A gearbox gear self-learning system and method

Abstract

The application discloses a gearbox gear self-learning system and method, and relates to the field of gearbox control.The system comprises: a gear selection displacement deviation learning module, which is used for pushing the gear selection displacement to the limit position by applying a negative gear selection duty ratio, calculating the gear selection displacement deviation, and correcting the gear selection value based on the deviation; a neutral displacement learning module, which is configured to perform temporary neutral position learning and multiple neutral position sampling, and to generate a neutral self-learning displacement by averaging; a gear shifting displacement learning module, which is used for switching the target gear cyclically, recording the gear-in displacement, and generating a gear shifting self-learning displacement by averaging multiple sampling values; and a gear selection displacement self-learning module, which is used for controlling the gear selection displacement to both ends of the gear shifting position during gear shifting, taking the intermediate value, and generating a gear selection self-learning displacement after multiple sampling.The application can significantly enhance the robustness of the gear shifting logic, realize full automation of the self-learning process, and provide a high-reliability solution for intelligent control of the gearbox.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of gearbox control, in particular to a gearbox gear self-learning system and method. BACKGROUND

[0002] Gearbox gear self-learning technology is a core means to improve the shift accuracy and reliability of automatic transmissions, which learns the displacement parameters in the actual operation of the gearbox, and eliminates the shift deviation caused by sensor installation error, mechanical wear or assembly tolerance.

[0003] The prior art mainly relies on the following methods: static calibration method, which calibrates the gear position by fixed duty cycle signal or preset displacement range when the gearbox is offline; single learning method, which learns the subsequent gear based on the default N gear position, and only records the last gear displacement value as the reference. The said methods have the following key problems: the installation deviation of gear selection and the displacement deviation of neutral gear are not considered, which leads to gear selection failure or gear entry error in the self-learning process; the fixed parameters or single sampling data are relied on, which cannot cover the sensor accuracy error and mechanical size change, leading to an increase in shift failure rate; the method does not combine with actual working conditions for dynamic verification, leading to the disconnection between self-learning results and actual needs; frequent manual intervention or offline calibration is required, increasing data maintenance and vehicle downtime. Therefore, based on the above problems, the present application proposes a gearbox gear self-learning system and method. SUMMARY

[0004] TECHNICAL OBJECTIVE

[0005] In order to solve the above problems, the purpose of the present application is to provide a gearbox gear self-learning system and method, which aims to solve the shift failure problem caused by the uncorrected installation deviation of gear selection, the default neutral position reference error and the contingency of single sampling data in the existing gearbox gear self-learning technology, to improve the accuracy and robustness of the self-learning process through systematic error compensation mechanism and multi-stage learning optimization strategy, to realize the dynamic adaptation and long-term stability of the shift logic, and to reduce the requirement of manual intervention and maintenance cost.

[0006] TECHNICAL SCHEME

[0007] In order to achieve the above-mentioned purpose, the application provides a gearbox gear self-learning system and method, which drives displacement to a limit position by a negative gear selection duty ratio to calculate gear selection displacement offset and dynamically correct software setting value; determines a reference position by taking an average value based on neutral displacement sampling at three different gear selection intermediate positions; records three times of gear-in displacement average values to generate a gear shift parameter in high-low gear cyclic switching; optimizes displacement data by combining timeout monitoring and three times of sampling abnormal value elimination mechanism, and finally feeds the corrected gear selection displacement, neutral reference and gear shift parameter to an execution system in real time, eliminates mechanical assembly error, sensor deviation and accidental interference, and ensures reliability of self-learning result and accuracy of gear shift action.

[0008] In a first aspect, the application provides a gearbox gear self-learning system, comprising:

[0009] A gear selection displacement deviation learning module is configured to push gear selection displacement to a limit position by applying a negative gear selection duty ratio, calculate gear selection displacement offset, and correct gear selection value of software setting based on the offset;

[0010] A neutral displacement learning module is configured to perform temporary neutral position learning and multiple neutral position sampling, and take an average value to generate neutral self-learning displacement;

[0011] A gear shift displacement learning module is configured to generate gear shift self-learning displacement by cyclically switching target gears and recording gear-in displacement, and taking an average value of multiple sampling;

[0012] A gear selection displacement self-learning module is configured to control gear selection displacement to both ends of gear shift position and take an intermediate value during gear shift, and generate gear selection self-learning displacement after multiple sampling;

[0013] An error elimination module takes an average value of more than three self-learning data to eliminate accidental error;

[0014] An execution control unit is configured to adjust gearbox gear shift logic according to self-learning result.

[0015] Further, the gear selection displacement deviation learning module determines the leftmost limit position of gear selection displacement through displacement limit detection, takes the difference between the limit position and the default value of preset gear selection displacement as gear selection displacement offset, and superimposes gear selection value and the offset in real time to generate temporary gear selection displacement.

[0016] Further, the neutral displacement learning module generates a temporary neutral position by hanging gears one and two and recording an average value of gear shift displacement, and performs neutral displacement sampling at three different gear selection intermediate positions to obtain neutral self-learning displacement by taking an average value of three times.

[0017] Further, the shift displacement learning module switches step by step from the lowest gear to the highest gear, and then switches step by step from the highest gear to the lowest gear, forming a complete cycle, and recording the last time the gear is entered in each cycle, and taking the average of the gear entry displacement in three cycles as the shift self-learning displacement.

[0018] Further, the selected gear displacement self-learning module pushes the selected gear displacement to the two end vertices of the current shift position, and takes the middle value of the displacement data of the two end vertices as the single selected gear displacement self-learning value, and stores and applies the average of three self-learning values.

[0019] Further, the error elimination module records three successful self-learning values during the shift displacement and selected gear displacement self-learning processes, and optimizes the final self-learning displacement by removing abnormal values and taking the average.

[0020] Further, the system has a timing threshold in each self-learning stage, and if the timing threshold is exceeded, the self-learning is determined to fail, and the self-learning process and abnormal alarm signal are output in real time.

[0021] Further, it also includes a selected gear offset correction module, which constructs a displacement response curve, combines multiple dynamic sampling data, fits the deviation between actual displacement and theoretical value, and the corrected offset calculation formula is:

[0022]

[0023] In the formula, is the corrected offset; is the selected gear displacement limit position; is the default value of the selected gear displacement; is a non-linear compensation factor; is the instantaneous slope of the displacement-duty cycle curve at the limit position; is the preset maximum slope threshold.

[0024] By constructing a displacement response curve and introducing a non-linear compensation factor, the selected gear displacement offset is dynamically corrected, effectively overcoming the influence of the non-linear characteristics of the mechanical system on the displacement accuracy. This mechanism can significantly reduce the selected gear position error and improve the robustness of the shift logic, ensuring the dynamic adaptation ability of the self-learning process to sensor installation deviation and mechanical assembly error.

[0025] Further, it also includes a weighted mean optimization module, which samples displacement at different selected gear intermediate positions, and gives dynamic weight based on displacement distribution variance, and the neutral gear reference position formula is:

[0026]

[0027] In the formula, is the neutral gear reference position; is the normalized weight coefficient; is the i-th sampled neutral displacement value.

[0028] By sampling the displacement at different gear selection intermediate positions and assigning dynamic weights based on the inverse variance, the interference of local errors on the neutral reference position is suppressed. This algorithm can significantly reduce systematic errors, improve the accuracy and stability of the neutral positioning, and especially in the case of asymmetric assembly or dynamic working conditions, it strengthens the anti-interference ability of the reference position.

[0029] In a second aspect, the present application further provides a gearbox gear self-learning method, which is based on the system of the first aspect and comprises:

[0030] The corrected gear selection displacement is generated by limit position detection and offset calculation;

[0031] The neutral reference position is determined based on the mean value of multiple sampling;

[0032] The gear shift self-learning displacement is generated by gear cycle switching and displacement recording;

[0033] The gear selection self-learning displacement is generated by combining gear shift position boundary detection and intermediate value calculation;

[0034] The mean value of more than three self-learning data is taken to optimize the final displacement parameters.

[0035] Further, in the neutral displacement learning process, neutral displacement sampling is performed at three different gear selection intermediate positions, and the average value of the three sampling results is taken as the reference value of the neutral self-learning displacement.

[0036] In a third aspect, the present application further provides a computer device, comprising a management platform and a memory, wherein the management platform is connected to the memory, the memory is used to store a computer program, and the management platform is used to execute the computer program stored in the memory, so that the computer device executes the gearbox gear self-learning method.

[0037] In a fourth aspect, the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a management platform to implement the gearbox gear self-learning method.

[0038] The present application solves the problem of insufficient gear shifting precision caused by sensor installation deviation, mechanical assembly error and accidental single sampling in the prior art by a phased self-learning process and a multi-source data fusion strategy. Specifically, the present application includes: dynamic selection offset correction based on displacement curve, which eliminates the influence of friction hysteresis and gear gap through a nonlinear compensation factor; multi-point weighted average optimization algorithm for neutral displacement, which uses variance inverse proportional weight to suppress local error; phased self-learning sequence optimization, which decouples selection offset learning, neutral positioning and gear shift displacement learning into independent steps to avoid error propagation; and three or more self-learning data mean combined with an outlier rejection mechanism to eliminate accidental interference. The system can significantly enhance the robustness of the gear shifting logic, improve the neutral reference precision and gear shift displacement parameter stability, while realizing full automation of the self-learning process, reducing the need for human intervention and maintenance costs, and providing a high-reliability solution for intelligent control of the gearbox.

[0039] Advantages

[0040] By implementing the gearbox gear self-learning system and method provided by the present application, the following technical effects are achieved:

[0041] (1) The present application performs selection offset learning, neutral position learning and gear shift displacement learning in phases, which decomposes the complex self-learning process into ordered steps to avoid error accumulation affecting subsequent stages. This improves the fault tolerance of the self-learning process, reduces the risk of overall process interruption due to single-stage failure, and enhances the logical rigor of multi-parameter collaborative optimization.

[0042] (2) By taking the mean of multiple self-learning samples, combined with timeout monitoring and outlier rejection mechanism, accidental errors in single sampling are effectively eliminated. This ensures the long-term stability of gear shift displacement and selection displacement parameters, reduces parameter fluctuations caused by mechanical vibration, sensor noise or transient sticking, and ensures the repeatability and reliability of the self-learning results.

[0043] (3) The displacement response curve is constructed and a nonlinear compensation factor is introduced to dynamically correct the selection displacement offset, effectively overcoming the influence of the nonlinear characteristics of the mechanical system on displacement accuracy. This mechanism can significantly reduce the selection position error, improve the robustness of the gear shifting logic, and ensure the dynamic adaptation ability of the self-learning process to sensor installation deviation and mechanical assembly error.

[0044] (4) Displacement sampling is performed at different selection intermediate positions, and dynamic weights are assigned based on variance inverse proportion to suppress local error interference on the neutral reference position. This algorithm can significantly reduce systematic errors, improve neutral positioning accuracy and stability, and especially in asymmetric assembly or dynamic working conditions, it can enhance the anti-interference ability of the reference position. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to make the gearbox gear self-learning system and method of the present application more obvious and easy to understand, the drawings required in the specific embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can also be obtained from these drawings without creative labor for those skilled in the art.

[0046] Figure 1 The figure shows the system principle of the present application.

[0047] Figure 2 The figure shows the flow chart of the gearbox gear self-learning method. DETAILED DESCRIPTION

[0048] Embodiment 1:

[0049] A gearbox gear self-learning system is provided, and the system principle is shown in Figure 1 The figure shows the system principle of the present application. The system includes: a gear shift displacement deviation learning module, configured to push the gear shift displacement to the limit position by applying a negative gear shift duty ratio, calculate the gear shift displacement deviation, and correct the software set gear shift value based on the deviation; a neutral displacement learning module, configured to perform temporary neutral position learning and multiple neutral position sampling, and generate the neutral self-learning displacement by averaging; a gear shift displacement learning module, configured to generate the gear shift self-learning displacement by cyclically switching the target gear and recording the gear shift displacement, and averaging the multiple sampling values; a gear shift displacement self-learning module, configured to control the gear shift displacement to the two ends of the gear shift position and take the intermediate value during the gear shift process, and generate the gear shift self-learning displacement after multiple sampling; an error elimination module, configured to eliminate accidental errors by averaging three or more self-learning data; and an execution control unit, configured to adjust the gearbox gear shift logic according to the self-learning result. Details are as follows.

[0050] The gear shift displacement deviation learning module determines the leftmost limit position of the gear shift displacement through displacement limit detection, takes the difference between the limit position and the default value of the preset gear shift displacement as the gear shift displacement deviation, and superimposes the gear shift value and the deviation in real time to generate a temporary gear shift displacement.

[0051] The neutral displacement learning module generates a temporary neutral position by hanging gears 1 and 2 and recording the average gear shift displacement, and performs neutral displacement sampling at three different gear shift intermediate positions, and obtains the neutral self-learning displacement by averaging three times.

[0052] The gear shift displacement learning module switches from the lowest gear to the highest gear in a step-by-step increasing manner, and then switches in a step-by-step decreasing manner, forming a complete cycle. In each cycle, the last gear shift displacement of each gear is recorded, and the gear shift self-learning displacement is obtained by averaging the gear shift displacements in three cycles.

[0053] The selected gear shift displacement self-learning module pushes the selected gear shift displacement to the two end vertices of the current shift position, and takes the middle value of the displacement data of the two end vertices as the single selected gear shift displacement self-learning value, and stores and applies the average value of the three self-learning values.

[0054] The error elimination module records three successful self-learning values during the shift displacement and selected gear shift displacement self-learning processes, and optimizes the final self-learning displacement by taking the average value after removing the abnormal value.

[0055] The system is provided with a timing threshold in each self-learning stage, and if the timing threshold is exceeded, the self-learning is determined to fail, and the self-learning process and abnormal alarm signal are output in real time.

[0056] A gearbox gear self-learning method is also provided, which is based on the foregoing system, and the flow is as shown in Figure 2 The method includes generating a corrected selected gear shift displacement through limit position detection and offset calculation, determining a neutral position reference based on the average value of multiple samplings, generating a shift self-learning displacement through gear cycle switching and displacement recording, generating a selected gear self-learning displacement by combining shift position boundary detection and intermediate value calculation, and taking the average value of more than three self-learning data to optimize the final displacement parameter.

[0057] During the neutral displacement learning process, neutral displacement sampling is performed at three different selected intermediate positions, and the average value of the three sampling results is taken as the reference value of the neutral self-learning displacement.

[0058] Embodiment 2:

[0059] On the basis of the foregoing embodiments, considering that the traditional calculation method of the selected gear shift displacement offset only relies on the linear difference value of the limit position and the preset value, and does not consider the influence of the nonlinear characteristics of the mechanical system on the displacement, a dynamic selected gear shift offset correction mechanism based on the displacement curve is added. This mechanism constructs a displacement response curve, combines multiple dynamic sampling data, introduces a nonlinear compensation factor, accurately fits the deviation between the actual displacement and the theoretical value, and realizes dynamic correction.

[0060] The selected gear shift displacement response data is recorded while different duty cycle signals are applied to generate a displacement-duty cycle curve.

[0061] According to the limit position displacement and the preset value, the corrected offset is calculated by combining the curve slope and the nonlinear compensation factor:

[0062]

[0063] In the formula, is the corrected offset; is the limit position of the selected gear shift displacement; is the default value of the selected gear shift displacement; is the nonlinear compensation factor; is the instantaneous slope of the displacement-duty ratio curve at the limit position; is a preset maximum slope threshold.

[0064] The calculation formula of the nonlinear compensation factor is:

[0065]

[0066] In the formula, is the current sampling time; is the system dynamic response time constant.

[0067] is superimposed on the software set value to generate a temporary gear selection displacement instruction.

[0068] Verification shows that in the case of obtaining similar average error as the above-mentioned embodiment, through nonlinear compensation, the gear selection position error is reduced from ±1.2 mm to ±0.4 mm, the gear shifting failure rate is reduced from 8% to 2%, and the displacement standard deviation is reduced by 40% under temperature change. The results show that the dynamic correction mechanism effectively neutralizes the inherent nonlinear disturbances such as friction hysteresis and gear clearance of the mechanical system through the nonlinear compensation factor, so that the actual response of the gear selection position is more in line with the theoretical expectation. Compared with the traditional linear difference method, the corrected gear selection displacement error range is significantly reduced, the gear shifting logic has improved adaptability to sensor installation deviation, and the gear shifting failure rate shows a systematic downward trend.

[0069] Example 3:

[0070] On the basis of the foregoing embodiment, in order to avoid the problem that the traditional neutral position learning relies on single intermediate position sampling and is easily affected by mechanical assembly asymmetry, a multi-point weighted mean optimization algorithm for neutral displacement is added. The algorithm performs displacement sampling at three different gear selection intermediate positions, and gives dynamic weights based on displacement distribution variance to suppress the interference of local error on the reference position.

[0071] Neutral displacement sampling is performed at left, center and right of the gear selection intermediate position respectively, displacement values are obtained, and weight coefficients are calculated according to the variance of three times sampling:

[0072]

[0073] In the formula, is the weight coefficient; is the variance of the i-th neutral displacement sampling; is a smoothing factor to prevent the denominator from being 0.

[0074] The normalized weight formula is:

[0075] ​

[0076] wherein, is the normalized weight coefficient, ensuring the sum of all sampling weights is 1, ensuring the rationality of weighted average; is the dynamic weight coefficient of the j-th neutral displacement sampling.

[0077] The weighted average generates the neutral self-learning displacement, and the neutral reference position formula is:

[0078]

[0079] wherein, is the final neutral reference position, eliminating local errors caused by mechanical assembly asymmetry through weighted average, improving position accuracy; is the i-th sampled neutral displacement value.

[0080] Suppose that three neutral displacement samplings are performed at the three selected gear positions, the left-biased position is 100.3 mm, the centered position is 100.0 mm, and the right-biased position is 100.5 mm, and the preset weight distribution is that the left-biased position weight is 0.25, the centered position weight is 0.50, and the right-biased position weight is 0.25.

[0081] The weighted reference position is calculated as:

[0082]

[0083] The traditional method reference is:

[0084]

[0085] The effect of the multi-point weighted mean optimization algorithm of the neutral displacement is shown in Table 1.

[0086] Table 1, Effect summary of multi-point weighted mean optimization algorithm of neutral displacement

[0087] Parameter Left bias position Center position Right bias position Single sampling value (mm) 100.3 100.0 100.5 Pre-set weight 0.25 0.50 0.25 Method Reference position (mm) Error (ideal value 100.0 mm) Error reduction ratio Conventional simple average method 100.27 +0.27 mm - Weighted average method 100.20 +0.20 mm 25.9%

[0088] According to the experiment table, the weighted mean method reduces the reference position error from +0.27 mm of the traditional method to +0.20 mm, with an error reduction of 25.9%. It only needs single sampling and preset weight, without variance calculation, and is suitable for fast self-learning scenarios. By giving higher weight to the central position, it directly suppresses the interference of left and right offset positions. Experimental data show that, by using the variance inverse weight distribution strategy, the algorithm effectively suppresses the influence of local assembly error and environmental interference on the neutral position. Compared with the traditional single-point sampling or simple average method, the neutral position positioning error fluctuation range after weighted mean optimization is significantly converged, and the repeatability and consistency of the reference position are improved. Especially in the asymmetric assembly or dynamic vibration scenario, the algorithm gives higher weight to the low-variance sampling point, and systematically reduces the risk of reference drift caused by mechanical asymmetry or instantaneous impact. Experiments further verify that the algorithm significantly improves the long-term stability of the neutral self-learning result, providing a reliable reference for the subsequent gear shifting process.

[0089] Those skilled in the art will appreciate that embodiments of the application can be provided as methods, systems or computer program products. Accordingly, the application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the application can be embodied in the form of a computer program product on one or more computer-usable non-transitory storage media having computer-usable program code embodied in the medium.

[0090] The application can provide computer program instructions to the management platform of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing devices to generate a machine, so that the instructions executed by the management platform of the computer or other programmable data processing devices generate a device for implementing the system.

[0091] These computer program instructions can also be stored in a computer-readable storage medium that can guide the computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable storage medium generate a product including instruction devices, which implement the functions of the system.

[0092] These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions of the system.

Claims

1. A gearbox gear position self-learning system, characterized in that, include: The gear selection displacement deviation learning module is used to push the gear selection displacement to the limit position by applying a negative gear selection duty cycle, calculate the gear selection displacement offset, and correct the gear selection value based on the offset. The neutral displacement learning module is configured to perform temporary neutral position learning and multiple neutral position sampling, and take the average value to generate the neutral self-learning displacement. The shift displacement learning module is used to generate a shift self-learning displacement by cyclically switching the target gear and recording the shift displacement, and taking the average value of multiple samples. The gear selection displacement self-learning module is used to control the gear selection displacement to both ends of the shift position and take the intermediate value during the gear shifting process. After multiple samplings, the gear selection self-learning displacement is generated. The error elimination module uses the average value of multiple self-learning data. The execution control unit is used to adjust the gearbox shift logic based on the self-learning results.

2. The system according to claim 1, characterized in that: The gear selection displacement deviation learning module determines the leftmost extreme position of the gear selection displacement through displacement limit detection, takes the difference between the extreme position and the preset default value of the gear selection displacement as the gear selection displacement offset, and superimposes the gear selection value and the offset in real time to generate a temporary gear selection displacement.

3. The system according to claim 1, characterized in that: The neutral gear displacement learning module generates a temporary neutral position by engaging first and second gear and recording the average value of the shift displacement. It then performs neutral gear displacement sampling at the midpoint of different gear selections and takes the average value to obtain the neutral gear self-learning displacement.

4. The system according to claim 1, characterized in that: The shift displacement learning module switches from the lowest gear to the highest gear in an incremental manner, and then switches from the lowest gear to the highest gear in an incremental manner. In each cycle, it records the last shift displacement of each gear and takes the average value of the shift displacement as the shift self-learning displacement.

5. The system according to claim 1, characterized in that: The gear selection displacement self-learning module pushes the gear selection displacement to the two vertices of the current gear shift position, and takes the middle value as the single gear selection displacement self-learning value based on the displacement data of the two vertices.

6. The system according to claim 1, characterized in that: The error elimination module records successful self-learning values ​​during the shift displacement and gear selection displacement self-learning process, and optimizes the final self-learning displacement by removing outliers and taking the average value.

7. The system according to claim 1, characterized in that: It also includes a gear selection offset correction module, which constructs a displacement response curve and combines multiple dynamic sampling data to fit the deviation between the actual displacement and the theoretical value. The corrected offset calculation formula is as follows: In the formula, This is the corrected offset; This refers to the limit position of the selected gear displacement. The default value for the selected gear displacement; This is a nonlinear compensation factor; This represents the instantaneous slope of the displacement-duty cycle curve at the extreme position. This is the preset maximum slope threshold.

8. The system according to claim 1, characterized in that: It also includes a weighted average optimization module, which samples displacement at the midpoint of different gear selections and assigns dynamic weights based on the variance of the displacement distribution. The formula for the neutral gear reference position is: In the formula, This is the neutral reference position; These are the normalized weighting coefficients; Let be the displacement value of the empty space during the i-th sampling.

9. A method for self-learning gear positions in a transmission, characterized in that: The method is implemented based on the system described in any one of claims 1-8: The method includes: The corrected gear selection displacement is generated by extreme position detection and offset calculation; The neutral reference position is determined based on the average of multiple samplings. Gear shift self-learning displacement is generated by cyclically switching gears and recording displacement. The gear selection self-learning displacement is generated by combining gear shift position boundary detection and intermediate value calculation. The average value of multiple self-learning data is taken to optimize the final displacement parameters.

10. A computer-readable storage medium storing a computer program, characterized in that: The computer program is executed to perform the method of claim 9.

Images