Continuously variable speed zeroing control system based on direct-axis electric drive axle

Abstract

The present invention relates to the technical field of continuously variable speed control, and discloses a continuously variable speed zeroing control system based on a direct-axis electric drive axle. The system includes signal acquisition, torque analysis, zeroing control, mode switching, reference calibration, and execution output modules. The signal acquisition module acquires speed signals and torque data and sets a zeroing detection interval; the torque analysis module divides torque compensation nodes and generates characteristic vectors; the zeroing control module establishes correction rules and obtains compensation parameters; the mode switching module identifies output modes and calculates speed fluctuation amplitudes; the reference calibration module derives the optimal zeroing threshold and generates a torque deviation sequence; and the execution output module integrates and forms a control scheme. The system realizes torque dynamic compensation, speed fluctuation suppression, and threshold adaptive calibration through multi-module collaboration, thereby improving the stability and control accuracy of the electric drive axle under complex working conditions, and is suitable for continuously variable speed control of electric vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of continuously variable speed control, and in particular to a continuously variable speed zeroing control system based on a direct-axis electric drive bridge. Background Art

[0002] In the transmission system field of new energy vehicles and electric vehicles, direct-axis electric drive axles are core power components. Their transmission efficiency and control accuracy directly affect the vehicle's dynamic performance, energy consumption, and driving comfort. Traditional electric drive axle speed control systems commonly suffer from problems such as unstable torque output and large speed fluctuations when facing complex operating conditions. In particular, during zero control, it is difficult to accurately compensate for dynamic torque changes. This leads to power interruptions, impact vibrations, and other phenomena during gear shifts or sudden load changes in the transmission system. This seriously restricts the application and promotion of electric drive axles in high-end equipment and new energy vehicles.

[0003] In existing technologies, most zeroing control systems employ fixed-threshold torque compensation strategies, lacking the ability to adapt to real-time operating conditions. For example, traditional solutions typically employ compensation based on a preset torque-speed mapping table. However, in actual operation, dynamic changes in the drive axle's load fluctuation rate, phase offset parameters, and speed shifting phases can lead to compensation lag or overcompensation, failing to meet the requirements for smooth torque transition during continuously variable transmission. Furthermore, existing systems analyze torque characteristics in a single dimension, failing to fully consider the coupling relationship between torque type, drive frequency, and phase characteristics. This makes it difficult to construct multi-parameter coordinated zeroing correction rules, resulting in insufficient system stability under high-frequency disturbance conditions.

[0004] With the increasing demand for intelligent and efficient transmission systems in electric vehicles, developing a continuously variable zeroing control system capable of real-time sensing of operating conditions and dynamically adjusting compensation strategies has become a pressing technical challenge for the industry. Key technical bottlenecks in improving electric drive axle transmission performance are achieving multi-node phase resolution of torque data, building an adaptive mode switching network, and deriving the optimal zeroing threshold. Existing technologies are limited in multi-module coordinated control, dynamic parameter matching, and adaptability to complex operating conditions, necessitating a more advanced control solution to overcome the constraints of traditional technologies. Summary of the Invention

[0005] The object of the present invention is to provide a continuously variable speed zeroing control system based on a direct-axis electric drive bridge to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a continuously variable speed zeroing control system based on a direct-axis electric drive bridge, the system comprising:

[0007] The signal acquisition module is used to obtain the speed signal and torque output data of the drive axle and set the zero detection interval corresponding to the speed change working condition;

[0008] The torque analysis module is used to divide the zero detection interval into multiple torque compensation nodes, perform phase analysis on the torque data of each node, and generate the torque characteristic vector corresponding to the node;

[0009] A zeroing control module is used to extract a dynamic compensation index from a torque characteristic vector, establish a zeroing correction rule associated with a torque compensation node, and obtain a phase compensation parameter corresponding to the rule;

[0010] The mode switching module is used to identify the drive output mode in the phase compensation parameters, adaptively switch the dynamic compensation indicators according to the output mode, and calculate the speed fluctuation amplitude of each node under different switching strategies;

[0011] A benchmark calibration module is used to derive the optimal zeroing threshold value based on the speed fluctuation amplitude, and generate a torque deviation sequence by comparing the current torque output value with the optimal zeroing threshold value;

[0012] The output module is executed to analyze the torque deviation sequence and integrate it into a continuously variable speed zeroing control scheme based on the speed fluctuation direction of the node.

[0013] Preferably, the torque analysis module is implemented by: constructing a driving feature library corresponding to the torque compensation node, the driving feature library including a driving parameter vector mapped to the speed signal and torque data;

[0014] Equivalent phase matching is performed on the driving parameter vector, and the torque compensation interval of the driving parameter vector is divided according to the matching result; the distribution critical point of the torque data is extracted from the torque compensation interval, and the distribution critical point is set as the torque compensation node.

[0015] Preferably, the torque compensation interval for dividing the driving parameter vector further includes:

[0016] According to the torque type and speed change phase in the drive parameter vector, the drive frequency, load fluctuation rate and phase offset parameters are extracted, and the drive feature label is generated based on the above parameters;

[0017] The driving feature labels are associated with the driving parameter vectors. By calculating the torque equivalence between the feature labels, the driving parameter vectors with equivalence higher than the preset torque threshold are screened to form a torque compensation interval.

[0018] Preferably, the method for implementing the torque characteristic vector corresponding to the generating node includes:

[0019] For each torque compensation node, obtain the node's speed fluctuation data within a preset period based on the node's timing position in the zero detection interval, and calculate the node's speed fluctuation coefficient;

[0020] When the speed fluctuation coefficient exceeds the first zeroing threshold, the node is marked as a high-frequency disturbance node, and its torque data is extracted to form a torque feature vector; when the speed fluctuation coefficient is lower than the first zeroing threshold, the node is marked as a steady-state node, and the torque data of the node's adjacent nodes are phase-superimposed, and the superimposed data are reconstructed into a torque feature vector.

[0021] Preferably, the implementation of the zeroing control module includes:

[0022] Separating the driving torque proportion, no-load torque proportion and phase deviation parameters from the torque characteristic vector, and generating a zero adjustment correction rule for the torque compensation node based on the above parameters;

[0023] If the number of compensation nodes covered by the current zeroing correction rule is less than the preset compensation threshold, the torque characteristic vectors of adjacent torque compensation nodes are traversed, and the compensation indicators not included in the correction rules of the adjacent nodes are added to the current rule.

[0024] Preferably, the mode switching module is implemented by: acquiring a phase factor of a switching frequency and a switching factor of a load response rate in a drive output mode;

[0025] A mode transition network associated with the phase factor and switching factor is constructed, and the speed fluctuation amplitude under different switching strategies is determined according to the transition probability of each path in the network.

[0026] Preferably, the constructing mode transfer network further comprises:

[0027] Identify the periodic characteristics of the phase factor, and if the current periodic characteristics completely match the preset driving period, set the phase factor as the starting node of the mode transfer network;

[0028] Calculate the switching correlation between the phase factor and the switching factor, and generate the intermediate nodes and terminal nodes of the mode transfer network in descending order of correlation;

[0029] The phase of the terminal node is traced back, and when the correlation degree of the terminal node is lower than the preset switching threshold, it is output as the final path of the mode transfer network.

[0030] Preferably, the calculation of the speed fluctuation amplitude is implemented by:

[0031] The mean value of the phase factor and the range of the switching factor of each terminal node in the mode transfer network are statistically analyzed, and the global covariance of all node factors is calculated;

[0032] The phase fluctuation coefficient is obtained by subtracting the mean phase factor of a single terminal node from the mean phase factor of the adjacent nodes and dividing it by the global covariance. At the same time, the ratio of the switching factor range to the global covariance is calculated, and the weighted sum of the two is taken as the speed fluctuation amplitude of the node.

[0033] Preferably, the implementation of deriving the optimal zeroing threshold includes:

[0034] Extract the disturbance pattern closest to the current speed fluctuation amplitude in the historical data, and calculate the Euclidean distance between the two in phase distribution as the first calibration reference value;

[0035] Counting the difference in the number of peak points between the current speed fluctuation amplitude and the historical disturbance pattern, and using the difference as a second calibration reference value;

[0036] Based on the linear combination of the first calibration reference value and the second calibration reference value, an optimal zeroing threshold in a preset zeroing threshold table is matched.

[0037] Preferably, the implementation of the execution output module includes: dividing the positive compensation interval and the negative compensation interval according to the speed fluctuation direction of each node in the torque deviation sequence;

[0038] The attenuation rate of the torque deviation in the positive compensation interval and the increase rate of the torque deviation in the negative compensation interval are extracted, and the two are weightedly fused according to the load weight of the torque compensation node to generate the compensation parameters of the continuously variable speed zeroing control scheme.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] Through the collaborative work of multiple modules, precise dynamic control of the torque output of the electric drive axle is achieved, significantly improving the stability and response speed of the transmission system under complex working conditions. The signal acquisition module acquires the speed signal and torque data in real time and sets the zero detection interval, providing basic data support for subsequent precise analysis, enabling the system to specifically process key data intervals under speed change conditions. The torque analysis module achieves multi-dimensional analysis of torque data by constructing a drive feature library, dividing the torque compensation interval and setting compensation nodes. It not only considers the influence of torque type and speed change stage, but also accurately selects parameter vectors with similar torque characteristics through drive feature labels and equivalence calculations, making the generation of torque feature vectors more in line with actual working conditions and providing a reliable basis for dynamic compensation.

[0041] The zeroing control module extracts multiple parameters from the torque characteristic vector to generate correction rules and supplements compensation indicators by traversing adjacent nodes, ensuring the comprehensiveness and adaptability of the rules. It can dynamically adjust the compensation strategy based on real-time torque characteristics, avoiding the lag and limitations of traditional fixed threshold solutions. The mode switching module constructs a mode transfer network and combines correlation analysis between phase factors and switching factors to achieve intelligent identification of drive output modes and optimize dynamic switching strategies. It accurately calculates the speed fluctuation amplitude at each node, enabling the system to automatically select the optimal compensation path under different operating conditions, effectively suppressing speed fluctuations and improving transmission smoothness.

[0042] The benchmark calibration module compares historical data with current fluctuation amplitudes and matches the optimal zeroing threshold with a linear combination of reference values, achieving dynamic adaptive threshold calibration. Compared to traditional preset threshold methods, this method is more adaptable to real-time changes in operating conditions and improves the accuracy of torque deviation detection. The execution output module divides the compensation interval according to the direction of speed fluctuation and generates compensation parameters based on the attenuation and amplification rates and load weight. This allows for refined processing of torque deviations, enabling the continuously variable speed zeroing control scheme to dynamically adjust compensation strength based on actual needs, further improving the system's control accuracy and energy efficiency.

[0043] Overall, the system forms a complete closed loop from data collection, feature analysis, rule establishment to dynamic control through the organic coordination of multiple modules and in-depth data mining and analysis. It not only solves the problems of compensation lag and insufficient stability of traditional systems under complex working conditions, but also significantly improves the transmission efficiency and driving comfort of the electric drive axle, providing advanced technical solutions for the optimization of transmission systems of electric vehicles and high-end equipment, and has significant engineering application value and market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a working principle diagram of the stepless speed zeroing control system based on the direct-axis electric drive axle of the present invention;

[0045] Figure 2 Construct the working principle diagram of the torque compensation node for the torque analysis module;

[0046] Figure 3 Working principle diagram for torque eigenvector generation;

[0047] Figure 4 This is the working principle diagram for establishing zero correction rules and expanding compensation indicators. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] See also Figures 1-4 The present invention relates to a continuously variable speed zeroing control system based on a direct-axis electric drive axle. The system includes: a signal acquisition module, a torque analysis module, a zeroing control module, a mode switching module, a reference calibration module, and an execution output module. The specific implementation steps are as follows:

[0050] Signal Acquisition Module: This module uses sensors to acquire real-time drive axle speed signals and torque output data. It also sets corresponding zero-check intervals based on different shifting conditions (such as starting, acceleration, and deceleration). For example, during vehicle start-up, the zero-check interval can be set from 0 to 30% of the rated speed to specifically monitor torque output stability under low-speed conditions.

[0051] Torque Analysis Module: This module divides the zero detection interval into multiple torque compensation nodes. Each node corresponds to a specific torque output state. The module performs phase analysis on the torque data of each node and generates a vector that reflects the torque characteristics of that node. This vector includes parameters such as torque amplitude, phase offset, and fluctuation frequency.

[0052] Zeroing control module: extracts dynamic compensation indicators (such as torque deviation rate, phase correction coefficient, etc.) from the torque characteristic vector, establishes zeroing correction rules associated with each torque compensation node, and obtains phase compensation parameters corresponding to the rules, such as phase advance compensation parameters for high-frequency disturbance nodes or phase lag compensation parameters for steady-state nodes.

[0053] Mode Switching Module: This module identifies the drive output mode (e.g., constant torque mode, variable torque mode) within the phase compensation parameters and adaptively switches the dynamic compensation indicators based on the output mode. It also calculates the speed fluctuation amplitude of each node under different switching strategies. This amplitude reflects the impact of torque adjustment on speed stability.

[0054] The benchmark calibration module derives the optimal zeroing threshold based on the speed fluctuation amplitude. This threshold is determined by comparing the torque output performance under similar operating conditions in historical data. The module compares the current torque output value with the optimal zeroing threshold to generate a sequence of data representing the torque deviation.

[0055] The execution output module analyzes the torque deviation sequence and integrates it into a specific continuously variable speed zeroing control scheme based on the speed fluctuation direction (positive or negative) of each node. The scheme includes the compensation direction and compensation amount for each node to achieve real-time correction of the drive axle torque output.

[0056] The technical solution of the present invention is further described in detail below with reference to specific embodiments.

[0057] Example 1: In this embodiment, the torque analysis module is implemented as follows: First, a drive feature library corresponding to the torque compensation node is constructed. This library is based on historical operating data of the direct-axis electric drive axle under different operating conditions. This library contains speed signals and torque data collected in real time by speed and torque sensors and preprocessed. This data is mapped into a multidimensional drive parameter vector and stored in the library. The drive parameter vector includes basic features such as speed amplitude, torque mean, and phase angle, as well as derived features such as trend terms and periodic components derived from time series analysis, forming a comprehensive description of the drive axle's operating status.

[0058] When performing equivalent phase matching on the drive parameter vectors, a similarity calculation method based on the absolute value of the phase difference is adopted. Specifically, for any two drive parameter vectors, the absolute value of the point-by-point difference in their phase angle sequence is calculated, and the average is taken as the phase matching index. According to the preset phase matching threshold, vectors with a phase matching degree higher than the threshold are divided into the same group, thereby forming different torque compensation intervals. For example, under a certain speed change condition, the drive parameter vector can be divided into multiple torque compensation intervals such as the starting acceleration interval, the constant speed cruising interval, and the deceleration braking interval through this matching method. The vectors in each interval have similar phase characteristics, reflecting the common operating characteristics of the drive axle in this stage.

[0059] To extract the critical points of the torque data distribution within the torque compensation interval, the torque data within each interval is first statistically analyzed to calculate the probability density distribution of the torque values. By finding the local minima of the probability density function, the boundary points of the torque data distribution are determined, and these boundary points are referred to as critical points. For example, in the torque data distribution during the starting acceleration interval, when the torque changes from a slow increase to a rapid increase, the corresponding torque value is a critical point in the distribution. This point is set as a torque compensation node, which is used to identify the transition point in the drive axle's operating state.

[0060] When dividing the torque compensation interval of the drive parameter vector, it is also necessary to extract characteristic parameters based on the torque type and speed change stage in the drive parameter vector. Torque types are divided into driving torque, no-load torque, braking torque, etc., and speed change stages include starting, low speed, medium speed, high speed, deceleration, and parking. For different torque types and speed change stages, the driving frequency (the number of torque changes per unit time), load fluctuation rate (the ratio of the load change to the average load), and phase offset parameter (the difference between the actual phase and the ideal phase) are extracted. For example, under the driving torque type of the high-speed and uniform speed stage, the driving frequency is low, the load fluctuation rate is small, and the phase offset parameter is stable within a small range. These parameters are combined to generate the corresponding driving characteristic label, such as "high speed-driving torque-low fluctuation-stable phase".

[0061] After associating the drive feature labels with the drive parameter vectors, the torque equivalence between the feature labels is calculated to filter the vectors and form a torque compensation interval. The torque equivalence is calculated based on the parameter differences in each dimension of the feature labels. A weighted Euclidean distance formula is used to assign different weights to dimensions such as torque type, speed change phase, drive frequency, load fluctuation rate, and phase offset parameter. The distance between two feature labels is then calculated; the smaller the distance, the higher the torque equivalence. The preset torque threshold is an empirically determined distance threshold. When the distance between two feature labels is less than this threshold, the corresponding drive parameter vectors are considered to have a high degree of equivalence and are placed in the same torque compensation interval. For example, within a given torque compensation interval, if the torque equivalence of multiple drive feature labels is higher than the preset threshold, this indicates that the drive parameter vectors corresponding to these labels have high similarity in torque characteristics and operating phase, and can be analyzed and processed uniformly as data samples within the same interval.

[0062] Through the above steps, the torque analysis module accurately divides the zeroing detection interval into multiple torque compensation nodes and analyzes the torque data of each node based on phase and characteristic parameters, providing basic data support for the subsequent generation of torque characteristic vectors and the establishment of zeroing correction rules. This process achieves refined modeling of the operating state of the direct-axis electric drive axle through multi-dimensional analysis and feature extraction of drive parameters, ensuring that the division of torque compensation nodes accurately reflects the torque variation characteristics of the drive axle under different operating conditions, laying the foundation for subsequent processing steps in the continuously variable speed zeroing control system.

[0063] Example 2: In this embodiment, the torque characteristic vector corresponding to the node is generated as follows: For each torque compensation node within the zero detection interval, first determine its time range within a preset period based on the node's position in the time series. The preset period can be set according to the characteristics of the speed change condition, for example, 0.5 seconds in the starting condition and 2 seconds in the high-speed cruising condition, to ensure that the acquired speed fluctuation data can fully reflect the dynamic characteristics of the node in the typical operating stage. The speed signal within this period is collected by a speed sensor, and after digital filtering to remove high-frequency noise, the deviation sequence between the real-time value of the speed and the mean value is calculated to obtain the speed fluctuation data.

[0064] The speed fluctuation coefficient is calculated based on the statistical characteristics of the deviation sequence. Specifically, it is calculated as the ratio of the standard deviation to the mean of the deviation sequence. This coefficient quantifies the relative magnitude of speed fluctuation. For example, if the mean speed of a node within a preset period is 1000 rpm and the standard deviation is 50 rpm, the speed fluctuation coefficient is 5%. By setting the first zeroing threshold (for example, 3%), nodes can be classified as high-frequency disturbance nodes or stable nodes.

[0065] When the speed fluctuation coefficient exceeds the first zeroing threshold, the node is determined to be a high-frequency disturbance node. At this point, the torque sensor output data for that node within a preset period is directly extracted, including the instantaneous torque value sequence, peak values, valley values, and other parameters. These data are arranged in chronological order to form a torque feature vector. For example, the torque data of a high-frequency disturbance node may exhibit periodic fluctuations or sudden spikes. Its feature vector can be represented as [t1, t2, t3, ..., tn], where ti is the torque value at the i-th sampling moment. This reflects the dynamic changes in torque at that node and is used to formulate subsequent zeroing correction rules.

[0066] When the speed fluctuation coefficient is lower than the first zeroing threshold, the node is marked as a steady-state node. Since the speed fluctuation of the steady-state node is small, it is difficult to capture significant features by analyzing its torque data alone. Therefore, it is necessary to combine the torque data of adjacent nodes for collaborative processing. Specifically, N adjacent nodes before and after the steady-state node are selected (N can be set according to the complexity of the working conditions, such as 2-5), and the torque data of these nodes are phase-superimposed. The method of phase superposition is: based on the phase of the steady-state node, the torque data of the adjacent nodes are time-shifted according to their phase difference, so that the torque waveforms of each node are aligned on the time axis, and then arithmetic averaging or weighted averaging is performed (the weight can be set according to the distance from the steady-state node, and the weight of the neighboring nodes is higher) to obtain the superimposed torque data sequence.

[0067] When reconstructing the superimposed data to form a torque eigenvector, dimensionality reduction is required to retain the primary eigenvalues. For example, principal component analysis (PCA) can be used to extract the first k principal components (k is determined based on the data complexity, such as 2-3). The coefficients of these principal components are then used as the elements of the eigenvector. This reconstructed torque eigenvector comprehensively reflects the torque distribution trend of the steady-state node and its adjacent areas, avoiding the problem of missing features caused by small fluctuations in the data of a single node.

[0068] In practical applications, the timing position needs to be determined in conjunction with the operating status signals of the drive axle, such as the gear position signal and accelerator pedal position signal obtained via the CAN bus, to ensure that the preset cycle is synchronized with the time node of the current operating condition. For example, when the accelerator pedal is pressed, a preset cycle is set starting at that moment to collect speed and torque data during the starting and acceleration phase, ensuring that the node analysis accurately corresponds to the operating condition phase.

[0069] Furthermore, the acquisition frequency of speed fluctuation data must comply with the Nyquist sampling theorem to avoid frequency aliasing. Typically, a sampling frequency at least twice the maximum speed signal frequency (e.g., 100 Hz) is used to ensure accurate calculation of the speed fluctuation coefficient. For the extraction and overlay of torque data, synchronous clock calibration is required to ensure consistency of data across different nodes on the timeline and avoid phase misalignment caused by clock skew.

[0070] Through this method, the generated torque feature vector can adaptively select different processing strategies based on the node's disturbance characteristics: for high-frequency disturbance nodes, the original torque data is directly extracted to retain its dynamic details; for steady-state nodes, the regionalized torque distribution characteristics are mined by superimposing and reconstructing the data of adjacent nodes. This differentiated processing method not only ensures a rapid response to abnormal fluctuations, but also improves the reliability of data features under steady-state conditions, providing accurate input data for the subsequent zeroing control module to extract dynamic compensation indicators. The entire process achieves efficient analysis of torque data and accurate generation of feature vectors through the synergistic effects of timing synchronization, data filtering, feature calculation, and reconstruction.

[0071] Example 3: In this embodiment, the zeroing control module is implemented as follows: First, the driving torque ratio, no-load torque ratio and phase deviation parameter are separated from the torque characteristic vector. The torque characteristic vector is a multidimensional data sequence generated by the torque analysis module, which contains information such as the instantaneous value, mean value, peak value, etc. of the torque on the time axis. The driving torque ratio is obtained by calculating the ratio of the numerical sum of the driving torque interval in the characteristic vector to the numerical sum of the entire vector, and the no-load torque ratio is determined by the ratio of the numerical sum of the no-load torque interval to the vector sum. The phase deviation parameter is obtained by comparing the difference between the actual phase sequence of the torque characteristic vector and the ideal phase sequence (preset based on the design working conditions of the direct-axis electric drive bridge), reflecting the degree of phase deviation of the torque output.

[0072] When generating zeroing correction rules for the torque compensation node based on the aforementioned parameters, a mapping relationship between the parameters and the correction strategy must be established. For example, when the driving torque ratio exceeds a preset threshold (e.g., 70%) and the phase deviation parameter is greater than the allowable range (e.g., ±5°), the node is determined to need to increase the no-load torque compensation to balance the phase deviation. When the no-load torque ratio is too high (e.g., exceeding 30%) and the phase deviation is in the negative range, the driving torque output ratio is determined to be reduced. These rules are developed through expert experience and historical data training and stored in the zeroing control module's rule library as tables or conditional statements for easy real-time access and matching.

[0073] If the number of compensation nodes covered by the current zero-correction rule is less than a preset compensation threshold (e.g., 50% of the total number of nodes), the adjacent node rule traversal mechanism is triggered. During this traversal, the torque feature vectors of the current node's adjacent torque compensation nodes (determined by their temporal position, e.g., three nodes each preceding and following) are obtained. The driving torque percentage, no-load torque percentage, and phase deviation parameters in these vectors are extracted and compared with the parameters in the current rule. If the adjacent node parameters contain a compensation metric not covered by the current rule (e.g., the phase deviation parameter of an adjacent node exceeds the current rule's processing range), the correction strategy corresponding to that metric (e.g., a step-by-step compensation method for out-of-range phase deviations) is added to the current rule.

[0074] During rule expansion, it's important to ensure the compatibility of newly added compensation metrics with existing rules. For example, if the current rule primarily targets positive phase deviation compensation, but the negative phase deviation metric of an adjacent node isn't covered, a branch addressing negative deviations can be added to the existing rule to create a bidirectional compensation rule. Furthermore, by calculating the correlation between the new metric and existing parameters (such as the Pearson correlation coefficient), the weight of its impact on zeroing control can be determined. If the correlation falls below a preset value (e.g., 0.3), the metric is ignored to avoid rule redundancy.

[0075] The update frequency of the zeroing correction rules is synchronized with the zeroing detection interval division period, typically triggering a rule optimization process after each zeroing detection interval. For example, after a complete speed change cycle (such as from start to stop), the system calculates the rule coverage of all torque compensation nodes within that interval. If the preset compensation threshold is not reached, the system initiates adjacent node traversal and rule expansion to ensure that the rule coverage in the next detection interval meets the control requirements.

[0076] In practical applications, the calculation of the driving torque percentage and the no-load torque percentage must account for dynamic switching of torque types. For example, when a direct-axis electric drive axle switches from driving to braking, the torque type changes from driving torque to braking torque. At this point, the torque range must be redefined and the percentage calculation method adjusted. The calculation of the phase deviation parameter is based on a real-time phase tracking algorithm. Using phase-locked loop (PLL) technology, the actual phase is extracted from the speed signal and compared in real time with the preset ideal phase to ensure the accuracy of the deviation calculation.

[0077] The zeroing control module and the torque analysis module interact via a real-time data interface. Once generated, torque eigenvectors are immediately transmitted to the zeroing control module for parameter separation and rule matching. The system uses a priority queue to process torque eigenvectors from high-frequency disturbance nodes, ensuring rapid response to sudden torque fluctuations. For reconstructed eigenvectors from steady-state nodes, batch processing is used to improve computational efficiency.

[0078] Through this approach, the zeroing control module dynamically generates and expands zeroing correction rules based on the real-time parameters of the torque characteristic vector, achieving precise control of the torque compensation node. This process, through the synergistic effects of parameter separation, rule matching, adjacent node traversal, and indicator expansion, ensures that the zeroing correction rules are adaptable to the torque output characteristics of the direct-axis electric drive axle under different operating conditions, providing a reliable rule basis for the subsequent mode switching module and benchmark calibration module. Furthermore, through the use of preset compensation thresholds and correlation filtering mechanisms, excessive complexity and redundancy of the rules are avoided, ensuring the real-time and stability of the control system.

[0079] Example 4: In this embodiment, the mode switching module is implemented as follows: First, the phase factor of the switching frequency in the drive output mode and the switching factor of the load response rate are obtained. The phase factor characterizes the phase periodicity characteristics when the drive output mode is switched, and can be obtained by extracting the phase angle sequence of the main frequency components by Fourier transforming the drive axle speed signal; the switching factor reflects the speed of the system response when the load changes, and is determined by calculating the ratio of the time required for the torque to reach a stable value after a load step change to a preset reference time. For example, in the economic mode, the switching frequency is low, the phase factor exhibits a long period characteristic, the load response rate is slow, and the switching factor value is large; in the power mode, the switching frequency is high, the phase factor period is short, the load response rate is fast, and the switching factor value is small.

[0080] When constructing a mode transfer network associated with a phase factor and a switching factor, the network node model is first established. Nodes are categorized into three types: starting nodes, intermediate nodes, and ending nodes. The starting node represents the initial state of the drive output mode, the intermediate nodes represent the transition state during the mode switching process, and the ending node represents the stable target mode state. To identify the periodic characteristics of the phase factor, the main period is determined by calculating the autocorrelation function of the phase factor sequence. If the current main period exactly matches the preset drive period (such as the standard switching period under the design conditions), the corresponding phase factor is set as the starting node of the mode transfer network. For example, if the main period of the phase factor is detected to be 1.5 seconds, which is consistent with the preset drive period, the state node corresponding to this phase factor is used as the starting point of the network.

[0081] When calculating the switching correlation between the phase factor and the switching factor, a mutual information algorithm is used to measure their dependence. A larger mutual information value indicates a higher correlation between the two factors. The intermediate nodes and terminal nodes of the mode transition network are generated sequentially from highest to lowest correlation. For example, the phase factor-switching factor combination with the highest correlation is selected as the first intermediate node, and so on, until the combination with a correlation below the preset switching threshold (e.g., a mutual information value less than 0.2) is selected as the terminal node. While generating nodes, the transition probabilities between each node are calculated based on historical switching data to form network path weights. For example, if the number of historical switching attempts from the starting node to an intermediate node is 100, and the number of successful attempts is 85, then the transition probability of this path is set to 0.85.

[0082] When performing phase backtracking on the ending node, the system traces back from the ending node to the starting node, checking whether the correlation of each node on the path is above a preset switching threshold. If the correlation of a certain ending node is below the threshold, the node is deemed invalid and the path is removed. If the correlation of all nodes meets the conditions, the node is output as the final path of the mode transition network. For example, if the mutual information value of a certain ending node is 0.15, which is below the threshold of 0.2, the corresponding path is excluded, ensuring that the final network path contains only state transitions with high correlation.

[0083] To calculate the speed fluctuation amplitude, we first calculate the mean phase factor and the range of the switching factor for each terminating node in the mode transition network. The mean phase factor is the arithmetic average of all historical phase factors for that node, reflecting the phase characteristics under steady-state conditions. The range of the switching factor is the difference between the maximum and minimum switching factors, reflecting the fluctuation range of the load response rate. The global covariance is obtained by calculating the covariance matrix of the phase factor and switching factor for all nodes, representing the overall correlation between the changes in these two factors.

[0084] For a single terminating node, the absolute value of the difference between its phase factor mean and the mean phase factor of its adjacent nodes (previous and subsequent nodes) is calculated and divided by the square root of the global covariance to obtain the phase fluctuation coefficient. This dimensionless coefficient quantifies the relative magnitude of phase variation. Simultaneously, the ratio of the switching factor range to the global covariance is calculated to obtain the load fluctuation coefficient. The phase fluctuation coefficient and the load fluctuation coefficient are weighted and summed according to preset weights (e.g., 60% for phase fluctuation and 40% for load fluctuation) to determine the speed fluctuation amplitude for that node. For example, if the phase fluctuation coefficient of a terminating node is 0.3 and the load fluctuation coefficient is 0.2, with weights of 0.6 and 0.4, respectively, the speed fluctuation amplitude is 0.3 × 0.6 + 0.2 × 0.4 = 0.26.

[0085] In practical applications, the acquisition of phase and switching factors must be synchronized with the drive axle's Controller Area Network (CAN) bus data to ensure factor calculations are based on the same operating status at the same moment. For example, the CAN bus acquires real-time transmission gear signals and motor torque commands, and combines them with speed and torque sensor data to dynamically update the phase and switching factors. The mode transition network is constructed using an incremental learning approach. As new operating data accumulates, node associations and path transition probabilities are continuously updated, improving the network's adaptability to real-time operating conditions.

[0086] The speed fluctuation amplitude calculation frequency is consistent with the switching frequency of the drive output mode, and the calculation process is usually triggered once after each mode switch is completed. For multi-path mode transition networks, a parallel computing architecture is used to simultaneously process the fluctuation amplitude calculation for each terminal node, shortening the response time. The statistics involved in the calculation process (such as mean, range, and covariance) are updated using a sliding window algorithm. The window size is dynamically adjusted based on the stability of the operating conditions (for example, the window size is 100 sampling points under stable conditions and 50 sampling points under fluctuating conditions), ensuring that the statistical results reflect the current operating status in real time.

[0087] Through this method, the mode switching module constructs a dynamic mode transition network based on phase factors and switching factors, accurately identifying the state transition paths of the drive output mode and quantifying the impact of different switching strategies by calculating the speed fluctuation amplitude. This process, through the synergistic effect of factor extraction, network construction, path optimization, and amplitude calculation, achieves adaptive switching control of the direct-axis electric drive axle output mode. This provides the benchmark calibration module with speed fluctuation data under multiple strategies, ensuring that the continuously variable speed zeroing control system can select the optimal switching strategy based on actual operating conditions, reducing speed fluctuations and improving operational stability.

[0088] Example 5: In this example, the optimal zeroing threshold is derived as follows: First, the disturbance pattern closest to the current speed fluctuation amplitude is extracted from the historical data. The historical data is stored in the system database. Each disturbance pattern includes features such as speed fluctuation amplitude, phase distribution sequence, and number of peak points. The amplitude difference between the current speed fluctuation amplitude and each pattern in the historical data is calculated using Euclidean distance. The top N (e.g., N = 5) historical patterns with the smallest difference are selected as the candidate set.

[0089] Calculate the Euclidean distance between the current disturbance mode and each historical mode in the candidate set in phase distribution as the first calibration reference value The phase distribution is represented by the phase spectrum after discrete Fourier transform. Let the phase spectrum of the current mode be , the phase spectrum of the historical mode is , then the Euclidean distance formula is:

[0090] in, Indicates the current mode The phase value of the frequency component, Indicates the history mode The phase value of the frequency component, n is the total number of frequency components.

[0091] Count the difference in the number of peak points between the disturbance pattern corresponding to the current speed fluctuation amplitude and the historical disturbance pattern as the second calibration reference value The peak point is defined as the local maximum point in the speed fluctuation curve, which is determined by comparing the values ​​of adjacent sampling points. If the current point value is greater than the values ​​of the previous and next points, it is determined to be the peak point. The number of peak points in the current mode Number of peak points in historical patterns The absolute value difference of .

[0092] Based on the first calibration reference value With the second calibration reference value The linear combination of , matches the optimal zeroing threshold in the preset zeroing threshold table. The preset zeroing threshold table is constructed through offline simulation and engineering experience, and stores different and The threshold corresponding to the combination The linear combination expression is ,in and is the weight coefficient (automatically adjusted by the system according to the type of working condition, such as under acceleration conditions ) The optimal zeroing threshold is obtained by interpolating the K value in the threshold table .

[0093] The implementation of the execution output module involves dividing the torque compensation nodes into positive and negative compensation intervals based on the speed fluctuation direction (positive or negative) of each node in the torque deviation sequence. Positive fluctuations refer to speeds exceeding the target speed, corresponding to a negative torque deviation; negative fluctuations refer to speeds falling below the target speed, corresponding to a positive torque deviation. By traversing the torque deviation sequence, nodes with consecutive fluctuations in the same direction are grouped into the same interval. For example, if the torque deviations of five consecutive nodes are all negative, the interval is classified as positive compensation.

[0094] Extract the decay rate of the torque deviation in the positive compensation interval and the increase rate of the torque deviation in the negative compensation interval. The decay rate is defined as the rate of change of the absolute value of the torque deviation in the positive compensation interval, which is obtained by the absolute value of the slope of the torque deviation sequence in the linear fitting interval; the increase rate is defined as the rate of change of the absolute value of the torque deviation in the negative compensation interval, which is obtained by the absolute value of the slope of the torque deviation sequence in the linear fitting interval. The two are weighted and fused according to the load weight of the torque compensation node. The load weight is determined according to the load size of the drive axle corresponding to the node (such as light load weight 0.3, medium load weight 0.6, heavy load weight 0.9). The calculation formula is:

[0095] Among them, C is the compensation parameter after fusion, For the The rate of the interval (decay rate or increase rate), is the load weight of the corresponding interval, and m is the total number of intervals. The resulting continuously variable speed zeroing control scheme adjusts the torque compensation of each node according to the C value to achieve closed-loop control of speed fluctuations.

[0096] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0097] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A continuously variable speed zeroing control system based on a direct-axis electric drive bridge, characterized in that: include: The signal acquisition module is used to obtain the speed signal and torque output data of the drive axle and set the zero detection interval corresponding to the speed change working condition; The torque analysis module is used to divide the zero detection interval into multiple torque compensation nodes, perform phase analysis on the torque data of each torque compensation node, and generate a torque characteristic vector corresponding to the torque compensation node; A zeroing control module is used to extract a dynamic compensation index from a torque characteristic vector, establish a zeroing correction rule associated with a torque compensation node, and obtain a phase compensation parameter corresponding to the rule; The mode switching module is used to identify the drive output mode in the phase compensation parameters, adaptively switch the dynamic compensation indicators according to the drive output mode, and calculate the speed fluctuation amplitude of each torque compensation node under different switching strategies; A benchmark calibration module is used to derive the optimal zeroing threshold value based on the speed fluctuation amplitude, and generate a torque deviation sequence by comparing the current torque output value with the optimal zeroing threshold value; The output module is used to analyze the torque deviation sequence and integrate the torque deviation sequence into a continuously variable speed zeroing control scheme based on the speed fluctuation direction of the torque compensation node.

2. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 1 is characterized in that: The implementation of the torque analysis module includes: constructing a drive feature library corresponding to the torque compensation node, the drive feature library containing a drive parameter vector mapped to the speed signal and torque data; Equivalent phase matching is performed on the driving parameter vector, and the torque compensation interval of the driving parameter vector is divided according to the matching result; the distribution critical point of the torque data is extracted from the torque compensation interval, and the distribution critical point is set as the torque compensation node.

3. The stepless speed zeroing control system based on a direct-axis electric drive bridge according to claim 2 is characterized in that: The torque compensation interval for dividing the drive parameter vector also includes: According to the torque type and speed change stage in the driving parameter vector, the driving frequency, load fluctuation rate and phase offset parameters are extracted, and the driving feature label is generated based on the driving frequency, load fluctuation rate and phase offset parameters; The driving feature labels are associated with the driving parameter vectors. By calculating the torque equivalence between the driving feature labels, the driving parameter vectors with torque equivalence higher than a preset torque threshold are screened to form a torque compensation interval.

4. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 1 is characterized in that: The implementation methods for generating the torque characteristic vector corresponding to the torque compensation node include: For each torque compensation node, according to the time sequence position of the torque compensation node in the zeroing detection interval, the speed fluctuation data of the torque compensation node within a preset period is obtained, and the speed fluctuation coefficient of the torque compensation node is calculated; When the speed fluctuation coefficient exceeds the first zeroing threshold, the torque compensation node is marked as a high-frequency disturbance node, and its torque data is extracted to form a torque feature vector; when the speed fluctuation coefficient is lower than the first zeroing threshold, the torque compensation node is marked as a steady-state node, and the torque data of the torque compensation nodes adjacent to the steady-state node are phase-superimposed, and the superimposed data are reconstructed into a torque feature vector.

5. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 1 is characterized in that: The implementation of the zero control module includes: Separating the driving torque ratio, no-load torque ratio and phase deviation parameters from the torque characteristic vector, and generating a zero adjustment correction rule for the torque compensation node based on the driving torque ratio, no-load torque ratio and phase deviation parameters; If the number of torque compensation nodes covered by the current zeroing correction rule is less than the preset compensation threshold, the torque characteristic vectors of adjacent torque compensation nodes are traversed, and the compensation indicators not included in the zeroing correction rules of the adjacent torque compensation nodes are added to the current zeroing correction rule.

6. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 1 is characterized in that: The mode switching module is implemented by: obtaining a phase factor of a switching frequency and a switching factor of a load response rate in a driving output mode; A mode transfer network associated with the phase factor and switching factor is constructed. According to the transition probability of each path in the mode transfer network, the speed fluctuation amplitude under different switching strategies is determined.

7. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 6 is characterized in that: Building a mode transfer network also includes: Identify the periodic characteristics of the phase factor, and if the current periodic characteristics completely match the preset driving period, set the phase factor as the starting node of the mode transfer network; Calculate the switching correlation between the phase factor and the switching factor, and generate the intermediate nodes and terminal nodes of the mode transfer network in descending order of correlation; The phase of the terminal node is traced back, and when the correlation degree of the terminal node is lower than the preset switching threshold, it is output as the final path of the mode transfer network.

8. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 7 is characterized in that: The implementation methods for calculating the speed fluctuation amplitude include: The mean of the phase factor and the range of the switching factor of each terminal node in the mode transfer network are statistically analyzed, and the global covariance is obtained by calculating the covariance matrix of the phase factor and switching factor of all nodes; The phase fluctuation coefficient is obtained by subtracting the mean phase factor of a single termination node from the mean phase factor of the adjacent nodes and dividing it by the global covariance. At the same time, the ratio of the switching factor range to the global covariance is calculated, and the weighted sum of the two is taken as the speed fluctuation amplitude of the torque compensation node.

9. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 1 is characterized in that: The implementation of deriving the optimal zeroing threshold includes: Extract the disturbance pattern closest to the current speed fluctuation amplitude in the historical data, and calculate the Euclidean distance between the two in phase distribution as the first calibration reference value; Counting the difference in the number of peak points between the current speed fluctuation amplitude and the historical disturbance pattern, and using the difference as a second calibration reference value; Based on the linear combination of the first calibration reference value and the second calibration reference value, an optimal zeroing threshold in a preset zeroing threshold table is matched.

10. The continuously variable speed zeroing control system based on a direct-axis electric drive bridge according to claim 1, characterized in that: The implementation method of the execution output module includes: dividing the positive compensation interval and the negative compensation interval according to the speed fluctuation direction of each torque compensation node in the torque deviation sequence; The attenuation rate of the torque deviation in the positive compensation interval and the increase rate of the torque deviation in the negative compensation interval are extracted, and the two are weightedly fused according to the load weight of the torque compensation node to generate the compensation parameters of the continuously variable speed zeroing control scheme.

Images