the parameter spectrum file management, and the required preprocessing programs are responsible for by the preprocessing management. Tables 1, 2, and 3 show the constructed parameter spectra of each level of the transmission system.

[0050] Table 1

[0051] Low-frequency monitoring value of Component 1's vibration in the X direction Vibration of Component 4 in the Y direction 1 High-frequency monitoring value of Component 1's vibration in the X direction Vibration of Component 4 in the Y direction 2 Low-frequency monitoring value of Component 1's vibration in the Y direction Vibration of Component 4 in the X direction 1 High-frequency monitoring value of Component 1's vibration in the Y direction Vibration of Component 4 in the X direction 2 Low-frequency monitoring value of Component 1's vibration in the Z direction Vibration of Component 4 in the Z direction 1 High-frequency monitoring value of Component 1's vibration in the Z direction Vibration of Component 4 in the Z direction 2 Vibration of Component 2 in the Z direction Vibration of Component 5 in the Y direction Vibration of Component 2 in the Y direction Vibration of Component 5 in the Z direction Vibration of Component 3 in the Y direction Vibration of Component 6 in the Z direction

[0052] Table 2

[0053]

[0054]

[0055] Table 3

[0056]

[0057] The quaternary parameter is set as the output of the trained model, the input data is the calculated secondary and tertiary parameters, and the output parameter is the reconstruction error, that is, the quaternary parameter.

[0058] Step 2: Import flight data into the system successfully through batch uploading according to the naming requirements. The imported data can also be deleted through the delete key. After the data is imported, restart the task to perform the calculation and analysis of specific parameters or all parameters according to the customer's requirements. The results support downloading, and screening and result display can be performed on the screening interface and the result display interface of the processing results.

[0059] Step 3: In this embodiment, the model training selects to use a convolutional autoencoder to perform PHM model modeling, and the modeling program is as follows.

[0060]

[0061]

[0062] Step 4: After the PHM model is established, the model is trained multiple times using the training data to obtain different models. The quantitative test results of the state monitoring algorithm are evaluated by comparing with the qualified line and the excellent line. The qualified line is the benchmark for judging whether the algorithm passes the test. The qualified line is that the state discrimination accuracy rate is greater than 80%, and the false negative rate of the abnormal state is less than 40%. The excellent line is the benchmark for judging the performance of the algorithm. The excellent line is that the state discrimination accuracy rate is greater than 90%, and the false positive rate of the abnormal state is less than 10%. The quantitative test results of the fault diagnosis algorithm are evaluated by comparing with the qualified line. The qualified line is the benchmark for judging whether the algorithm passes the test. The qualified indicators of the accuracy rate, precision rate, and recall rate of the machine learning algorithm are greater than 70%, and both the macro-average and micro-average should meet the qualified indicator requirements of the precision rate and recall rate. The quantitative test results of the prediction algorithm are evaluated by comparing with the qualified line. The qualified line is the benchmark for judging whether the algorithm passes the test. The qualified line of the prediction accuracy rate is 60%. When the recognition accuracy rate of the model reaches the set requirements, the model can be released for calculating the system-level health indicators.

[0063] Step 5: Visualize the health assessment of the transmission system. In this step, the calculated multi-level characteristic parameters will be displayed, and predictive maintenance of the transmission system can be achieved based on these parameters.

[0064] In a specific implementation, the present application provides a computer storage medium and a corresponding data processing unit. Among them, the computer storage medium can store a computer program, and when the computer program is executed by the data processing unit, it can run the inventive content of the method for predicting and maintaining the transmission system based on multi-level characteristic parameters provided by the present invention and some or all of the steps in each embodiment. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0065] Those skilled in the art can clearly understand that the technical solutions in the embodiments of the present invention can be implemented by means of a computer program and its corresponding general hardware platform. Based on such an understanding, the essence of the technical solutions in the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a computer program, that is, a software product. The computer program software product can be stored in a storage medium, including several instructions for causing a device (which can be a personal computer, a server, a single-chip microcomputer, a MUU, or a network device, etc.) containing a data processing unit to execute the methods described in each embodiment or some parts of the embodiments of the present invention.

[0066] The present invention provides a predictive maintenance modeling method for a transmission system based on multi-level characteristic parameters. There are many methods and ways to specifically implement this technical solution. The above description is only a preferred implementation manner of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Each component not clearly defined in this embodiment can be implemented by using the prior art.

Claims

1. A predictive maintenance modeling method for a transmission system based on multi-level characteristic parameters, characterized in that It includes the following steps: Step 1: Construct a multi-level characteristic parameter spectrum for the transmission system, support the design of the multi-level parameter spectrum, and provide functions for characteristic spectrum file management, preprocessing script management, and characteristic dependency relationship; the multi-level includes component original signals, mathematical statistics, complex mathematical statistics, and system health indicators; Step 2: Perform data import and data analysis, support the upload of original data from the transmission system through csv. files, calculate the first three levels of multi-level characteristic parameters using the multi-level characteristic parameter spectrum, and be able to calculate system-level health indicators using the model in Step 4; Step 3: Perform prediction and health management, establish a general algorithm library for realizing correlation analysis, trend prediction, and anomaly detection, model the PHM model, and develop different PHM models according to different customer requirements; Step 4: Train the PHM model, perform model training and optimization more than twice for the selected machine learning algorithm, and when the recognition accuracy of the model reaches the set requirements, release the PHM model for calculating system-level health indicators; Step 5: Visualize the health assessment of the transmission system, display the multi-level characteristic parameters of the transmission system according to different requirements, and realize fault detection by evaluating the health state of the transmission system through visualization, so as to realize predictive maintenance of the transmission system.

2. The method according to claim 1, wherein In Step 1, the multi-level characteristic parameter spectrum contains four levels of parameters, namely the first-level parameter, the second-level parameter, the third-level parameter, and the fourth-level parameter. According to the function, the second-level parameter and the third-level parameter are called state indicators, and the fourth-level parameter is called a health indicator; The first-level parameter is the original signal directly collected from the aircraft, that is, the monitoring parameter; The second-level parameter is the mathematical statistic obtained by mathematical transformation of the first-level parameter, including time-domain statistics and frequency-domain statistics; The third-level parameter is obtained by further calculation based on the second-level parameter, including energy ratio, sideband level factor, and sixth-order central moment of the differential signal; The fourth-level parameter is the output obtained by taking the second-level parameter and the third-level parameter as the input of the PHM model, and is used to characterize the health state of the transmission system; The characteristic spectrum file management and preprocessing script management are to manage the files of the multi-level characteristic parameter spectrum and the involved preprocessing scripts; The characteristic dependency relationship is to display the dependency relationship of each level of characteristics through a visual way.

3. The method according to claim 2, wherein In Step 2, the data import needs to realize data upload according to specific naming requirements, that is, in the format of flight time + flight number + sampling frequency. At the same time, data analysis can calculate and store single characteristics or two or more characteristic parameters for single-group or two or more groups of data.

4. The method according to claim 3, wherein In Step 3, based on the multi-level characteristic parameter spectrum, model the PHM model using the general algorithm library, or the customer develops an algorithm by himself and imports it to realize PHM model modeling.

5. The method according to claim 4, characterized in that In Step 4, use the training data to train the PHM model. During training, the PHM model arranges the fourth-level parameters from small to large to determine the discrimination boundary; then compare with the actual health state and define the health indicator at the boundary as the threshold; After completing the training of the PHM model, input the test data. According to the results output by the PHM model, evaluate the PHM model using the test criteria in GB / T 43555-2023. Select an excellent model or a suitable model according to the customer-defined criteria, and store the model in the parameter spectrum file management.

6. The method according to claim 5, wherein In step 5, the visualization of the health assessment of the transmission system includes the visualization display of multi-level characteristic parameters under different aircraft models, different flights, and different working conditions. The four-level characteristic parameters displayed at the top, that is, the system health indicators, can realize the assessment of the health status of the transmission system, and the first three-level characteristic parameters are screened and displayed at the bottom as required.

7. An electronic device, characterized in that, It includes a processor and a memory. The memory stores program codes. When the program codes are executed by the processor, the processor is caused to execute the steps of the method according to any one of claims 1 to 6.

8. A storage medium, characterized in that, It stores a computer program or instruction. When the computer program or instruction is run, the method according to any one of claims 1 to 6 is implemented.

Images