Turbocharger performance evaluation method and system

Abstract

The invention provides a turbocharger performance evaluation method and system, and relates to the technical field of artificial intelligence, and the method comprises the steps of data collection, model establishment, model training, performance state output, performance state evaluation and the like. The system comprises a data acquisition module, a model building module, a model training module, a performance state output module and a performance state evaluation module, the performance of the turbocharger can be comprehensively evaluated based on the equipment aging condition through a built performance degradation model, and the accuracy of turbocharger evaluation is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a turbocharger performance evaluation method and system. Background Art

[0002] With the rapid development of industrialization and the automobile industry, turbochargers, as key components for improving engine power performance and fuel economy, are increasingly used in modern industrial equipment and automobile engines. Turbochargers use the exhaust gas energy emitted by the engine to drive the turbine to rotate, thereby increasing the intake volume, improving combustion efficiency, and reducing pollutant emissions. The performance of the turbocharger has an important impact on the overall performance and operating economy of the engine. Its key performance indicators (such as compressor efficiency, turbine efficiency, and flow characteristics) will decline with the use time and operating conditions. When the performance of the turbocharger declines, it will lead to a decline in engine performance, increased fuel consumption, deterioration of exhaust emissions, and a significant decline in equipment reliability and operating safety. This not only affects the economic benefits and service life of the equipment, but also puts higher requirements on environmental protection and production safety. Therefore, how to scientifically evaluate the performance of the turbocharger and promptly discover performance problems is the key to ensuring stable operation of the equipment and extending its service life.

[0003] The Chinese invention patent with the application publication date of June 14, 2024 and publication number CN118194518A provides a real-time prediction method and device for the risk of turbocharger failure. The patent compares the working condition of the turbocharger with the corresponding compressor characteristic diagram, and then combines the surge boundary, overspeed boundary and blockage boundary to conduct risk assessment, thereby ensuring the normal use of the user and effectively improving the prediction work efficiency and the reliability of the results. With respect to the above-mentioned technical scheme, the main focus is on comparing the working condition of the turbocharger with the compressor characteristic diagram, and evaluating the turbocharger in combination with specific boundary conditions, but the performance degradation of the turbocharger due to equipment aging during long-term operation is not considered, and the performance degradation caused by equipment aging cannot be discovered in time, thereby affecting the accuracy of the turbocharger performance evaluation. Summary of the invention

[0004] In order to improve the accuracy of turbocharger evaluation, the present application provides a turbocharger performance evaluation method and system.

[0005] In a first aspect, the present application provides a turbocharger performance evaluation method, which adopts the following technical solution: A method for evaluating turbocharger performance comprises the following steps: Data collection: including real-time data collection and historical data collection; Collect real-time data: Collect real-time data of turbocharger; Collect historical data: Collect historical data of turbocharger; Model building: Construct a performance degradation model based on the nonlinear Gamma process; Model training: Use the historical data of the turbocharger to train the performance degradation model to obtain the trained performance degradation model; Performance status output: input the real-time data into the trained performance degradation model and output the current performance status value of the turbocharger; Performance status evaluation: Determine whether the current performance status value of the turbocharger is lower than the preset alarm threshold: If yes, then the result that the current performance status of the turbocharger is unqualified is displayed, and an abnormal performance status alarm of the turbocharger is given; If not, the result indicating that the current performance status of the turbocharger is acceptable is displayed.

[0006] By adopting the above technical solution, a performance degradation model based on the nonlinear Gamma process is constructed, and the performance degradation process caused by equipment aging is incorporated into the performance evaluation system of the turbocharger. The current performance status and degradation trend are dually evaluated to help staff accurately grasp the performance status of the equipment, timely discover potential problems caused by equipment aging, and avoid evaluation blind spots caused by ignoring aging effects. In addition, the performance degradation model based on the nonlinear Gamma process constructed can effectively describe the performance degradation law of the turbocharger under different working conditions. Through the performance degradation model, this application can adapt to the evaluation needs of the turbocharger under different working conditions, capture the nonlinear characteristics of equipment performance degradation, especially in scenarios of long-term continuous operation, and have stronger robustness.

[0007] Optionally, after the step of executing the model training and before the step of executing the performance status output, the method further includes: Marginal distribution processing: select the marginal distribution type for the historical data, use the maximum likelihood estimation method to fit the distribution parameters, perform marginal distribution processing on the historical data, and obtain the marginal distribution probability of the historical data; Generate joint distribution: Input the marginal distribution probability of historical data into the Copula function, and generate joint distribution by fitting the dependency structure of historical data; Generate simulated sample data: sample from the joint distribution to obtain simulated joint distribution probability values, convert the simulated joint distribution probability values ​​back to the space of historical data, and generate simulated sample data; Simulation training: Input the simulated sample data into the trained performance degradation model for simulation training to obtain the simulated trained performance degradation model as a new trained performance degradation model.

[0008] By adopting the above technical solutions, introducing edge distribution processing and joint distribution fitting, the data correlation under complex operating conditions can be captured, so that the model has the ability to adapt to different operating conditions, especially in extreme environments or non-standard operating conditions, and can still provide reliable performance evaluation results. In addition, generating simulated sample data as an extension of historical data reduces the dependence on high-frequency real-time data collection, helps reduce data collection costs, and optimizes data distribution coverage during model training to reduce model training complexity.

[0009] Optionally, after executing the step of building the model and before executing the step of joint distribution processing, the method further includes: Similarity screening: Calculate the similarity between historical data, compare the similarity with the preset similarity filtering threshold, and screen out historical data with similarity greater than the preset similarity filtering threshold as new historical data.

[0010] By adopting the above technical solution and introducing similarity screening between marginal distribution and joint distribution, data quality can be effectively improved, the joint distribution fitting effect can be optimized, the model's ability to capture degradation laws can be enhanced, and computational complexity can be reduced to adapt to a variety of application scenarios. Moreover, through this step, the accuracy and robustness of the performance degradation model can be significantly improved, thereby improving the accuracy and reliability of turbocharger performance evaluation.

[0011] Optionally, after executing the step of generating simulated sample data and before executing the step of simulated training, the method further includes: Global verification: Calculate the Wasserstein distance between historical data and generated simulated sample data, and determine whether the calculated Wasserstein distance is greater than the preset Wasserstein distance threshold: If so, the marginal distribution type and distribution parameters of the historical data are adjusted, and the marginal distribution processing step is performed; If not, then perform the local verification step; Local verification: Use the KernelDensity function to perform kernel density estimation on the historical data and the generated simulated sample data, compare the density functions of the two, and determine whether the similarity of the two in each local area is greater than the preset similarity threshold: If yes, then execute the simulation training steps; If not, adjust the Copula function type and execute the steps to generate a joint distribution.

[0012] By adopting the above technical solution, a multi-level and multi-link verification mechanism is constructed by introducing global verification and local verification, which ensures that the generated simulation sample data can truly reflect the characteristics of historical data both globally and locally, optimizes the data generation quality, enhances the model training effect, and reduces the risk of model failure. At the same time, it improves the accuracy and robustness of identifying turbocharger performance degradation and abnormal conditions.

[0013] Optionally, after the step of outputting the performance status and before the step of evaluating the performance status, the method further includes: Predicting performance status: inputting real-time data into the trained performance degradation model to predict the performance status value of the turbocharger at a certain point in the future, which is recorded as the predicted performance status value; Prediction performance status evaluation: Determine whether the prediction performance status value is lower than the preset prediction performance threshold: If so, the time point at which the predicted performance status value is displayed; If not, no processing is performed.

[0014] By adopting the above technical solution, the performance status value at a specific time point is predicted, and the maintenance window is reasonably planned to avoid unnecessary frequent maintenance and equipment downtime, thereby reducing operating costs. Moreover, when the performance status is predicted to degrade to the preset prediction threshold, a prompt is given so that the staff can take preventive maintenance measures to reduce the damage to the equipment due to long-term operation in abnormal conditions and delay equipment aging.

[0015] Optionally, after the step of outputting the performance status and before the step of evaluating the performance status, the method further includes: Setting an initial threshold: setting a first preset threshold according to a turbocharger equipment standard; Initial performance status assessment: Determine whether the current performance status value of the turbocharger is lower than the first preset threshold: If not, the result indicating that the current performance status of the turbocharger is excellent is displayed; If so, it is displayed that the current performance status of the turbocharger is not excellent.

[0016] By adopting the above technical solutions, the initial classification and evaluation of performance status is carried out based on the initial threshold set by the standard, which effectively improves the efficiency of equipment monitoring and the accuracy of operation and maintenance decisions. The preliminary evaluation results can provide a basis for subsequent deeper analysis and intervention, help discover potential problems in advance, and improve the operating efficiency and life of the equipment.

[0017] Optionally, after the step of performing the initial performance status evaluation and before the step of performing the performance status evaluation, the method further includes: Calculating the performance degradation rate: using the performance degradation model to output the performance state value of the turbocharger at the previous moment, and calculating the rate of change of the performance value at the current moment and the previous moment; Dynamically adjust the threshold: calculate the preset performance status threshold, record the preset performance status threshold as the preset alarm threshold, and the calculation model of the preset performance status threshold is as follows: ; in, Indicates the current moment, Indicates the preset alarm threshold at the current moment. represents the first preset threshold, represents the adjustment factor, Indicates the rate of change of the performance value between the current moment and the previous moment.

[0018] By adopting the above technical solutions, introducing the mechanism of performance change rate calculation and dynamic adjustment of alarm threshold, more accurate tracking and management of equipment performance can be achieved. In addition, the dynamically adjusted threshold can respond to changes in equipment degradation rate in real time, ensuring that the equipment monitoring system always maintains high efficiency and accuracy under different operating environments. This not only improves the timeliness and accuracy of fault warnings, but also optimizes the flexibility of equipment management, helps staff to accurately diagnose and manage equipment at all stages, and extends the life cycle of the turbocharger.

[0019] Optionally, after executing the step of providing an alarm prompt for abnormal turbocharger performance status, the method further includes: First processing: inputting the historical data into the performance degradation model to obtain the performance degradation data of the turbocharger, and using the performance degradation data to draw a performance degradation curve, which is recorded as a first curve; The second process: a period of time before the current time point is recorded as a first period of time, and the data of the turbocharger in the first period of time is collected and recorded as first data; The third process: inputting the first data into the performance degradation model to obtain the first performance data, and drawing a performance change curve according to the first performance data, which is recorded as the second curve; First judgment: Calculate the curve similarity between the first curve and the second curve, and judge whether the curve similarity is greater than a preset curve similarity threshold: If so, it will display the serious aging result of the turbocharger equipment and issue a turbocharger equipment maintenance alarm; If not, the turbocharger equipment failure result is displayed, and a turbocharger equipment failure alarm is issued.

[0020] By adopting the above technical solution, the historical data of the equipment is combined with the real-time data, the similarity between the performance change curve and the performance degradation curve is calculated, and the aging or fault status of the equipment is judged based on this, which effectively improves the accuracy and intelligence level of equipment evaluation. At the same time, by accurately distinguishing between equipment aging and failure, equipment maintenance or component replacement can be carried out in advance, reducing the losses caused by equipment failure, improving the efficiency of equipment management, and greatly enhancing the reliability of the turbocharger in long-term operation, ensuring efficient and stable operation of the equipment.

[0021] Optionally, after executing the step of displaying the qualified result of the current performance status of the turbocharger, the method further includes: Fourth process: inputting the historical data into the performance degradation model to obtain the performance degradation data of the turbocharger, and using the performance degradation data to draw a performance degradation curve, which is recorded as the third curve; The fifth process: recording a period of time before the current time point as a second period of time, collecting data of the turbocharger in the second period of time, and recording it as second data; Sixth processing: inputting the second data into the performance degradation model to obtain second performance data, and drawing a performance change curve according to the second performance data, which is recorded as a fourth curve; Second judgment: Map the fourth curve in the coordinate system of the third curve to determine whether the fourth curve is within the upper bound of the third curve: If yes, it will show that the turbocharger equipment is well maintained and prompt to extend the equipment maintenance cycle; If not, the result of insufficient maintenance of the turbocharger equipment will be displayed, and a prompt will be given to shorten the equipment maintenance cycle.

[0022] By adopting the above technical solution, the performance change curve and the performance degradation curve are mapped and analyzed, which realizes the accurate evaluation of the equipment maintenance status and provides a scientific basis for the dynamic adjustment of the maintenance cycle. In addition, this solution effectively improves the scientificity, flexibility and economy of equipment maintenance, while improving the reliability and efficiency of equipment operation, reducing operation and maintenance costs, and providing important technical support for equipment management and maintenance.

[0023] In a second aspect, the present invention provides a turbocharger performance evaluation system, which adopts the following technical solution: A turbocharger performance evaluation system, comprising: Data collection module, including real-time data collection module and historical data collection module; A real-time data collection module is used to collect real-time data of the turbocharger; A historical data collection module is used to collect historical data of the turbocharger; Establish a model module to construct a performance degradation model based on the nonlinear Gamma process; A model training module, used for training the performance degradation model using historical data of the turbocharger to obtain a trained performance degradation model; A performance status output module, used to input real-time data into the trained performance degradation model and output the current performance status value of the turbocharger; A performance status evaluation module, used to determine whether the current performance status value of the turbocharger is lower than a preset alarm threshold; If the current performance status value of the turbocharger is lower than the preset alarm threshold, the result that the current performance status of the turbocharger is unqualified is displayed, and an abnormal turbocharger performance status alarm is given; If the current performance status value of the turbocharger is not lower than the preset warning threshold, then a result indicating that the current performance status of the turbocharger is qualified is displayed.

[0024] By adopting the above technical solution, a model module is established to construct a performance degradation model based on the nonlinear Gamma process, and the performance degradation process caused by equipment aging is incorporated into the performance evaluation system of the turbocharger. The performance status evaluation module realizes a dual evaluation of the current performance status and degradation trend of the turbocharger. This helps the staff to accurately grasp the equipment status, promptly discover potential problems caused by aging, avoid evaluation blind spots caused by ignoring aging effects, and improve the accuracy of turbocharger performance evaluation.

[0025] In summary, the present application includes at least one of the following beneficial technical effects: 1. Construct a performance degradation model based on the nonlinear Gamma process, incorporate the performance degradation process caused by equipment aging into the performance evaluation system of the turbocharger, and conduct a dual evaluation of the current performance status and degradation trend, which will help the staff to accurately grasp the equipment status, promptly discover potential problems caused by aging, and avoid evaluation blind spots caused by ignoring the aging effect. In addition, the performance degradation model based on the nonlinear Gamma process can effectively describe the performance degradation law of the turbocharger under different working conditions. Through the performance degradation model, this application can adapt to the evaluation needs of the turbocharger under different working conditions, capture the nonlinear characteristics of equipment performance degradation, and have stronger robustness, especially in scenarios of long-term continuous operation.

[0026] 2. The introduction of edge distribution processing and joint distribution fitting can capture data correlation under complex operating conditions, enabling the model to adapt to different operating conditions, especially in extreme environments or non-standard operating conditions, and still provide reliable performance evaluation results. In addition, generating simulated sample data as an extension of historical data reduces the reliance on high-frequency real-time data collection, helps reduce data collection costs, and optimizes data distribution coverage during model training, reducing the complexity of model training.

[0027] 3. By combining historical equipment data with real-time data, the similarity between the performance change curve and the performance degradation curve is calculated, and the aging or failure status of the equipment is judged accordingly, effectively improving the accuracy and intelligence level of equipment evaluation. In addition, the performance change curve and the performance degradation curve are mapped and analyzed to achieve an accurate assessment of the equipment maintenance status, providing a scientific basis for the dynamic adjustment of the maintenance cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flow chart of Example 1 of the present application; Figure 2 is a flow chart of the S6 performance status pre-evaluation of Example 1 of the present application; Figure 3 is a flow chart of S8 performance abnormality analysis of Example 1 of the present application; Figure 4 is a flow chart of the S9 performance qualification analysis of Example 1 of the present application; Figure 5 This is a flowchart of the S10 simulation model training of Example 2 of the present application. DETAILED DESCRIPTION

[0029] The following combination Figures 1 to 5 This application is described in further detail.

[0030] Embodiment 1: This embodiment discloses a method for evaluating the performance of a turbocharger, such as Figure 1 As shown, the evaluation method includes: collecting real-time data and historical data of the turbocharger, then building a performance degradation model based on the nonlinear Gamma process, using the historical data of the turbocharger to train the performance degradation model, obtaining the trained performance degradation model, and finally inputting the collected real-time data into the trained performance degradation model, outputting the current performance state value of the turbocharger, and judging whether the current performance state value of the turbocharger is above the preset alarm threshold. This embodiment includes the following steps: S1 data collection: includes S11 real-time data collection and S12 historical data collection.

[0031] S11 collects real-time data: collects real-time data of the turbocharger, including but not limited to turbine speed, intake pressure, exhaust pressure, vibration signal, intake temperature, exhaust temperature, intake flow, exhaust flow, metal particle concentration of lubricating oil, humidity, etc.

[0032] In this embodiment, the turbine speed of the turbocharger can be collected by a speed sensor installed on the turbine shaft, near the turbine blade or bearing position. The intake pressure of the turbocharger is collected by a pressure sensor installed at the air inlet of the turbocharger. The exhaust pressure of the turbocharger is collected by a pressure sensor installed at the air outlet of the turbocharger. The vibration signal of the turbocharger is collected by an accelerometer or vibration sensor installed at a vibration sensitive part near the turbocharger housing or bearing seat. The intake temperature is collected by a thermocouple or temperature sensor installed in the intake duct of the turbocharger, near the air inlet of the turbocharger. The exhaust temperature is collected by a temperature sensor installed in the exhaust duct at the exhaust port of the turbocharger, near the exhaust impeller area. The intake flow rate is collected by a flow sensor installed in the intake duct of the turbocharger. The intake flow rate is collected by a flow sensor installed in the exhaust duct of the turbocharger. The metal particle concentration of the lubricating oil is collected by a lubricating oil particle sensor installed in the lubricating oil return pipe or oil storage tank of the turbocharger.

[0033] In this embodiment, a single data is used to express the degradation process of the turbocharger performance, so the real-time data collected in this embodiment is any one of the turbine speed, intake pressure, exhaust pressure, vibration signal, intake temperature, exhaust temperature, intake flow, exhaust flow, lubricating oil metal particle concentration, and humidity.

[0034] S12 collects historical data: collects historical data of the turbocharger, which includes historical turbine speed, historical intake pressure, historical exhaust pressure, historical vibration signal, historical intake temperature, historical exhaust temperature, historical intake flow, historical exhaust flow, historical lubricating oil metal particle concentration, historical humidity and the total historical operation time of the turbocharger when each data is collected.

[0035] In this embodiment, the collected historical data is data corresponding to the collected real-time data and the total historical operation time of the turbocharger when it is collected.

[0036] In this embodiment, after collecting the real-time data and historical data of the turbocharger, the data needs to be preprocessed. The steps of data preprocessing include data cleaning, data normalization and feature extraction.

[0037] Data cleaning: De-noising, processing of missing values, and removal of outliers are performed on the collected real-time data and historical data.

[0038] Denoising includes using smoothing filters (such as sliding average) to eliminate small fluctuations in exhaust temperature, boost pressure, and turbine speed; using wavelet transform to decompose the vibration signal data, and reconstructing it after removing the noise component. Processing missing values ​​includes directly deleting data with missing values. Removing outliers includes using statistical analysis to eliminate outliers in the data and deleting detected outlier data.

[0039] Data normalization: Normalize the collected real-time data and historical data.

[0040] Feature extraction: Extract the time domain features (including mean, variance, skewness, kurtosis, peak, envelope features, etc.) and frequency domain features (main frequency, spectrum energy, harmonic features) of vibration signal data.

[0041] S2 Model building: Construct a performance degradation model based on the nonlinear Gamma process.

[0042] In this embodiment, the Gamma process model is used as the basic model of the performance degradation model.

[0043] S3 model training: Use the historical data of the turbocharger as a sample training set, use the sample training set to train the performance degradation model, and obtain the trained performance degradation model. It includes S31 loss function definition, S32 model reasoning, S33 loss calculation, S34 parameter update, and S35 iterative training.

[0044] In this embodiment, S3 model training includes the following: S31 loss function definition: Define the log-likelihood function as the loss function, and use the log-likelihood function to maximize the fit of the data.

[0045] .

[0046] in, represents the log-likelihood function, N represents the number of samples, Indicates The degradation data of samples, represents the probability density function of the Gamma distribution, represents the distribution parameter.

[0047] S32 model reasoning: The historical data of the turbocharger is used as a sample training set, and the sample training set is input into the performance degradation model. The performance degradation model predicts the degradation amount of the turbocharger based on the input sample and outputs the performance status value of the turbocharger.

[0048] S33 calculates the loss: uses the defined log-likelihood function to calculate the loss value of the performance degradation model.

[0049] S34 parameter update: Use the back propagation algorithm to calculate the model parameter gradient of the performance degradation model, and use the gradient descent method to update the model parameters of the performance degradation model.

[0050] S35 iterative training: loop through the above S32 model inference to S34 parameter update until the preset number of iterative training times is completed or the calculated loss value of the performance degradation model no longer decreases, completing the model training and obtaining the trained performance degradation model.

[0051] S4 Performance Status Output: Input real-time data into the trained performance degradation model and output the current performance status value of the turbocharger.

[0052] The real-time data is input into the trained performance degradation model. The model predicts the current degradation amount of the turbocharger. Through the initial performance state value of the turbocharger, the current performance state value of the turbocharger is calculated and output.

[0053] - f .

[0054] in, Indicates the current performance status of the turbocharger. Indicates the initial performance status value of the turbocharger. Indicates the current moment, represents the degradation amount of the turbocharger at the current moment, f represents the degradation function of the model, Represents the input features, that is, the input real-time data.

[0055] S5 predicted performance status output: includes S51 predicted performance status and S52 predicted performance status evaluation.

[0056] S51 predicts performance status: inputs the real-time data into the trained performance degradation model to predict the performance status value of the turbocharger at a certain point in the future, which is recorded as the predicted performance status value.

[0057] In this embodiment, the value at a certain time point in the future may be set according to actual needs to achieve prediction of the performance state value of the turbocharger in the future.

[0058] S52 Prediction performance status evaluation: determine whether the prediction performance status value is lower than a preset prediction performance threshold.

[0059] If so, the time point of the predicted performance status value is displayed to remind the user to maintain the turbocharger in time so that the turbocharger can maintain a certain performance and work normally when the user is expected to use the turbocharger.

[0060] If not, no processing is performed, indicating that the performance of the turbocharger remains good and can work normally when the user is expected to use the turbocharger.

[0061] S6 Performance status pre-assessment: including S61 setting initial threshold, S62 initial performance status assessment, S63 calculating performance degradation rate and S64 dynamically adjusting threshold, such as Figure 2 shown.

[0062] S61 sets an initial threshold: according to the turbocharger equipment standard, sets a first preset threshold, where the first preset threshold is a fixed threshold.

[0063] S62: Initial performance status assessment: determining whether the current performance status value of the turbocharger is lower than a first preset threshold.

[0064] If not, the current performance status of the turbocharger is displayed as excellent.

[0065] If so, it is displayed that the current performance status of the turbocharger is not an excellent result.

[0066] S63 calculates the performance degradation rate: uses the performance degradation model to output the performance state value of the turbocharger at the previous moment, and calculates the change rate of the performance value at the current moment and the previous moment.

[0067] S64 dynamically adjusts the threshold: calculates a preset performance status threshold, and records the preset performance status threshold as a preset alarm threshold. The preset alarm threshold is a dynamic threshold, and the corresponding alarm threshold can be adjusted according to the performance change of the turbocharger.

[0068] The calculation model of the preset performance status threshold is as follows: .

[0069] in, Indicates the current moment, Indicates the preset alarm threshold at the current moment. represents the first preset threshold, represents the adjustment factor, , Indicates the rate of change of the performance value between the current moment and the previous moment.

[0070] In this embodiment, based on the experimental calibration of turbocharger operation data, combined with the optimization of false alarm rate and missed alarm rate, a series of candidate values, and evaluate the calculation model of the preset performance status threshold value in different The performance under different values ​​is selected to find the adjustment coefficient with the best performance under different working conditions. .

[0071] S7 Performance Status Evaluation: Determine whether the current performance status value of the turbocharger is lower than the preset alarm threshold.

[0072] If so, the current performance status of the turbocharger is displayed as unqualified, and an abnormal performance status alarm of the turbocharger is issued, and S8 performance abnormality analysis is performed.

[0073] If not, the current performance status of the turbocharger is displayed as qualified, and S9 performance qualification analysis is performed.

[0074] S8 Performance Abnormal Analysis: including S81 first processing, S82 second processing, S83 third processing and S84 first judgment, such as Figure 3 shown.

[0075] S81 First processing: inputting historical data into the performance degradation model, predicting all performance degradation data of the turbocharger throughout its life cycle, and using the obtained performance degradation data to draw a performance degradation curve, recorded as a first curve, wherein the coordinates of the first curve are respectively the total operating time of the turbocharger and the corresponding performance status value of the turbocharger.

[0076] S82 Second process: Record a period of time from the current time point to the front as a first time period, collect the data of the turbocharger in the first time period, and record it as first data. In this embodiment, the length of the first time period can be set according to actual needs.

[0077] S83: The third process: input the first data into the performance degradation model to obtain all the performance degradation data in the first time period, recorded as the first performance data, and draw a performance change curve according to the first performance data, recorded as the second curve.

[0078] S84 first judgment: slide the first curve with the length of the second curve as the window, take out a window each time, and obtain several windows of the first curve. Calculate the curve similarity between each window and the second curve by Euclidean distance, cosine similarity or dynamic time warping, and determine whether there is a window with a curve similarity greater than a preset curve similarity threshold.

[0079] If it exists, the serious aging result of the turbocharger equipment will be displayed, and the turbocharger equipment maintenance alarm will be issued to remind the user to carry out maintenance in time.

[0080] If it does not exist, the turbocharger equipment failure result will be displayed, and the turbocharger equipment failure alarm will be issued to remind the user to perform fault inspection and repair in time.

[0081] S9 Performance Qualification Analysis: including S91 fourth processing, S92 fifth processing, S93 sixth processing and S94 second judgment, such as Figure 4shown.

[0082] S91 Fourth processing: inputting historical data into the performance degradation model to obtain the performance degradation data of the turbocharger, and using the performance degradation data to draw a performance degradation curve, recorded as the third curve, the coordinates of the third curve are the total operating time of the turbocharger and the corresponding performance state value of the turbocharger.

[0083] S92 Fifth process: Record a period of time from the current time point to the front as a second period of time, collect the data of the turbocharger in the second period of time, and record it as second data. In this embodiment, the length of the second period of time can be set according to actual needs.

[0084] S93 Sixth processing: input the second data into the performance degradation model to obtain all performance degradation data in the second time period, record them as second performance data, and draw a performance change curve based on the second performance data, record them as a fourth curve.

[0085] S94 Second judgment: Map the horizontal axis value of the fourth curve to the horizontal axis range of the third curve so that the two have the same horizontal axis scale. If the horizontal axis ranges of the two are different, scale the horizontal axis range of the fourth curve until the two have the same horizontal axis scale. At this time, ensure that the vertical axis range of the fourth curve is consistent with the vertical axis range of the third curve. If not, linearly normalize the vertical axis value of the fourth curve and map it to the vertical axis range of the third curve to complete the mapping. It will be judged whether the value of the lowest point in the fourth curve is higher than the value of the highest point of the third curve.

[0086] If so, it will show that the turbocharger equipment is well maintained, and will prompt the user to extend the equipment maintenance cycle, suggesting that the user extend the equipment maintenance cycle.

[0087] If not, the result of insufficient maintenance of the turbocharger equipment will be displayed, and a prompt will be given to shorten the equipment maintenance cycle. The user is advised to extend or shorten the maintenance cycle and perform maintenance in time.

[0088] This embodiment constructs a performance degradation model based on the nonlinear Gamma process, incorporates the performance degradation process caused by equipment aging into the performance evaluation system of the turbocharger, and performs a dual evaluation of the current performance status and degradation trend. This helps the staff to accurately grasp the equipment status and promptly discover potential problems caused by aging, avoids evaluation blind spots caused by ignoring aging effects, and improves the accuracy of turbocharger performance evaluation.

[0089] Example 2: Figure 5 As shown, the difference between this embodiment and embodiment 1 is that if the processed turbocharger real-time data is multiple data, after executing the step of S3 model training and before executing the step of S4 performance status output, it also includes: S10 simulation model training: including S101 edge distribution processing, S102 similarity screening, S103 generating joint distribution, S104 generating simulated sample data, S105 global verification, S106 local verification and S107 simulation training.

[0090] S101 Marginal distribution processing: select the appropriate marginal distribution type for each historical data, use the maximum likelihood estimation method to estimate the marginal distribution parameters of each historical data, and convert each data into a marginal distribution probability through the cumulative distribution function.

[0091] S102 Similarity screening: Calculate the similarity between each historical data sequence through methods such as dynamic time warping, cosine similarity or Euclidean distance, compare the calculated similarity with the preset similarity filter threshold, and screen out historical data with a similarity greater than the preset similarity filter threshold as new historical data.

[0092] S103 generates a joint distribution: takes the marginal distribution probability of the historical data as input, uses the maximum likelihood estimation method to fit the Copula function, so as to fit the dependency relationship between the various historical data and generate a joint distribution.

[0093] S104 generates simulated sample data: samples the marginal distribution probabilities of several simulated samples from the generated joint distribution, and uses the inverse cumulative distribution function of the marginal distribution to map the sampling points of the several simulated samples back to the space of historical data to obtain simulated sample data.

[0094] S105 Global Verification: Use Wasserstein distance to calculate the global distribution difference between historical data and simulated sample data, and determine whether the calculated Wasserstein distance is greater than a preset Wasserstein distance threshold.

[0095] If yes, the marginal distribution type and distribution parameters of the historical data are adjusted, and the marginal distribution processing step S101 is executed.

[0096] If not, execute the local verification step S106.

[0097] S106 Local Verification: Use the kernel density estimation function to perform kernel density estimation on the historical data and the generated simulated sample data respectively. In each local area, calculate the similarity of the two kernel density functions to determine whether the similarity between the two in each local area is greater than the preset similarity threshold.

[0098] If yes, execute step S107 of simulation training.

[0099] If not, adjust the Copula function type and execute S103 to generate a joint distribution.

[0100] S107 simulation training: input the simulated sample data into the trained performance degradation model for model training, output the performance status value of the turbocharger, use the log-likelihood function to maximize the data fit, use the back-propagation algorithm to calculate the model parameter gradient, and update the model parameters through the gradient descent method, and execute model reasoning and parameter update in a loop until the preset number of training rounds is reached or the loss value converges, and obtain the performance degradation model after simulated training as the new trained performance degradation model.

[0101] In this embodiment, the historical data of the turbocharger is a plurality of data, including historical turbine speed, historical intake pressure, historical exhaust pressure, historical vibration signal, historical intake temperature, historical exhaust temperature, historical intake flow, historical exhaust flow, historical lubricant metal particle concentration, historical humidity, and the total historical operation time of the turbocharger when each data item is collected. The historical turbine speed, historical intake pressure, historical exhaust pressure, historical vibration signal, historical intake temperature, historical exhaust temperature, historical intake flow, historical exhaust flow, historical lubricant metal particle concentration, and historical humidity under the total historical operation time of the same device and the same turbocharger are a historical data time series.

[0102] This application introduces marginal distribution processing and joint distribution fitting to capture data correlations under complex operating conditions, enabling the model to adapt to different operating conditions, especially in extreme environments or non-standard operating conditions, and still provide reliable performance evaluation results.

[0103] For example, researchers collected data on a turbocharger at a certain moment, including turbine speed, intake pressure, exhaust pressure, vibration signal, intake temperature, exhaust temperature, intake flow, exhaust flow, lubricating oil metal particle concentration, and humidity, and evaluated the performance status of the turbocharger at the current moment. The specific steps are as follows: 1. Collect the performance degradation trend of the turbocharger in the past year as the historical data of the turbocharger.

[0104] 2. Data processing: The collected historical data of the turbocharger are cleaned, normalized and features are extracted to obtain processed historical data.

[0105] 3. Model construction: Construct a performance degradation model based on the nonlinear Gamma process.

[0106] 4. Use the processed historical data and generated simulation data to train the model.

[0107] Input the processed historical data into the performance degradation model for model training. After the model training is completed, select an appropriate marginal distribution type for each piece of historical data, perform marginal distribution processing to obtain the marginal distribution probabilities of the historical data. After similarity screening, use the marginal distribution probabilities of the selected historical data as input, fit the dependence relationship between each historical data through the Copula function to generate a joint distribution, sample from the joint distribution to obtain simulated joint distribution probability values, convert the simulated joint distribution probability values back to the space of the historical data to generate simulated sample data, input the simulated sample data into the trained performance degradation model for model training, and obtain the simulated and trained performance degradation model as the new trained performance degradation model.

[0108] 5. Input the collected data into the trained performance degradation model to obtain the current performance state value of the turbocharger and conduct an evaluation.

[0109] Prediction and evaluation: Obtain the current performance state value as a, predict the predicted performance state value b of the turbocharger in the next three months, and preset the prediction performance threshold as c. If b < c, it indicates that the turbocharger needs maintenance after the next three months, and preparations for early maintenance should be made; if b > c, it indicates that the turbocharger equipment is normal in the next three months.

[0110] Current evaluation: Set the first preset threshold as d and the preset alarm threshold as e.

[0111] If a ≥ d, it proves that the current performance state of the turbocharger is excellent and no maintenance or inspection is required. If a < d, it proves that the current performance state of the turbocharger is not excellent, and further evaluation is carried out. If a < e, it indicates that the current performance state of the turbocharger is unqualified, then draw the data of the turbocharger recorded 15 days forward from the current moment and draw the performance change curve of the turbocharger. Compare the similarity between the drawn performance change curve and the performance degradation curve of the turbocharger. If the curve similarity is less than or equal to the preset curve similarity threshold, it proves that the turbocharger has a fault and needs to be repaired; if the curve similarity is greater than the preset curve similarity threshold, it proves that the turbocharger has normal aging and needs maintenance.

[0112] If a ≥ e, it indicates that the current performance state of the turbocharger is qualified. Draw the data of the turbocharger recorded 15 days forward from the current moment and draw the performance change curve of the turbocharger. Map the drawn performance change curve in the coordinate system of the performance degradation curve. If the performance change curve is located above the upper bound of the performance degradation curve, it indicates that the turbocharger equipment is well maintained, and it is recommended that the user extend the equipment maintenance cycle; if the performance change curve is located below the lower bound of the performance degradation curve, it indicates that the turbocharger equipment is insufficiently maintained, and it is recommended that the user shorten the equipment maintenance cycle.

[0113] Embodiment 3: This embodiment discloses a turbocharger performance evaluation system, the evaluation system comprising: The data acquisition module includes a real-time data acquisition module and a historical data acquisition module.

[0114] The real-time data collection module is used to collect real-time data of the turbocharger.

[0115] The historical data collection module is used to collect historical data of the turbocharger.

[0116] A model module is built to construct a performance degradation model based on the nonlinear Gamma process.

[0117] The model training module is connected to the model building module and the data acquisition module for training the performance degradation model using the historical data of the turbocharger to obtain the trained performance degradation model.

[0118] The performance status output module is used to input real-time data into the trained performance degradation model and output the current performance status value of the turbocharger.

[0119] The performance status evaluation module is used to determine whether the current performance status value of the turbocharger is lower than a preset alarm threshold.

[0120] If the current performance status value of the turbocharger is lower than the preset alarm threshold, the result that the current performance status of the turbocharger is unqualified is displayed, and an abnormal turbocharger performance status alarm prompt is issued, triggering the performance abnormality analysis module.

[0121] If the current performance status value of the turbocharger is not lower than the preset alarm threshold, then the result that the current performance status of the turbocharger is qualified is displayed, triggering the performance qualification analysis module.

[0122] The performance anomaly analysis module is respectively communicated with the data acquisition module, the model building module, the model training module, the performance status output module and the performance status evaluation module, and includes a first processing subunit, a second processing subunit, a third processing subunit and a first judgment subunit.

[0123] The first processing subunit is used to input the historical data into the performance degradation model to obtain the performance degradation data of the turbocharger, and to draw a performance degradation curve using the performance degradation data, which is recorded as a first curve.

[0124] The second processing subunit is used to record a period of time from the current time point to the front as a first time period, and collect data of the turbocharger in the first time period as first data.

[0125] The third processing subunit is used to input the first data into the performance degradation model to obtain the first performance data, and draw a performance change curve according to the first performance data, which is recorded as the second curve.

[0126] The first judgment subunit is used to calculate the curve similarity between the first curve and the second curve, and judge whether the curve similarity is greater than a preset curve similarity threshold.

[0127] If the curve similarity is greater than a preset curve similarity threshold, a serious aging result of the turbocharger equipment is displayed, and a turbocharger equipment maintenance alarm is issued.

[0128] If the curve similarity is not greater than the preset curve similarity threshold, a turbocharger device failure result is displayed, and a turbocharger device failure alarm is issued.

[0129] The performance qualification analysis module is respectively communicated with the data acquisition module, the model building module, the model training module, the performance status output module and the performance status evaluation module, and includes a fourth processing subunit, a fifth processing subunit, a sixth processing subunit and a second judgment subunit.

[0130] The fourth processing subunit is used to input the historical data into the performance degradation model to obtain the performance degradation data of the turbocharger, and to draw a performance degradation curve using the performance degradation data, which is recorded as the third curve.

[0131] The fifth processing sub-unit is used to record a period of time from the current time point to the front as a second time period, and collect data of the turbocharger in the second time period as second data.

[0132] The sixth processing subunit is used to input the second data into the performance degradation model, obtain all performance degradation data in the second time period, record them as second performance data, and draw a performance change curve according to the second performance data, record them as a fourth curve.

[0133] The second judging subunit is used to map the fourth curve into the coordinate system of the third curve, and judge whether the fourth curve is at the upper limit of the third curve.

[0134] If the fourth curve is at the upper limit of the third curve, it indicates that the turbocharger equipment is well maintained, and a reminder to extend the equipment maintenance period is given.

[0135] If the fourth curve is not within the upper limit of the third curve, then it is displayed that the turbocharger equipment is under-maintained, and a prompt is given to shorten the equipment maintenance cycle.

[0136] In this embodiment, a performance degradation model based on the nonlinear Gamma process is constructed by establishing a model module, the performance degradation process caused by equipment aging is incorporated into the performance evaluation system of the turbocharger, and a dual evaluation of the current performance status and degradation trend of the turbocharger is achieved through the performance status evaluation module. This helps the staff to accurately grasp the equipment status, promptly discover potential problems caused by aging, avoid evaluation blind spots caused by ignoring aging effects, and improve the accuracy of turbocharger performance evaluation.

[0137] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A turbocharger performance evaluation method, characterized in that: include: Data collection: including real-time data collection and historical data collection; Collect real-time data: Collect real-time data of turbocharger; Collect historical data: Collect historical data of turbocharger; Model building: Construct a performance degradation model based on the nonlinear Gamma process; Model training: Use the historical data of the turbocharger to train the performance degradation model to obtain the trained performance degradation model; Performance status output: input the real-time data into the trained performance degradation model and output the current performance status value of the turbocharger; Performance status evaluation: Determine whether the current performance status value of the turbocharger is lower than the preset alarm threshold: If yes, then the result that the current performance status of the turbocharger is unqualified is displayed, and an abnormal performance status alarm of the turbocharger is given; If not, the result indicating that the current performance status of the turbocharger is acceptable is displayed.

2. The turbocharger performance evaluation method according to claim 1, characterized in that: After executing the model training step and before executing the performance status output step, it also includes: Marginal distribution processing: select the marginal distribution type for the historical data, use the maximum likelihood estimation method to fit the distribution parameters, perform marginal distribution processing on the historical data, and obtain the marginal distribution probability of the historical data; Generate joint distribution: Input the marginal distribution probability of historical data into the Copula function, and generate joint distribution by fitting the dependency structure of historical data; Generate simulated sample data: sample from the joint distribution to obtain simulated joint distribution probability values, convert the simulated joint distribution probability values ​​back to the space of historical data, and generate simulated sample data; Simulation training: Input the simulated sample data into the trained performance degradation model for simulation training to obtain the simulated trained performance degradation model as a new trained performance degradation model.

3. The turbocharger performance evaluation method according to claim 2, characterized in that: After executing the step of building the model and before executing the step of joint distribution processing, the method further includes: Similarity screening: Calculate the similarity between historical data, compare the similarity with the preset similarity filtering threshold, and screen out historical data with similarity greater than the preset similarity filtering threshold as new historical data.

4. The turbocharger performance evaluation method according to claim 2, characterized in that: After executing the step of generating simulated sample data and before executing the step of simulated training, the method further includes: Global verification: Calculate the Wasserstein distance between historical data and generated simulated sample data, and determine whether the calculated Wasserstein distance is greater than the preset Wasserstein distance threshold: If so, the marginal distribution type and distribution parameters of the historical data are adjusted, and the marginal distribution processing step is performed; If not, then perform the local verification step; Local verification: Use the KernelDensity function to perform kernel density estimation on the historical data and the generated simulated sample data, compare the density functions of the two, and determine whether the similarity of the two in each local area is greater than the preset similarity threshold: If yes, then execute the simulation training steps; If not, adjust the Copula function type and execute the steps to generate a joint distribution.

5. The turbocharger performance evaluation method according to claim 1, characterized in that: After the step of outputting the performance status, and before the step of evaluating the performance status, the step further includes: Predicting performance status: inputting real-time data into the trained performance degradation model to predict the performance status value of the turbocharger at a certain point in the future, which is recorded as the predicted performance status value; Prediction performance status evaluation: Determine whether the prediction performance status value is lower than the preset prediction performance threshold: If so, the time point at which the predicted performance status value is displayed; If not, no processing is performed.

6. The turbocharger performance evaluation method according to claim 1, characterized in that: After the step of outputting the performance status and before the step of evaluating the performance status, the step further includes: Setting an initial threshold: setting a first preset threshold according to a turbocharger equipment standard; Initial performance status assessment: Determine whether the current performance status value of the turbocharger is lower than the first preset threshold: If not, the result indicating that the current performance status of the turbocharger is excellent is displayed; If so, it is displayed that the current performance status of the turbocharger is not excellent.

7. The turbocharger performance evaluation method according to claim 6, characterized in that: After the step of performing the initial performance status assessment and before the step of performing the performance status assessment, the step further includes: Calculating the performance degradation rate: using the performance degradation model to output the performance state value of the turbocharger at the previous moment, and calculating the rate of change of the performance value at the current moment and the previous moment; Dynamically adjust the threshold: calculate the preset performance status threshold, record the preset performance status threshold as the preset alarm threshold, and the calculation model of the preset performance status threshold is as follows: ； in, Indicates the current moment, Indicates the preset alarm threshold at the current moment. represents the first preset threshold, represents the adjustment factor, Indicates the rate of change of the performance value between the current moment and the previous moment.

8. The turbocharger performance evaluation method according to claim 1, characterized in that: After executing the step of providing an abnormal turbocharger performance status warning prompt, the method further includes: First processing: inputting the historical data into the performance degradation model to obtain the performance degradation data of the turbocharger, and using the performance degradation data to draw a performance degradation curve, which is recorded as a first curve; The second process: a period of time before the current time point is recorded as a first period of time, and the data of the turbocharger in the first period of time is collected and recorded as first data; The third process: inputting the first data into the performance degradation model to obtain the first performance data, and drawing a performance change curve according to the first performance data, which is recorded as the second curve; First judgment: Calculate the curve similarity between the first curve and the second curve, and judge whether the curve similarity is greater than a preset curve similarity threshold: If so, it will display the serious aging result of the turbocharger equipment and issue a turbocharger equipment maintenance alarm; If not, the turbocharger equipment failure result is displayed, and a turbocharger equipment failure alarm is issued.

9. The turbocharger performance evaluation method according to claim 1, characterized in that: After the step of displaying the qualified result of the current performance status of the turbocharger is performed, the method further includes: Fourth process: inputting the historical data into the performance degradation model to obtain the performance degradation data of the turbocharger, and using the performance degradation data to draw a performance degradation curve, which is recorded as the third curve; The fifth process: recording a period of time before the current time point as a second period of time, collecting data of the turbocharger in the second period of time, and recording it as second data; Sixth processing: inputting the second data into the performance degradation model to obtain second performance data, and drawing a performance change curve according to the second performance data, which is recorded as a fourth curve; Second judgment: Map the fourth curve in the coordinate system of the third curve to determine whether the fourth curve is within the upper bound of the third curve: If yes, it will show that the turbocharger equipment is well maintained and prompt to extend the equipment maintenance cycle; If not, the result of insufficient maintenance of the turbocharger equipment will be displayed, and a prompt will be given to shorten the equipment maintenance cycle.

10. A turbocharger performance evaluation system, the system being applicable to the method according to any one of claims 1 to 9, characterized in that: include: Data collection module, including real-time data collection module and historical data collection module; A real-time data collection module is used to collect real-time data of the turbocharger; A historical data collection module is used to collect historical data of the turbocharger; Establish a model module to construct a performance degradation model based on the nonlinear Gamma process; A model training module, used for training a performance degradation model using historical data of a turbocharger to obtain a trained performance degradation model; A performance status output module, used to input real-time data into the trained performance degradation model and output the current performance status value of the turbocharger; The performance status evaluation module is used to determine whether the current performance status value of the turbocharger is lower than the preset alarm threshold: If the current performance status value of the turbocharger is lower than the preset alarm threshold, the result that the current performance status of the turbocharger is unqualified is displayed, and an abnormal turbocharger performance status alarm is given; If the current performance status value of the turbocharger is not lower than the preset warning threshold, then a result indicating that the current performance status of the turbocharger is qualified is displayed.

Images