A method for predicting the service life of an engine cylinder head based on the cumulative damage of the driving state profile

Abstract

The present invention discloses a method for predicting the service life of an engine cylinder head based on the cumulative damage of the driving state profile. This method first obtains the on-vehicle variable data during the entire operation of the vehicle, and divides the driving segments of the vehicle according to the vehicle speed and engine speed. Secondly, eigenvalue extraction and principal component analysis are performed on the data of all driving segments. With the help of the clustering method, a driving state profile is constructed. Further, a cumulative damage model is constructed to calculate the cylinder head life degradation factor of each driving state profile. Finally, the cumulative damage and remaining service life values of the cylinder head of the vehicle to be predicted are obtained through the linear cumulative method. This method is easy to implement, fully considers the influence of the vehicle driving conditions on the life damage, can efficiently, accurately and comprehensively realize the prediction of the cylinder head life, effectively solves the problems of low prediction efficiency and low prediction accuracy of the traditional life prediction method based on the mechanism model, and provides technical support for the condition-based maintenance of the vehicle and the operation management of the enterprise.

Description

Technical Field

[0001] The present invention relates to the technical field of life analysis in the field of mechanical engineering, and particularly relates to a method for predicting the life of an engine cylinder head based on cumulative damage of a driving state profile. Background Art

[0002] As the core part of a vehicle, the engine is the source of vehicle power, and its operating stability and reliability directly affect the operating conditions of the vehicle. The engine cylinder head is an important part of the engine. During the operation of the engine, it bears various complex loads such as high temperature, high pressure, impact and vibration, and is prone to problems such as fatigue cracking and deformation deterioration. In order to avoid the adverse consequences caused by the failure of the cylinder head, it is necessary to conduct life assessment and maintenance management on it.

[0003] There may be great differences in the usage environments and usage methods of different users, which will have an important impact on the service life of the engine cylinder head. For example, some users may drive at high load for a long time on the highway, while some users often drive at low speed on urban congested roads. This difference will lead to great differences in the load, temperature, vibration and other conditions borne by the engine. In addition, there may also be differences in the maintenance levels of different users for the engine in daily use, such as changing engine oil, cleaning the radiator, and regularly replacing the filter, which will also affect the life of the engine cylinder head. The service life of engines of the same batch varies greatly under the use of different users. It is of great significance to statistically analyze the usage data of different user groups to understand the usage conditions and failure characteristics of the engine under different conditions, so as to further improve the product design and maintenance management strategies.

[0004] Existing life prediction methods mainly rely on experimental verification and empirical formulas. By establishing a finite element model of the cylinder head, simulating the load conditions endured by the cylinder head structure during actual operation, loading it accordingly, and combining finite element theory to conduct transient dynamic analysis on the structure, dynamic response data such as the local load-time history of each point on the structure can be obtained. Subsequently, the fatigue life of the cylinder head is predicted and the reliability of the structure is evaluated. For example, Patent CN111896361A provides a method for predicting the thermo-mechanical fatigue life of an engine cylinder head based on the energy method, calculating the thermo-mechanical fatigue life of the firing surface of the cylinder head through a finite element model calibrated by experiments and measured material boundaries. Patent CN114912211A provides a method for predicting the fatigue life of a cylinder head based on the energy method and an equivalent component model, completing the thermo-mechanical fatigue life analysis based on the energy method theory by extracting the load spectrum curve under service conditions and importing the thermo-mechanical load spectrum obtained from finite element simulation. Patent CN109598079A provides a method for predicting the fatigue life of different zones of a cylinder head, partitioning the cylinder head according to the failure modes of different zones, and using different fatigue damage and life prediction models for specific failure modes to predict the fatigue life of the cylinder head in different zones. Existing remaining life prediction methods based on mechanism models require accurate description of parameters such as the load and boundary conditions of the structure, and fail to incorporate real-time monitoring data of actual vehicles when predicting the life, making it difficult to accurately reflect the actual state of the vehicle's current operation. Especially when the actual operating conditions of the vehicle change frequently, if real-time monitoring data cannot be used to update the model, large prediction deviations will occur. However, the changes in the actual operating conditions of the vehicle are usually complex, so it is difficult to accurately describe them. Therefore, there is a need to provide an efficient, accurate, and comprehensive method for predicting the life of an engine cylinder head to meet the needs of engine manufacturers and users. Summary of the Invention

[0005] The object of the present invention is to overcome the deficiencies of the prior art and provide a method for predicting the life of an engine cylinder head based on the cumulative damage of the driving state profile. By integrating in-vehicle data mining technology and fatigue damage mechanism models, it analyzes and evaluates the life damage of the diesel engine cylinder head under different working conditions and mission profiles, and accurately predicts the service life of the engine. Its characteristics are as follows:

[0006] Step 1: During the vehicle driving process, through sensors and recorder devices installed on the vehicle, collect, monitor, and record various vehicle operation data, summarize and analyze the vehicle operation data, and screen out in-vehicle network general variables related to the cylinder head life, including torque, throttle opening, exhaust gas flow, ambient pressure, cumulative mileage, oil temperature, gear position, vehicle speed, water temperature, ambient temperature, engine speed, and engine cumulative operation time; process and clean the full life cycle data of N vehicles collected, including data calibration, outlier processing, and missing value filling.

[0007] Step 2: Divide the driving segments of the vehicle according to the vehicle speed and engine speed in the in-vehicle network variables.

[0008] Step 2.1: Define the data segment from the vehicle speed starting from 0 km / h to greater than 0 km / h and then back to 0 km / h as one motion segment.

[0009] Step 2.2: Define the data segment where the vehicle state maintains a vehicle speed of 0 km / h and the engine speed is greater than 0 as one idling segment.

[0010] Step 2.3: One motion segment and an adjacent idling segment form one driving segment. Segment the vehicle driving data of the in-vehicle terminals of N vehicles throughout their life cycles according to Step 2.1 and Step 2.2, delete the segments with a short trip duration less than 10 s, and screen out the driving segments that meet the requirements to form a driving segment set.

[0011] Step 3: Feature value extraction and principal component analysis of the driving state segment variables.

[0012] Step 3.1: Based on the driving segment set of N vehicles, select the characteristic parameters representing the driving state characteristics and the characteristic parameters representing the driving state distribution characteristics as the evaluation parameters for a single short driving segment, and calculate the characteristic parameters of all driving segment sets. Suppose there are m driving segment samples X = [x (1) ,x (2) ,…x (i) …,x (m) , where x (i) represents the i-th driving segment sample, which contains n characteristics of the in-vehicle network general variables, that is where T is the transpose symbol, represents the j-th characteristic of the i-th driving segment sample.

[0013] Step 3.2: Centralize the original data samples, that is, subtract the mean of each characteristic in all samples from each characteristic That is: to obtain the decentralized driving parameter sample data

[0014] Step 3.3: Calculate the covariance matrix of the driving parameter sample data Perform eigenvalue decomposition on the covariance matrix, calculate the eigenvalues of the covariance matrix C, and arrange them in descending order, denoted as {λ1, λ2, … λ τ …, λ p}, where λ1 ≥ λ2 ≥ … ≥ λ p , and the corresponding eigenvectors are denoted as {e1, e2, … e τ …, e p}, where eτ is the τ-th eigenvalue λ τ corresponding eigenvector, and p is the number of eigenvalues.

[0015] Step 3.4: For the v-th principal component, calculate the contribution rate of the principal component Calculate the cumulative value of the contribution rates of the first v principal components where λ v is the v-th eigenvalue. Take the first v eigenvalues with the cumulative contribution rate CR > ε, where ε is the set cumulative contribution rate, usually ε ∈ (0.85, 0.9]. The larger the value of ε, the more characteristic information is desired to be retained.

[0016] Step 3.5: Select the eigenvectors corresponding to the first v largest eigenvalues to form the projection matrix W, i.e., W = [e1, e2, …, e v , and multiply the original driving parameter sample data by the projection matrix W to obtain the new reduced-dimensional driving parameter matrix Y, i.e.,

[0017] Step Four: Construct the driving state profile.

[0018] Step 4.1: Based on the driving parameter matrix Y calculated in Step 3.5, denoted as Y = [y (1) , y (2) , … y (i) …, y (m) , where each is a v-dimensional vector, where represents the real number field;

[0019] Step 4.2: Randomly select k clustering centroid points as {C1, C2, …, C k}, where 1 < k ≤ m, and each is a v-dimensional vector.

[0020] Step 4.3: Calculate the Euclidean distance from each object to each clustering center where represents the t-th characteristic parameter of the i-th object, 1 ≤ i ≤ m, 1 ≤ t ≤ v; C jt is the t-th characteristic parameter of the j-th clustering center, 1 ≤ j ≤ k.

[0021] Step 4.4: Compare the distances from each driving segment to each clustering center in turn, and assign the driving segments to the clusters of the nearest clustering center to obtain k driving state clusters.

[0022] Step 4.5: For each driving state cluster, recalculate its centroid and use it as the new k clustering centroid points. Repeat steps 4.3 to 4.5 until the driving state clustering centroid remains unchanged, obtaining k driving state clusters, which represent k driving state profiles.

[0023] Step Five: Calculate the cylinder head life degradation factor of the driving state profile.

[0024] Step 5.1: Classify the driving segments of each vehicle into driving state profiles, obtaining the driving state profiles experienced during the vehicle's full life cycle and the cumulative driving mileage under each driving state profile.

[0025] Step 5.2: Construct a cumulative damage model. Each profile causes a certain degree of damage to the cylinder head life. The damage caused by each profile experienced during the engine operation accumulates. When the accumulated damage equals 1, the cylinder head will undergo fatigue failure. Construct the cumulative damage model where S is the mileage matrix, is the cumulative driving mileage of the j-th vehicle under the ξ-th driving state profile, and the driving mileage under the unexperienced driving state profile is assigned 0; D is the cylinder head life degradation factor matrix, D ξ is the cylinder head life degradation factor per unit mileage of the vehicle under the ξ-th driving state profile; b is the constant coefficient matrix, and k is the total number of driving state profiles.

[0026] Step 5.3: Solve the cumulative damage model in step 5.2 to obtain the least squares solution D = (S T S) -1 S T b, which is the cylinder head life degradation factor of each driving state profile.

[0027] Step Six: Construct a cylinder head life prediction model.

[0028] Step 6.1: Obtain the historical driving data of the vehicle whose cylinder head life is to be predicted. Segment the historical driving segments of the vehicle according to step two, and classify the historical driving segment set of the vehicle to be predicted into driving state profiles. Assume that the vehicle to be predicted has currently experienced w driving state profiles, and the cumulative driving mileage under each driving state profile is Determine the life degradation factors of the w driving state profiles according to the cylinder head life degradation factors of the driving state profiles obtained in step five, denoted as D hp = [D1, D2, …, D w .

[0029] Step 6.2: Calculate the historical cumulative damage degree up to the current time according to the formula

[0030] Step 6.3, set the future driving states to include all the experienced historical state profiles, and allocate the driving mileage under each future driving state profile according to the historical state profile ratio, and construct a cylinder head life prediction model. Where S c is the remaining drivable mileage before the cylinder head fails, and the predicted result of the remaining service life of the cylinder head is obtained by calculation. Brief Description of the Drawings

[0031] Figure 1 is a schematic flow chart of a method for predicting the life of an engine cylinder head based on the cumulative damage of the driving state profile proposed by the present invention;

[0032] Figure 2 is a schematic diagram showing that a motion segment and an adjacent idle segment form a driving segment;

[0033] Figure 3 is a schematic diagram for comparing different driving segments;

[0034] Figure 4 is a schematic diagram showing the changes of various parameters under a driving segment;

[0035] Figure 5 is the life prediction result of the cylinder heads of 7 vehicles. Detailed Embodiments

[0036] For the purpose, technical solution and advantages in production practice of the present invention to be presented better, it is necessary to further and more clearly explain and illustrate the present invention in combination with production examples and relevant drawings. Obviously, the specific implementation examples described in the text are only a part of the specific implementation examples of the present invention, rather than all the implementation examples. Based on the content of the present invention, all other implementation examples obtained by those skilled in the relevant art through interpreting the present invention without creative labor belong to the protection scope of the present invention.

[0037] The following combines the drawings to detail the specific technical solutions provided by the implementation examples of the present application.

[0038] The technical solution adopted by the present invention to solve the above problems is: a method for predicting the life of an engine cylinder head based on the cumulative damage of the driving state profile, and the method flow is shown in the attached Figure 1 . The driving data used in this technical method is collected, monitored and recorded for various vehicle operation data during vehicle driving through sensors and recorders installed on the vehicle, and the vehicle operation data of each vehicle is summarized and analyzed.

[0039] The first step, acquisition and processing of on-vehicle data:

[0040] Based on vehicle networking technology, using the vehicle driving data of in-vehicle terminals, summarize and analyze 88 types of operating data of each vehicle, screen out the general variables of the vehicle network related to the cylinder head life, as shown in the following table, and process and clean the data, including data correction, outlier processing, and missing value filling.

[0041]

[0042] Second step, driving segment division:

[0043] 1) Define the data segment from the vehicle speed starting from 0 km / h to greater than 0 km / h and then back to 0 km / h as one motion segment. Define the data segment where the vehicle state maintains a vehicle speed of 0 km / h and the engine speed is greater than 0 as one idle segment. See the appendix for segment division Figure 2 , and see the appendix for the comparison chart of the operating states of different driving segments Figure 3 .

[0044] 2) One motion segment and an adjacent idle segment form one driving segment. Divide the in-vehicle terminal vehicle driving data of the vehicle's entire life cycle into segments, calculate and screen out the qualified short driving segments to form a driving segment set. See the appendix Figure 4 .

[0045] Third step, eigenvalue extraction and principal component analysis of driving state segment variables:

[0046] 1) Select 30 characteristic parameters of driving state segments, including characteristic parameters representing driving state characteristics and characteristic parameters representing driving state distribution characteristics, as shown in the following table.

[0047]

[0048] 2) Calculate the characteristic parameters of all driving segment sets, perform centering processing on the original data samples, calculate the covariance matrix, eigenvalues, and eigenvectors of the driving parameter sample data, calculate the contribution rate of the principal components. In this example, take the cumulative contribution rate ε = 0.9, and select 11 principal component parameters with a cumulative contribution rate greater than 90% as the characteristic variables for clustering analysis.

[0049] Fourth step, construct the driving state profile:

[0050] Use the principal component parameters of all driving segment sets to perform clustering analysis on the driving state profile, and obtain the maximum silhouette coefficient when k = 4. Cluster the driving segments into 4 driving state clusters, that is, representing 4 driving state profiles.

[0051] Fifth step, calculate the cylinder head life degradation factor of the driving state profile:

[0052] 1) Classify the driving segment sets of each vehicle into 4 driving state profiles respectively to obtain the driving state profiles experienced during the vehicle's full life cycle and the cumulative driving mileage under each driving state profile.

[0053] 2) Substitute the cumulative driving mileage of all vehicles under each driving state profile into the cumulative damage model. Assign a value of 0 to the driving mileage under the driving state profiles that have not been experienced, and calculate the cylinder head life degradation factors under 4 driving state profiles as shown in the following table.

[0054]

[0055] Step 6, Cylinder head life prediction:

[0056] 1) Select the historical driving data of 7 groups of vehicles to be predicted. Divide the driving data into historical driving segments, classify the historical driving segment sets of the vehicles to be predicted into 4 driving state profiles, and calculate the cumulative driving mileage under each driving state profile.

[0057] 2) Use the cylinder head life degradation factors under 4 driving state profiles and the cumulative driving mileage under each driving state profile to calculate the cumulative damage degree up to the current historical driving data.

[0058] 3) Assuming that the future driving states include all the historical state profiles experienced and the driving mileage under each future driving state profile is distributed according to the proportion of the historical state profiles, set the future driving states of 7 groups of vehicles to be predicted, calculate the predicted results of the remaining service life of the cylinder head, and compare the predicted results with the fatigue damage life of the actual service components as shown in the following table.

[0059]

[0060] The life prediction results of the cylinder heads of 7 vehicles are shown in the appendix Figure 5 , and it is verified that the cylinder head life prediction method proposed by the present invention is accurate and feasible.

Claims

1. A method for predicting the service life of an engine cylinder head based on the cumulative damage of the driving state profile, characterized in that: It includes the following steps: Step 1: During the vehicle driving process, collect, monitor and record various vehicle operation data through sensors and recorder devices installed on the vehicle, summarize and analyze each vehicle operation data, and screen out in-vehicle network general variables related to the cylinder head life, including torque, throttle opening, exhaust gas flow, ambient pressure, cumulative mileage, oil temperature, gear position, vehicle speed, water temperature, ambient temperature, engine speed and engine cumulative operation time; process and clean the full life cycle data of N vehicles collected, including data correction, outlier processing, and missing value filling; Step 2: Divide the driving segments of the vehicle according to the vehicle speed and engine speed in the in-vehicle network variables; Step 2.1: Define the data segment from the vehicle speed starting from 0 km / h to greater than 0 km / h and then equal to 0 km / h as one motion segment; Step 2.2: Define the data segment with the vehicle state maintaining a vehicle speed of 0 km / h and the engine speed greater than 0 as one idle segment; Step 2.3: One motion segment and an adjacent idle segment form one driving segment. According to Steps 2.1 and 2.2, divide the in-vehicle terminal driving data of N vehicles during the full life cycle into segments, delete the segments with a short trip segment duration less than 10 s, and screen out the qualified driving segments to form a driving segment set; Step 3: Feature value extraction and principal component analysis of driving segment variables; Step 3.1, based on the driving segment sets of N vehicles, select the characteristic parameters representing the driving state characteristics and the driving state distribution characteristics as the evaluation parameters of a single driving short segment, and calculate the characteristic parameters of all driving segments; assume there are m driving segment samples X = [x (1) , x (2) , … x (i) …, x (m) , where x (i) represents the i-th driving segment sample, including the characteristics of n on-vehicle network general variables, that is where T is the transpose symbol, represents the j-th characteristic of the i-th driving segment sample. Step 3.2, perform a decentralization process on the original data samples, that is, subtract the mean of each feature in all samples from each feature That is: Obtain the decentralized driving parameter sample data Step 3.3, calculate the covariance matrix of the driving parameter sample data Perform eigenvalue decomposition on the covariance matrix, calculate the eigenvalues of the covariance matrix C, and arrange them in descending order, denoted as {λ1, λ2, …, λ τ …, λ p}, where λ1 ≥ λ2 ≥ … ≥ λ p , and the corresponding eigenvectors are denoted as {e1, e2, …, e τ …, e p}, where e τ is the eigenvector corresponding to the τ-th eigenvalue λ τ , and p is the number of eigenvalues; Step 3.4: For the v-th principal component, calculate the contribution rate of the principal component Calculate the cumulative value of the contribution rates of the first v principal components where λ v is the v-th eigenvalue. Take the first v eigenvalues with the cumulative contribution rate CR > ε, where ε is the set cumulative contribution rate, ε ∈ (0.85, 0.9]. The larger the value of ε, the more characteristic information is desired to be retained; Step 3.5, select the eigenvectors corresponding to the top v largest eigenvalues to form the projection matrix W, i.e., W = [e1, e2, …, e v , and multiply the original driving parameter sample data by the projection matrix W to obtain the new reduced-dimensional driving parameter matrix Y, i.e., Step 4: Construct a driving state profile; Step 4.1, based on the driving parameter matrix Y calculated in Step 3.5, denoted as Y = [y (1) , y (2) , … y (i) …, y (m) , where each is a v-dimensional vector, and where represents the real number field; Step 4.2, randomly select k clustering centroid points, denoted as {C1, C2, …, C k}, where 1 < k ≤ m, and each is a v-dimensional vector; Step 4.3, calculate the Euclidean distance from each object in the driving parameter matrix to each cluster center where represents the t-th feature parameter of the i-th object, 1 ≤ i ≤ m, 1 ≤ t ≤ v; C jt is the t-th feature parameter of the j-th cluster center, 1 ≤ j ≤ k; Step 4.4: Compare the distance from each driving segment to each cluster center in turn, and assign the driving segment to the cluster of the nearest cluster center to obtain k driving state clusters; Step 4.5: For each driving state cluster, recalculate its centroid and use it as the new k cluster centroid points; Repeat Steps 4.3 - 4.5 until the driving state clustering centroid remains unchanged, and obtain k driving state clusters, which represent k driving state profiles; Step 5: Calculate the cylinder head life degradation factor of the driving state profile; Step 5.1: Classify the driving segment set of each vehicle into driving state profiles to obtain the driving state profiles experienced during the full life cycle of the vehicle and the cumulative driving mileage under each driving state profile; Step 5.2: Build a cumulative damage model. Each profile causes a certain degree of damage to the cylinder head life. The damage caused by each profile experienced during the engine operation accumulates. When the accumulated damage equals 1, the cylinder head will undergo fatigue failure. Build a cumulative damage model where S is the mileage matrix, is the cumulative driving mileage of the j-th vehicle under the ξ-th driving state profile, and the driving mileage under the unexperienced driving state profile is assigned 0; D is the cylinder head life degradation factor matrix, D ξ is the cylinder head life degradation factor per unit mileage of the vehicle under the ξ-th driving state profile; b is the constant coefficient matrix, and k is the total number of driving state profiles; Step 5.3, solve the cumulative damage model in Step 5.2 to obtain the least squares solution D = (S T S) -1 S T b, which is the cylinder head life degradation factor for each driving state profile; Step 6: Construct a cylinder head life prediction model; Step 6.1: Obtain the historical driving data of the vehicle whose cylinder head life is to be predicted. Segment the historical driving segments of the vehicle according to Step 2, and classify the set of historical driving segments of the vehicle to be predicted into driving state profiles. Assume that the vehicle to be predicted has experienced w driving state profiles, and the cumulative driving mileage under each driving state profile is Determine the life degradation factors of the w driving state profiles according to the cylinder head life degradation factors of the driving state profiles obtained in Step 5, denoted as D hp =[D1, D2, …, D w ; Step 6.2, according to the formula calculate the historical cumulative damage degree up to the current time; Step 6.3, set the future driving states to include all the experienced historical state profiles, and allocate the driving mileage under each future driving state profile according to the historical state profile ratio, and construct a cylinder head life prediction model where S c is the remaining mileage that can be driven before the cylinder head fails, and the prediction result of the remaining service life of the cylinder head is obtained by calculation.

Images