A method for estimating equivalent drag coefficient of special vehicles

Abstract

The present invention belongs to the technical field of development of automatic shift control systems for special vehicles, and specifically relates to a method for estimating the equivalent drag coefficient of special vehicles. The method comprises: extracting features from signals; using the recursive least squares method to identify and estimate the vehicle mass and the equivalent drag coefficient; performing confidence preprocessing on the estimated values; further performing confidence judgment on the credible estimated values ​​screened out; and finally performing weight fusion calculations. The method further reduces the deviation of the estimation results caused by abnormal data and calculation parameter errors, and has high real-time performance. The method for estimating the equivalent drag coefficient of special vehicles has been successfully applied to the calculation and simulation analysis of the control system of a certain wheeled armored vehicle. Verification shows that the method is reasonably designed, has clear logic, and reduces estimation deviations. It can meet the current application and subsequent development needs, and has significant progressive significance for the development of the project.

Description

Technical Field

[0001] The present invention belongs to the technical field of development of automatic shift control systems for special vehicles, and in particular relates to a method for estimating an equivalent drag coefficient of a special vehicle. Background Art

[0002] Traditional automatic transmission control technologies include shift schedule design and shift process control. In shift schedule design, traditional shift schedules use throttle and vehicle speed as control parameters. Due to the unique demands of military use, special vehicles must adapt to a variety of complex and changing environments and harsh conditions. Therefore, making the most reasonable gear decision based on road resistance to improve vehicle control quality under complex operating conditions is particularly important. To improve the adaptability of traditional control strategies to road conditions, a series of intelligent algorithms are being used to study the equivalent drag coefficient estimation method for special vehicles. This is of great significance for optimizing gear decisions and improving vehicle dynamics and road adaptability. Summary of the Invention

[0003] (1) Technical issues to be solved

[0004] The technical problem to be solved by the present invention is: how to provide a method for estimating the equivalent drag coefficient of a special vehicle.

[0005] (2) Technical solution

[0006] To solve the above technical problems, the present invention provides a method for estimating the equivalent drag coefficient of a special vehicle, the method comprising:

[0007] Step 1: Extract features from the signal;

[0008] Step 2: After extracting the signal features, the recursive least squares method with multiple forgetting factors is first used to identify and estimate the vehicle mass and equivalent drag coefficient. A deadline T is set. When the deadline is reached, the current estimation is stopped and the next estimation is started. A total of K estimates are performed to obtain estimated values ​​1, 2, ..., K.

[0009] Step 3: After the estimation is completed, the confidence level of the estimated value is preprocessed. The preprocessing method is to set a limit threshold to ensure that the error between two adjacent estimated values ​​is less than the threshold;

[0010] In the case of multiple estimated values, statistical probability methods are used to handle them, correlation analysis is performed on the estimated values, data consistency is tested based on the degree of confidence, and abnormal data are eliminated;

[0011] Step 4: After confidence preprocessing, the confidence level of the filtered out reliable estimates is further judged. A confidence level judger is established, which takes the vehicle speed signal, torque signal, brake signal, and gear position signal as inputs and outputs the confidence level of each estimate after setting rules.

[0012] Step 5: Finally, perform weight fusion calculation and define the confidence threshold. When the confidence of a certain estimate is very high and the confidence of other estimates is low, the estimate with high confidence is output as the final estimate.

[0013] When the confidence of a certain estimate is very low, and the confidence of other estimates is neither high nor low, the weight of this estimate is set to 0 to reduce its influence on the fusion result;

[0014] When the confidence levels of all estimated values ​​are similar, the confidence levels of the estimated values ​​are normalized and used as their weights for weighted fusion to obtain the final estimated value.

[0015] Wherein, in said step 1, said signals for feature extraction include: vehicle speed signal, torque signal, brake signal and gear signal.

[0016] Among them, in step 2, due to the particularity of special vehicles, the mass of the vehicle remains basically unchanged, so the forgetting factor corresponding to the mass of the vehicle can take a larger value, close to 1, while the equivalent resistance coefficient is a time-varying value, and the corresponding forgetting factor value should be smaller.

[0017] In step 3, the estimated value confidence level is preprocessed as follows:

[0018] Step 31: First, perform vehicle mass confidence preprocessing. For special vehicles, set M1 as the empty mass boundary value and M2 as the fully loaded mass boundary value, and remove the vehicle mass estimate that exceeds the boundary value; define the equivalent drag coefficient estimate after processing as x i , i=1,2...n, n is less than or equal to K;

[0019] An estimated value x i and another estimate x j The deviation is expressed as Δ ij express:

[0020] Δ ij =|x i -x j |

[0021] Δ ij The larger the value, the estimated value x i and the estimated value x j The larger the deviation between them, the error matrix is ​​constructed. Indicates the degree of deviation between the two K estimates;

[0022]

[0023] Step 32: Set the confidence threshold E. If Δ ij >E, indicating estimated value x i and the estimated value x j The deviation is large, the data do not support each other, and the confidence is 0. On the contrary, it means that the two support each other, the correlation between the data is large, and the confidence is 1. We can get a confidence matrix R between the estimated values n ;

[0024]

[0025] Step 33: If the confidence array r of the estimated value i ij When most of them are 1, it means that the estimated value is highly correlated with most of the estimated values, with high confidence and high validity. Otherwise, the correlation is low, the confidence is low, and the validity is low, and it can be eliminated as abnormal data.

[0026] Wherein, i=1,2...n, j=1,2...n, n is less than or equal to K.

[0027] Wherein, in said step 4, after the confidence preprocessing in step 3, K' estimated values ​​are obtained and input into the confidence calculator. In combination with the signal characteristics, a confidence judgment rule is given in the confidence calculator to perform confidence judgment.

[0028] Wherein, K' is less than or equal to K.

[0029] In step 4, the vehicle speed signal, torque signal, throttle signal, and throttle change rate signal are used as discrimination conditions, and a corresponding [0, 1] confidence value is output for each estimated value; the vehicle speed signal in the discrimination conditions is divided into three standards: low, medium, and high; the torque signal is divided into three standards: low, medium, and high; the throttle signal is divided into three standards: low, medium, and high; and the throttle change rate is divided into two standards: large and small. 54 discrimination conditions are combined, and a confidence value is given for each combination of discrimination conditions; the estimated value with a small throttle and a large throttle change rate has the lowest confidence, and the estimated value with a large throttle and a small throttle change rate has the highest confidence;

[0030] Each estimated value is calculated and judged according to the vehicle speed signal, torque signal, throttle signal and throttle change rate state at the time of estimation, and the corresponding confidence R is obtained. i ,i=K'.

[0031] Among them, in step 5, the weight fusion calculator obtains weights according to the confidence of each estimated value, and then performs weighted fusion on them, and the fusion result is the final estimated value.

[0032] Among them, in the step 5, when the confidence of a certain estimated value is very high and the confidence of other estimated values ​​is low, the estimated value with high confidence is output as the final estimated value; when the confidence of a certain estimated value is very low and the confidence of other estimated values ​​is neither high nor low, the weight of this estimated value is set to 0 to reduce its impact on the fusion result; when the confidence of all estimated values ​​is similar, the confidence of the estimated value is normalized and used as its weight W i , the normalization formula is:

[0033]

[0034] Finally, weighted fusion is performed to obtain the final estimated value.

[0035] (3) Beneficial effects

[0036] Compared with the prior art, the present invention addresses the problem of large estimation errors in the vehicle equivalent drag coefficient using the recursive least squares method with multiple forgetting factors alone. By proposing a method for weighted fusion of estimated values, the method further reduces the deviations in the estimation results caused by abnormal data and calculation parameter errors, and offers high real-time performance. The method for estimating the equivalent drag coefficient of special vehicles has been successfully applied to the calculation and simulation analysis of the control system of a wheeled armored vehicle. Verification has shown that the method is rationally designed, has clear logic, and reduces estimation errors. It can meet the needs of current applications and subsequent development, and has significant progressive significance for the development of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the principle of the technical solution of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, content, and advantages of the present invention more clear, the specific implementation methods of the present invention are further described in detail below with reference to the accompanying drawings and examples.

[0039] At present, the least squares method with forgetting factor has been widely used in the estimation of road resistance coefficient. However, due to the harsh operating environment of military special vehicles, the error in extracting the longitudinal acceleration signal features increases, which will lead to deviation in the estimated value. The data saturation phenomenon of the least squares method will also make the estimation error larger and larger. Under some special working conditions, such as low-speed high torque, high-speed gliding, high-speed small throttle, braking, etc., and the different power losses in different gears will cause the calculation parameters to change and lead to biased estimates.

[0040] To address these issues, the present invention proposes a method for estimating the equivalent drag coefficient of special vehicles. This method first sets a cutoff frequency for the least squares method and performs multiple estimations. These estimates are then subjected to confidence preprocessing, followed by confidence assessment. Finally, a weighted fusion is performed to obtain the final estimate. This method can, to a certain extent, suppress bias in the estimated values ​​and improve their accuracy.

[0041] In order to solve the above technical problems, the present invention provides a method for estimating the equivalent drag coefficient of a special vehicle, such as Figure 1 As shown, the method includes:

[0042] Step 1: Extract features from the signal;

[0043] Step 2: After extracting the signal features, similar to traditional methods, we first use the recursive least squares method with multiple forgetting factors to identify and estimate the vehicle mass and equivalent drag coefficient. Due to the special characteristics of special vehicles, the vehicle mass remains essentially unchanged, so the forgetting factor corresponding to the vehicle mass can take a large value, close to 1. The equivalent drag coefficient is a time-varying value, so the corresponding forgetting factor should be small. To address data saturation, a deadline T is set. When the deadline is reached, the current estimation is terminated and the next estimation is started. A total of K estimates are performed, resulting in estimated values ​​1, 2, ..., K.

[0044] Step 3: After the estimation is completed, the estimated value confidence preprocessing is performed. The general method of preprocessing is to set a limit threshold to ensure that the error between two adjacent estimated values ​​is less than the threshold;

[0045] In the case of multiple estimated values, statistical probability methods are used to handle them, correlation analysis is performed on the estimated values, data consistency is tested based on the degree of confidence, and abnormal data are eliminated;

[0046] Step 4: After confidence preprocessing, the confidence level of the filtered out reliable estimates is further judged. A confidence level judger is established, which takes the vehicle speed signal, torque signal, brake signal, and gear position signal as inputs and outputs the confidence level of each estimate after setting rules.

[0047] Step 5: Finally, perform weight fusion calculation and define the confidence threshold. When the confidence of a certain estimate is very high and the confidence of other estimates is low, the estimate with high confidence is output as the final estimate.

[0048] When the confidence of a certain estimate is very low, and the confidence of other estimates is neither high nor low, the weight of this estimate is set to 0 to reduce its influence on the fusion result;

[0049] When the confidence levels of all estimated values ​​are similar, the confidence levels of the estimated values ​​are normalized and used as their weights for weighted fusion to obtain the final estimated value.

[0050] Wherein, in said step 1, said signals for feature extraction include: vehicle speed signal, torque signal, brake signal and gear signal.

[0051] Among them, in step 2, due to the particularity of special vehicles, the mass of the vehicle remains basically unchanged, so the forgetting factor corresponding to the mass of the vehicle can take a larger value, close to 1, while the equivalent resistance coefficient is a time-varying value, and the corresponding forgetting factor value should be smaller.

[0052] In step 3, the estimated value confidence level is preprocessed as follows:

[0053] Step 31: First, perform vehicle mass confidence preprocessing. For special vehicles, set M1 as the empty mass boundary value and M2 as the fully loaded mass boundary value, and remove the vehicle mass estimate that exceeds the boundary value; define the equivalent drag coefficient estimate after processing as x i , i=1,2...n, n is less than or equal to K;

[0054] An estimated value x i and another estimate x j The deviation is expressed as Δ ij express:

[0055] Δ ij =|x i -x j |

[0056] Δ ij The larger the value, the estimated value x i and the estimated value x j The larger the deviation between them, the error matrix is ​​constructed. Indicates the degree of deviation between the two K estimates;

[0057]

[0058] Step 32: Set the confidence threshold E. If Δ ij >E, indicating estimated value x i and the estimated value x j The deviation is large, the data do not support each other, and the confidence is 0. On the contrary, it means that the two support each other, the correlation between the data is large, and the confidence is 1. We can get a confidence matrix R between the estimated values n ;

[0059]

[0060] Step 33: If the confidence array r of the estimated value i ij When most of them are 1, it means that the estimated value is highly correlated with most of the estimated values, with high confidence and high validity. Otherwise, the correlation is low, the confidence is low, and the validity is low, and it can be eliminated as abnormal data.

[0061] Wherein, i=1,2...n, j=1,2...n, n is less than or equal to K.

[0062] Wherein, in said step 4, after the confidence preprocessing in step 3, K' estimated values ​​are obtained and input into the confidence calculator. In combination with the signal characteristics, a confidence judgment rule is given in the confidence calculator to perform confidence judgment.

[0063] Wherein, K' is less than or equal to K.

[0064] In step 4, the vehicle speed signal, torque signal, throttle signal, and throttle change rate signal are used as discrimination conditions, and a corresponding [0, 1] confidence value is output for each estimated value; the vehicle speed signal in the discrimination conditions is divided into three standards: low, medium, and high; the torque signal is divided into three standards: low, medium, and high; the throttle signal is divided into three standards: low, medium, and high; and the throttle change rate is divided into two standards: large and small. 54 discrimination conditions are combined, and a confidence value is given for each combination of discrimination conditions; the estimated value with a small throttle and a large throttle change rate has the lowest confidence, and the estimated value with a large throttle and a small throttle change rate has the highest confidence;

[0065] Each estimated value is calculated and judged according to the vehicle speed signal, torque signal, throttle signal and throttle change rate state at the time of estimation, and the corresponding confidence R is obtained. i ,i=K'.

[0066] Among them, in step 5, the weight fusion calculator obtains weights according to the confidence of each estimated value, and then performs weighted fusion on them, and the fusion result is the final estimated value.

[0067] Among them, in the step 5, when the confidence of a certain estimated value is very high and the confidence of other estimated values ​​is low, the estimated value with high confidence is output as the final estimated value; when the confidence of a certain estimated value is very low and the confidence of other estimated values ​​is neither high nor low, the weight of this estimated value is set to 0 to reduce its impact on the fusion result; when the confidence of all estimated values ​​is similar, the confidence of the estimated value is normalized and used as its weight W i , the normalization formula is:

[0068]

[0069] The global estimate after fusion

[0070]

[0071] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for estimating the equivalent drag coefficient of a special vehicle, characterized in that: The method comprises: Step 1: Extract features from the signal; Step 2: After extracting the signal features, first use the recursive least squares method with multiple forgetting factors to identify and estimate the vehicle mass and equivalent drag coefficient; set the deadline , stop the current estimation after reaching the deadline and start the next estimation. Estimation, get estimated values ​​1, 2...K; Step 3: After the estimation is completed, the confidence level of the estimated value is preprocessed. The preprocessing method is to set a limit threshold to ensure that the error between two adjacent estimated values ​​is less than the threshold; In the case of multiple estimated values, statistical probability methods are used to handle them, correlation analysis is performed on the estimated values, data consistency is tested based on the degree of confidence, and abnormal data are eliminated; Step 4: After confidence preprocessing, the confidence level of the filtered out reliable estimates is further judged. A confidence level judger is established, which takes the vehicle speed signal, torque signal, brake signal, and gear position signal as inputs and outputs the confidence level of each estimate after setting rules. Step 5: Finally, perform weight fusion calculation and define the confidence threshold. When the confidence of a certain estimate is very high and the confidence of other estimates is low, the estimate with high confidence is output as the final estimate. When the confidence of a certain estimate is very low, and the confidence of other estimates is neither high nor low, the weight of this estimate is set to 0 to reduce its influence on the fusion result; When the confidence levels of all estimated values ​​are similar, the confidence levels of the estimated values ​​are normalized and used as their weights for weighted fusion to obtain the final estimated value. In step 3, the estimated value confidence level is preprocessed as follows: Step 31: First, pre-process the vehicle quality trustworthiness. For special vehicles, set is the no-load mass boundary value, is the full-load mass boundary value, and the value of the vehicle mass estimate that exceeds the boundary is eliminated; the equivalent drag coefficient estimate after processing is defined as , i=1,2...n, n is less than or equal to K; An estimated value and another estimate The deviation of express: The larger the value, the more estimated it is. and estimated values The larger the deviation between them, the error matrix is ​​constructed. ,express The degree of bias between the two estimates; Step 32: Set the trust threshold ,if , which represents the estimated value and estimated values The deviation is large, the data do not support each other, and the confidence is 0. On the contrary, it means that the two support each other, the correlation between the data is large, and the confidence is 1. A confidence matrix between estimates can be obtained ; Step 33: If the estimated value The confidence array r ij When most of them are 1, it means that the estimated value is highly correlated with most of the estimated values, with high confidence and high validity. Otherwise, the correlation is low, the confidence is low, and the validity is low, and it can be eliminated as abnormal data.

2. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 1, wherein: In step 1, the signals for feature extraction include: vehicle speed signal, torque signal, brake signal and gear signal.

3. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 1, wherein: In step 2, due to the particularity of special vehicles, the mass of the entire vehicle remains basically unchanged, so the forgetting factor corresponding to the mass of the entire vehicle can take a larger value, close to 1, while the equivalent resistance coefficient is a time-varying value, and the corresponding forgetting factor value should be smaller.

4. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 1, wherein: i=1,2...n, j=1,2...n, n is less than or equal to K.

5. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 1, wherein: In step 4, after the confidence preprocessing in step 3, K' estimated values ​​are obtained and input into the confidence calculator. In combination with the signal characteristics, a confidence judgment rule is given in the confidence calculator to perform confidence judgment.

6. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 5, wherein: K' is less than or equal to K.

7. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 5, wherein: In step 4, the vehicle speed signal, torque signal, throttle signal and throttle change rate signal are used as the judgment conditions, and the corresponding output of each estimated value is Confidence value: The vehicle speed signal is divided into three standards: low, medium and high; the torque signal is divided into three standards: low, medium and high; the throttle signal is divided into three standards: low, medium and high; and the throttle change rate is divided into two standards: large and small. 54 different discrimination conditions are combined, and a confidence value is given for each combination of discrimination conditions. Among them, the estimated value with a small throttle and a large throttle change rate has the lowest confidence, while the estimated value with a large throttle and a small throttle change rate has the highest confidence. Each estimated value is calculated and judged according to the vehicle speed signal, torque signal, throttle signal and throttle change rate state at the time of estimation, and the corresponding confidence R is obtained. i ,i=K'.

8. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 7, wherein: In step 5, the weight fusion calculator obtains weights according to the confidence of each estimated value, and then performs weighted fusion on them, and the fusion result is the final estimated value.

9. The method for estimating the equivalent drag coefficient of a special vehicle according to claim 8, wherein: In step 5, when the confidence of a certain estimate is very high and the confidence of other estimates is low, the estimate with high confidence is output as the final estimate; when the confidence of a certain estimate is very low and the confidence of other estimates is neither high nor low, the weight of this estimate is set to 0 to reduce its impact on the fusion result; when the confidence of all estimates is similar, the confidence of the estimate is normalized and used as its weight. , the normalization formula is: Finally, weighted fusion is performed to obtain the final estimated value.

Images