A shift controller safety control test method based on edge computing

Abstract

The application discloses a kind of based on edge computing's gear controller safety control test method, comprising: collection gear controller test data and executes pre-processing, generates stage gear state sequence;Extract gear dynamics state characteristic parameters, and construct sin structure state matrix;Improved sin geometric modal decomposition is executed in edge computing node, generates gear modal evolution sequence;Dynamic envelope module based on Zonotope is constructed, and gear dynamic safety envelope sequence is recursively generated;Based on stage gear state sequence and gear modal evolution sequence, envelope crossing detection result sequence is generated;Envelope crossing detection result sequence is based on backtracking modal evolution result, and safety control test report is generated.The application realizes dynamic risk identification and safety control test function in the gear process of gear controller by introducing improved sin geometric modal decomposition method and Zonotope method.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology, and in particular to a safety control test method for a shift controller based on edge computing. Background Technology

[0002] With the development of edge computing technology, the safety testing methods for complex control systems are gradually evolving from centralized offline analysis to real-time testing at the edge. Traditional safety testing methods typically assess the safety of the control system's operating status by collecting multi-channel operational data and performing statistical analysis or threshold judgments on key parameters. However, during multi-stage dynamic operation, various operating parameters often exhibit coupling relationships and nonlinear changes. Especially during state transitions or load changes, the system's operating status can experience transient fluctuations and local anomalies. Traditional testing methods, relying on fixed thresholds or single parameter judgments, struggle to accurately identify dynamic trends under complex operating conditions, making it difficult to detect potential risks in a timely manner. Furthermore, offline testing methods usually require uploading collected data to a central server for processing, resulting in significant response delays and insufficient real-time performance, failing to meet the real-time safety testing requirements of complex control systems.

[0003] Current safety testing technologies for gear shift controllers typically involve collecting test data such as speed, torque, and actuator status, extracting feature parameters based on time-domain analysis or frequency-domain decomposition methods, and then combining this with fixed safety boundaries for status determination. These methods often employ traditional signal decomposition techniques, making it difficult to effectively separate the coupling relationships between different dynamic modes. This results in the inaccurate extraction of dynamic features such as execution hysteresis, torque surges, and speed mismatch during gear shifting. Furthermore, existing testing methods usually use static safety thresholds or fixed boundaries for risk assessment, which are ill-suited to the dynamically changing operating characteristics at different gear shifting stages, easily leading to misjudgments or omissions.

[0004] Therefore, how to provide a safety control testing method for shift controllers based on edge computing is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a safety control testing method for shift controllers based on edge computing. This invention utilizes edge computing technology, symplectic geometric mode decomposition, and a Zonotope-based dynamic envelope modeling method to dynamically model and assess the safety of multi-source operational data during shift controller testing. It details the process of achieving safe control testing of the shift controller through shift mode evolution analysis, dynamic safety envelope construction, and disturbance injection testing. During shifting, a shift mode evolution sequence is constructed, and the actual shift response trajectory and disturbance test trajectory are mapped and analyzed using the dynamic safety envelope boundary. This enables risk state identification and risk propagation path tracing during the shifting process. Edge computing nodes are used to complete real-time data processing and safety determination, reducing system latency and improving testing efficiency. This invention can accurately characterize the dynamic changes in the shifting process and has the advantages of high testing accuracy, strong dynamic adaptability, and strong risk identification capability.

[0006] A safety control test method for a shift controller based on edge computing according to an embodiment of the present invention includes:

[0007] The test data of the shift controller is collected and preprocessed at the edge computing node. The single shift process is segmented into stages to generate a staged shift state sequence.

[0008] Based on the phased shift state sequence, the shift dynamic state feature parameters are extracted, position-type state components and momentum-type state components are constructed, a symplectic state vector sequence is generated, and a symplectic structure state matrix is ​​constructed.

[0009] Improved symplectic geometric mode decomposition is performed at the edge computing node. The position-type state components and momentum-type state components are rewritten as symplectic state components in screw form. A screw extended symplectic basis is constructed, and the symplectic structure state matrix is ​​extended into a sparse state tensor. Tensorized symplectic decoupling is performed to generate a shift mode evolution sequence.

[0010] Based on the shift mode evolution sequence, a dynamic envelope module based on Zonotope is constructed. The envelope center bias and generation vector are constructed based on the shift mode features. The weights are adjusted in combination with the shift stage to generate a Zonotope generation vector group sensitive to the generation stage. A Zonotope safety set is constructed and a shift dynamic safety envelope sequence is generated recursively.

[0011] Perturbation injection is performed based on the phased shift state sequence and shift mode evolution sequence, and the actual shift response trajectory and the perturbation test trajectory are mapped to the shift dynamic safety envelope sequence to generate an envelope crossing detection result sequence.

[0012] Based on the envelope crossing detection result sequence backtracking modal evolution results, the dominant unstable mode is identified and the risk propagation path is determined, and a safety control test report is generated.

[0013] Optionally, the test data of the shift controller specifically includes input shaft speed data, output shaft speed data, transmission torque data, actuator displacement data, shift solenoid valve drive current data, clutch engagement pressure data, and shift stage indicator signal data.

[0014] Optionally, generating the phased shift state sequence includes:

[0015] The various data in the shift controller test data are arranged in time sequence according to a unified sampling time to form the original test data sequence. The original test data sequence is preprocessed, including time alignment, outlier removal and normalization, to generate a standard test data sequence.

[0016] The single gear shift process is segmented into stages according to the sliding time window and the shift stage identifier signal. Within each time window, the corresponding shift stage label is determined according to the changing trend of the standard test data sequence, forming a staged shift state sequence.

[0017] Optionally, constructing the symplectic structure state matrix includes:

[0018] Based on the phased shift state sequence, the input shaft speed data, output shaft speed data, transmission torque data and actuator displacement data corresponding to each time window are read, and the difference between the data of two adjacent sampling times is calculated in chronological order to obtain the change in input shaft speed, change in output shaft speed, change in transmission torque and change in actuator displacement.

[0019] Position-type state components are constructed based on the changes in input shaft speed and output shaft speed, and momentum-type state components are constructed based on the changes in transmission torque and actuator displacement.

[0020] A single-time symplectic state vector is constructed based on position-type state components and momentum-type state components. The single-time symplectic state vectors corresponding to each sampling time are concatenated in chronological order to generate a symplectic state vector sequence.

[0021] Based on the symplectic state vector sequence, the symplectic state vectors corresponding to each sampling time are used as column vectors of the symplectic structure state matrix, and arranged sequentially according to the sampling time to form the symplectic structure state matrix.

[0022] Optionally, generating the shift mode evolution sequence includes:

[0023] Based on the symplectic structure state matrix, an improved symplectic geometric mode decomposition is performed. The position-class state components and momentum-class state components corresponding to each sampling time are read and rewritten as symplectic state components in spinor form. The Lie algebra quantity, which characterizes the angular momentum rotation relationship, is introduced. Heterogeneous symplectic-Lie algebra splicing is performed. The blocks are spliced ​​in the order of spinor symplectic state components first and Lie algebra quantity second to form a heterogeneous symplectic Lie state block sequence.

[0024] Based on the heterogeneous syn-Li state block sequence, syn-base construction processing is performed on the state blocks corresponding to each sampling time to obtain the corresponding syn-base matrix. The main diagonal structure is extracted as the syn-base principal block. A syn-fractal basis recursive structure is introduced. According to the golden ratio recursive scaling method, the syn-base principal block corresponding to each sampling time is subjected to hierarchical expansion to form multi-layer self-similar syn-base sub-blocks, and a syn-fractal basis recursive sequence is constructed.

[0025] Based on the symplectic fractal basis recursive sequence, spectral domain compression processing is performed on the symplectic structure state matrix, and the symplectic state components, symplectic principal blocks and symplectic sub-blocks corresponding to each sampling time are mapped to a holographic symplectic spectrum matrix. The principal spectrum components are extracted based on the state coupling relationship between adjacent sampling times, and a sparse state tensor is constructed.

[0026] Tensorized symplectic decoupling is performed based on sparse state tensors. The main state component, temporal variation component, and coupling variation component are extracted along the state dimension, time dimension, and coupling dimension. They are then combined to obtain the modal coefficient sequence corresponding to each sampling time. The modal coefficient sequence is mapped to the main spectrum component in the holographic symplectic spectrum matrix to obtain the main modal component and hierarchical modal component corresponding to each sampling time.

[0027] Based on the modal coefficient sequence, principal modal components, and hierarchical modal components, each modal component is classified according to the modal amplitude change, modal temporal evolution trend, and modal coupling change relationship to generate a shift modal evolution sequence.

[0028] Optionally, the recursive generation of the shift dynamic safety envelope sequence includes:

[0029] A dynamic envelope module based on Zonotope is constructed. The dynamic envelope module consists of a modal state input unit, a vector generation construction unit, a state envelope construction unit, a dynamic envelope evolution unit, and a safety boundary determination unit.

[0030] The modal state input unit reads the dominant shift dynamic mode, execution hysteresis instability mode, torque impact disturbance mode and speed mismatch oscillation mode corresponding to each sampling moment in the shift mode evolution sequence, extracts the mode amplitude, mode hierarchy relationship and mode coupling change results, forms the modal state input sequence, and introduces the Koopman operator to perform high-dimensional linear lifting to generate the lifting state sequence;

[0031] The vector generation unit is based on the improved state sequence. It constructs the envelope center bias by using the amplitude of the dominant shift dynamics mode, and constructs the generation vector by using the execution hysteresis instability mode, torque impact disturbance mode and speed mismatch oscillation mode. It also performs weight adjustment in combination with the shift preparation stage, torque unloading stage, gear switching stage and torque reconstruction stage to generate a stage-sensitive Zonotope generation vector group.

[0032] The state envelope construction unit constructs a Zonotope safe set corresponding to each sampling time based on the envelope center bias and the stage-sensitive Zonotope generation vector group, and introduces homology theory to perform topological homology coding processing to generate homology ring labeling results, forming the initial dynamic envelope sequence;

[0033] The dynamic envelope evolution unit recursively updates the envelope center offset, the change in the generated vector, and the expansion of the envelope boundary between adjacent sampling times based on the initial dynamic envelope sequence, and synchronously updates the homology ring marking results to form a stage-recursive safe envelope set.

[0034] The safety boundary determination unit performs state determination on the envelope boundary and the homology loop marking results corresponding to each sampling time based on the stage recursive safety envelope set, and updates the determination results to the corresponding envelope boundary to form a shift dynamic safety envelope sequence.

[0035] Optionally, generating the envelope traversal detection result sequence includes:

[0036] Read the standard test data corresponding to each time window in the phased shift state sequence, and combine the dominant shift dynamic mode, execution hysteresis instability mode, torque impact disturbance mode and speed mismatch oscillation mode corresponding to each time window in the shift mode evolution sequence to determine the dominant mode type and corresponding disturbance channel corresponding to each time window;

[0037] Based on the dominant mode type and disturbance channel, disturbance injection is performed. Hysteresis disturbances are injected into the actuator displacement data channel, shift solenoid valve drive current data channel and clutch engagement pressure data channel in the time window corresponding to the hysteresis instability mode. Impact disturbances are injected into the transmission torque data channel in the time window corresponding to the torque impact disturbance mode. Offset disturbances are injected into the input shaft speed data channel and output shaft speed data channel in the time window corresponding to the speed mismatch oscillation mode, generating a disturbance test sequence.

[0038] The actual shift response trajectory is extracted based on the phased shift state sequence, and the disturbance test trajectory is extracted based on the disturbance test sequence. The actual shift response trajectory and the disturbance test trajectory are mapped to the corresponding dynamic safety envelope boundary in the shift dynamic safety envelope sequence according to the time window order to form the trajectory mapping result.

[0039] Based on the trajectory mapping results, the distance between the corresponding trajectory points of the real shift response trajectory and the disturbance test trajectory and the dynamic safety envelope boundary within each time window is calculated. When the trajectory point is inside the dynamic safety envelope boundary, it is marked as the envelope-in-the-boundary state. When the trajectory point exceeds the dynamic safety envelope boundary, it is marked as the envelope-crossing state. The corresponding trajectory states are output in the order of the time windows to form an envelope-crossing detection result sequence.

[0040] Optionally, generating the security control test report includes:

[0041] Based on the envelope crossing detection result sequence, the trajectory state corresponding to each time window is read, and the target time window corresponding to the envelope crossing state is extracted. When the target time window is the envelope crossing state, the shift mode evolution results corresponding to the target time window and the previous three time windows are read to form a backtracking mode window sequence.

[0042] Based on the backtracking modal window sequence, the modal amplitudes of the dominant shifting dynamics mode, execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode are extracted in each backtracking time window. The modal amplitudes of the same type of mode in each backtracking time window are accumulated to obtain the cumulative modal intensity of each type of mode.

[0043] Based on the cumulative modal intensities corresponding to various modes, the mode category with the largest cumulative modal intensity value is determined as the dominant instability mode, and the dominant instability modes corresponding to each time window are connected in order of time window to construct the risk propagation path;

[0044] Based on the dominant instability mode, risk propagation path, envelope crossing detection result sequence and corresponding shift stage, a safety control test report is generated.

[0045] The beneficial effects of this invention are:

[0046] This invention proposes a safety control testing method for a shift controller based on edge computing. It utilizes an improved symplectic geometric mode decomposition (SMD) method and the Zonotope method to dynamically model and test the multi-source operational data of the shift controller at different shift stages. Multi-source operational data is collected and the execution stages of a single shift process are segmented to construct symplectic state vectors and symplectic structured state matrices. The shift mode evolution sequence is obtained through symplectic geometric mode decomposition, and a dynamic safety envelope based on Zonotope is constructed to continuously characterize the dynamic range of changes during the shift process. A disturbance test trajectory is generated through disturbance injection, and the actual shift response trajectory and the disturbance test trajectory are mapped onto the dynamic safety envelope. Envelope crossing states are identified, and the dominant instability mode is traced back to determine the risk propagation path.

[0047] This invention reduces data transmission latency and improves the real-time performance of safety testing by completing dynamic modeling and safety envelope construction of the shifting process at edge computing nodes. Furthermore, by combining symplectic geometric mode decomposition and Zonotope dynamic envelope, it provides unified modeling of multivariate coupled changes, accurately characterizing complex dynamic behaviors such as hysteresis instability, torque shock, and speed mismatch during shifting. Moreover, it achieves early identification of potential risks through disturbance injection and envelope crossing detection, improving the accuracy and reliability of shift controller safety testing. This invention possesses advantages such as strong dynamic adaptability, high risk identification accuracy, and good testing process stability. Attached Figure Description

[0048] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0049] Figure 1 This is a flowchart of a safety control test method for a shift controller based on edge computing proposed in this invention;

[0050] Figure 2 This is a functional flowchart of the improved symplectic geometric mode decomposition of a safety control test method for a shift controller based on edge computing proposed in this invention.

[0051] Figure 3 This is a schematic diagram of the dynamic envelope module of a safety control test method for a shift controller based on edge computing proposed in this invention. Detailed Implementation

[0052] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0053] refer to Figure 1 , Figure 2 and Figure 3 A safety control test method for a shift controller based on edge computing, comprising:

[0054] The test data of the shift controller is collected and preprocessed at the edge computing node. The single shift process is segmented into stages to generate a staged shift state sequence.

[0055] Based on the phased shift state sequence, the shift dynamic state feature parameters are extracted, position-type state components and momentum-type state components are constructed, a symplectic state vector sequence is generated, and a symplectic structure state matrix is ​​constructed.

[0056] Improved symplectic geometric mode decomposition is performed at the edge computing node. The position-type state components and momentum-type state components are rewritten as symplectic state components in screw form. A screw extended symplectic basis is constructed, and the symplectic structure state matrix is ​​extended into a sparse state tensor. Tensorized symplectic decoupling is performed to generate a shift mode evolution sequence.

[0057] Based on the shift mode evolution sequence, a dynamic envelope module based on Zonotope is constructed. The envelope center bias and generation vector are constructed based on the shift mode features. The weights are adjusted in combination with the shift stage to generate a Zonotope generation vector group sensitive to the generation stage. A Zonotope safety set is constructed and a shift dynamic safety envelope sequence is generated recursively.

[0058] Perturbation injection is performed based on the phased shift state sequence and shift mode evolution sequence, and the actual shift response trajectory and the perturbation test trajectory are mapped to the shift dynamic safety envelope sequence to generate an envelope crossing detection result sequence.

[0059] Based on the envelope crossing detection result sequence backtracking modal evolution results, the dominant unstable mode is identified and the risk propagation path is determined, and a safety control test report is generated.

[0060] In this embodiment, the test data of the shift controller specifically includes input shaft speed data, output shaft speed data, transmission torque data, actuator displacement data, shift solenoid valve drive current data, clutch engagement pressure data, and shift stage indicator signal data.

[0061] In this embodiment, generating the phased shift state sequence includes:

[0062] The various data in the shift controller test data are arranged in time sequence according to a unified sampling time to form the original test data sequence. The original test data sequence is preprocessed, including time alignment, outlier removal and normalization, to generate a standard test data sequence.

[0063] The single gear shift process is segmented into stages according to the sliding time window and the shift stage identifier signal. Within each time window, the corresponding shift stage label is determined based on the changing trend of the standard test data sequence, forming a staged shift state sequence, where:

[0064] The execution of phase segmentation is as follows:

[0065] Within each time window, the changes in input shaft speed, output shaft speed, transmission torque, actuator displacement, shift solenoid valve drive current, and clutch engagement pressure are calculated. Combined with the shift stage indicator signals within the corresponding time window, the shift preparation stage is marked when the actuator displacement begins to change continuously; the torque unloading stage is marked when the transmission torque decreases and the clutch engagement pressure decreases; the gear switching stage is marked when the input shaft speed and output shaft speed change crosswise; and the torque reconstruction stage is marked when the clutch engagement pressure increases and the speed difference decreases.

[0066] In this embodiment, constructing the symplectic structure state matrix includes:

[0067] Based on the phased shift state sequence, the input shaft speed data, output shaft speed data, transmission torque data and actuator displacement data corresponding to each time window are read, and the difference between the data of two adjacent sampling times is calculated in chronological order to obtain the change in input shaft speed, change in output shaft speed, change in transmission torque and change in actuator displacement.

[0068] Position-type state components are constructed based on the changes in input shaft speed and output shaft speed, and momentum-type state components are constructed based on the changes in transmission torque and actuator displacement, wherein:

[0069] The construction of position-class state components is as follows:

[0070] The input shaft speed difference and output shaft speed difference between adjacent sampling times are calculated in chronological order and aligned in chronological order. The corresponding input shaft speed change and output shaft speed change within the same time window are combined to form a two-dimensional state vector. The two-dimensional state vectors are arranged in chronological order to form a sequence of position-type state components.

[0071] The construction of momentum-class state components is as follows:

[0072] The transmission torque difference and actuator displacement difference between adjacent sampling times are calculated in chronological order and aligned in chronological order. The corresponding transmission torque change and actuator displacement change within the same time window are combined to form a two-dimensional state vector. The two-dimensional state vectors are arranged in chronological order to form a momentum-type state component sequence.

[0073] A single-time symplectic state vector is constructed based on position-class and momentum-class state components. The single-time symplectic state vectors corresponding to each sampling time are concatenated in chronological order to generate a symplectic state vector sequence, where:

[0074] The construction of the symplectic state vector at a single time step is as follows:

[0075] At the same sampling moment, the changes in input shaft rotation speed and output shaft rotation speed in the position-type state components are taken as the first half of the data, and the changes in transmission torque and actuator displacement in the momentum-type state components are taken as the second half of the data. They are concatenated in the order of position-type state components first and momentum-type state components last to form a four-dimensional vector, which is used as the single-time symplectic state vector of the corresponding sampling moment.

[0076] Based on the symplectic state vector sequence, the symplectic state vectors corresponding to each sampling time are used as column vectors of the symplectic structure state matrix, and arranged sequentially according to the sampling time to form the symplectic structure state matrix, where:

[0077] The symplectic structure state matrix is ​​formed as follows:

[0078] Based on the symplectic state vector sequence, the symplectic state vector corresponding to each sampling time is taken as a column vector of the matrix, and the column vectors are arranged in order from early to late sampling time. The column vectors form a matrix structure in time order, forming a symplectic structure state matrix with the time dimension as the column direction and the state dimension as the row direction.

[0079] In this embodiment, generating the shift mode evolution sequence includes:

[0080] Based on the symplectic structure state matrix, an improved symplectic geometric mode decomposition is performed. Position-class and momentum-class state components corresponding to each sampling time are read and rewritten as symplectic state components in spinor form. A Lie algebra quantity representing the angular momentum rotation relationship is introduced, and heterogeneous symplectic-Lie algebra concatenation is performed. The blocks are concatenated in the order of spinor symplectic state components first, followed by the Lie algebra quantity, forming a heterogeneous symplectic-Lie state block sequence, where:

[0081] The symplectic state components, rewritten in spinor form, are as follows:

[0082] At each sampling time, the position-type state components and momentum-type state components corresponding to the symplectic structure state matrix are read. The input shaft speed change and output shaft speed change in the position-type state components are combined to form the first two-dimensional state vector. The transmission torque change and actuator displacement change in the momentum-type state components are combined to form the second two-dimensional state vector. The first two-dimensional state vector and the second two-dimensional state vector are subjected to complex number combination processing. The first two-dimensional state vector is used as the real part data and the second two-dimensional state vector is used as the imaginary part data to construct a complex form screw state vector. The screw state vectors corresponding to each sampling time are arranged in order according to the sampling time to form the symplectic state components in screw form.

[0083] The Liedai quantity characterizing the rotational relationship of angular momentum refers to the antisymmetric matrix form of data used to describe the rotational relationship of state components in phase space. The antisymmetric matrix is ​​composed of the difference relationship between state components, with zero elements on the main diagonal and the off-diagonal elements composed of the difference between position-type state components and momentum-type state components.

[0084] Perform heterogeneous symplectic-Li algebra splicing, specifically as follows:

[0085] At each sampling time, read the symplectic state components in the form of screws and the corresponding Lie numbers. Use the symplectic state components in the form of screws as the first half of the data and expand the Lie numbers into vector form as the second half of the data. Then, concatenate them in the order of symplectic state components in the form of screws first and Lie numbers last to form heterogeneous symplectic-Lie state blocks corresponding to a single sampling time. Arrange the heterogeneous symplectic-Lie state blocks corresponding to each sampling time in the order of sampling time to form a sequence of heterogeneous symplectic-Lie state blocks.

[0086] Based on the heterogeneous syn-Li state block sequence, syn-base construction processing is performed on the state blocks corresponding to each sampling time to obtain the corresponding syn-base matrix. The main diagonal structure is extracted as the syn-base principal block. A syn-fractal basis recursive structure is introduced, and the syn-base principal block corresponding to each sampling time is subjected to hierarchical expansion according to the golden ratio recursive scaling method to form multi-layer self-similar syn-base sub-blocks. Finally, a syn-fractal basis recursive sequence is constructed, where:

[0087] The octahedral construction process is performed on the state blocks corresponding to each sampling time point, specifically as follows:

[0088] The spinor form symmetric state components and Lie generation quantities in the heterogeneous symmetric-Lie state block are constructed into a square matrix structure according to the row and column order. Antisymmetric structure constraint processing is applied to the square matrix, the elements in the upper triangular region of the matrix maintain the opposite numerical relationship with the corresponding elements in the lower triangular region, and the diagonal elements are set to zero. Eigendecomposition processing is performed on the antisymmetric matrix to calculate the eigenvalues ​​and corresponding eigenvectors. The eigenvectors are arranged in descending order of eigenvalues ​​to form the symmetric matrix at the corresponding sampling time.

[0089] The symplectic-fractal recursive structure refers to a matrix construction method that generates multi-layer self-similar structures in a symplectic matrix through recursion. In the symplectic-fractal recursive structure, each layer of sub-blocks maintains the same arrangement and numerical change trend as the original symplectic main block after the matrix dimension is scaled.

[0090] The golden ratio recursive scaling method refers to the process of recursively reducing the number of rows and columns of the sigma block according to the golden ratio value. In each recursion, the number of rows and columns of the previous sigma block is multiplied by the golden ratio value and rounded to obtain the matrix dimension of the next sub-block, and the corresponding sub-matrix is ​​extracted within the reduced matrix dimension.

[0091] Hierarchical expansion is performed on the sigma block corresponding to each sampling time point, specifically as follows:

[0092] At each sampling time, the octaki master block is read, and the next layer of sub-block dimension is obtained by recursively scaling according to the golden ratio. The corresponding dimension of the sub-matrix is ​​extracted from the octaki master block as the first layer of octaki sub-block. Based on the first layer of octaki sub-block, the recursive scaling process is continued to generate the second layer of octaki sub-block and the third layer of octaki sub-block in sequence. The octaki sub-blocks of each layer are arranged along the matrix dimension according to the generation order to form a multi-layer self-similar octaki sub-block structure.

[0093] Constructing the symplectic fractal basis recursive sequence is as follows:

[0094] The symmetric sub-blocks of each layer within the same sampling time are arranged sequentially according to the hierarchical order to form a single-time symmetric fractal base sequence, and the single-time symmetric fractal base sequences corresponding to each sampling time are arranged sequentially according to the sampling time order to form a symmetric fractal base recursive sequence;

[0095] Based on the symplectic fractal basis recursive sequence, spectral domain compression is performed on the symplectic structure state matrix, mapping the symplectic state components, symplectic principal blocks, and symplectic sub-blocks corresponding to each sampling time step to a holographic symplectic spectrum matrix. The principal spectrum components are then extracted based on the state coupling relationship between adjacent sampling times to construct a sparse state tensor, where:

[0096] The spectral domain compression process is performed on the symplectic structure state matrix, specifically as follows:

[0097] The symplectic structure state matrix is ​​projected onto the symplectic fractal basis recursive sequence at the corresponding sampling time. The projection coefficients of the symplectic structure state matrix in the direction of each symplectic sub-block are calculated by matrix multiplication. The projection coefficients of each layer are arranged in hierarchical order to form a multi-layer spectral domain coefficient matrix. The spinor symplectic state components, symplectic principal blocks and each layer of symplectic sub-blocks corresponding to each sampling time are mapped to the row direction, column direction and hierarchical direction of the corresponding spectral domain coefficient matrix, respectively. The three types of data are aligned and arranged in a unified index order to form a holographic symplectic spectrum matrix.

[0098] The extraction of the main spectral components is as follows:

[0099] The holographic symplectic matrix is ​​read at each sampling time, the change of each projection coefficient between adjacent sampling times is calculated, and the absolute value of the change is sorted from large to small. The projection coefficients with the first change value are selected as the main spectrum components, and the projection coefficient sequence corresponding to each sampling time is extracted in chronological order to form the main spectrum component set.

[0100] Constructing a sparse state tensor is specifically as follows:

[0101] At each sampling time, the set of main spectrum components is read, and the main spectrum components are constructed into a three-dimensional matrix structure according to the state dimension, time dimension and hierarchical dimension. The positions that are not selected as main spectrum components are assigned zero, and the positions that are selected as main spectrum components are filled with corresponding values ​​to form a three-dimensional sparse matrix. The three-dimensional sparse matrices are arranged in the order of sampling time to form a sparse state tensor.

[0102] Tensorized symplectic decoupling is performed based on sparse state tensors. The principal state component, temporal variation component, and coupling variation component are extracted along the state dimension, time dimension, and coupling dimension, and combined to obtain the modal coefficient sequence corresponding to each sampling time. The modal coefficient sequence is then mapped to the principal spectral components in the holographic symplectic spectral matrix to obtain the principal modal components and hierarchical modal components corresponding to each sampling time, where:

[0103] The execution of Zhang's quantitative decoupling process is as follows:

[0104] In the sparse state tensor, the change sequence corresponding to each state component is extracted along the state dimension to form the main state component, the change sequence corresponding to each time window is extracted along the time dimension to form the temporal change component, and the change sequence corresponding to each level is extracted along the hierarchical dimension to form the coupled change component. The main state component, the temporal change component and the coupled change component are then spliced ​​and combined according to the sampling time order to generate the modal coefficient sequence corresponding to each sampling time.

[0105] The modal coefficient sequence is mapped to the principal spectral components in the holographic symplectic spectral matrix, specifically as follows:

[0106] At each sampling time, the modal coefficient sequence and the main spectral component in the holographic symplectic moment matrix are read. The two types of data are time-aligned according to the sampling time index. Each modal coefficient in the modal coefficient sequence is matched with the main spectral component at the corresponding sampling time. The matched modal coefficients are taken as the main modal components, and the unmatched modal coefficients are taken as the hierarchical modal components, thus forming the main modal components and hierarchical modal components corresponding to each sampling time.

[0107] Based on the modal coefficient sequence, principal modal components, and hierarchical modal components, each modal component is classified according to its modal amplitude variation, modal temporal evolution trend, and modal coupling relationship, generating a shift modal evolution sequence, wherein:

[0108] The modal components are categorized according to their modal amplitude variations, modal temporal evolution trends, and modal coupling relationships, specifically as follows:

[0109] The amplitude change of each modal component between consecutive sampling times is calculated, and the temporal evolution trend of the modality is determined based on the continuous change direction of the amplitude change. The degree of synchronization of changes between each modal component is calculated, the coupling change relationship between modes is determined, and modal components with consistent amplitude change trends and coupling change degree higher than the change degree threshold are classified into the same modal category. The modal categories are arranged in the order of sampling time to form a shift modal evolution sequence.

[0110] In this embodiment, the recursive generation of the shift dynamic safety envelope sequence includes:

[0111] A dynamic envelope module based on Zonotope is constructed. This module consists of a modal state input unit, a vector generation unit, a state envelope construction unit, a dynamic envelope evolution unit, and a safety boundary determination unit.

[0112] Constructing a dynamic envelope module based on Zonotope, specifically as follows:

[0113] The modal state input unit, the vector generation construction unit, the state envelope construction unit, the dynamic envelope evolution unit, and the safety boundary determination unit are connected in series to form a dynamic envelope module;

[0114] The modal state input unit reads the dominant shift dynamics mode, execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode corresponding to each sampling moment in the shift modal evolution sequence. It extracts the modal amplitude, modal hierarchy relationship, and modal coupling change results to form the modal state input sequence. Then, it introduces the Koopman operator to perform high-dimensional linear lifting, generating a lifted state sequence, where:

[0115] The results of extracting modal amplitude, modal hierarchy, and modal coupling changes are as follows:

[0116] At each sampling moment, the dominant shift dynamic mode, execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode corresponding to the shift mode evolution sequence are read. The modal coefficient values ​​corresponding to each mode at the current sampling moment are obtained, and the modal coefficient values ​​are used as the modal amplitudes of the corresponding modes. The modal amplitude values ​​are sorted from largest to smallest to determine that the dominant shift dynamic mode is located in the first level. The remaining modes are divided into the second and third levels according to the amplitude values, forming a modal hierarchy relationship.

[0117] The amplitude change of each mode is calculated between adjacent sampling times, and the degree of synchronization of amplitude change between different modes is calculated. When the two modes change in the same direction and have similar amplitude in consecutive sampling times, the two modes are determined to have a coupling relationship, and the modal coupling change results corresponding to each sampling time are recorded to form a modal state input sequence.

[0118] The Koopman operator is a linear transformation matrix used to map a nonlinear modal evolution process to a linear evolution process;

[0119] The execution of high-dimensional linear promotion is as follows:

[0120] At each sampling time, the modal state input sequence is read, and the modal amplitudes of the dominant shift dynamics mode, execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode are combined in time order to form a modal state vector. An extended state vector containing modal amplitude, modal amplitude squared terms, and product terms between different modes is constructed. The extended state vector is then subjected to matrix multiplication with the Koopman operator to obtain the high-dimensional state vector at the corresponding sampling time. The high-dimensional state vectors corresponding to each sampling time are arranged to form an improved state sequence.

[0121] The vector generation unit is based on the improved state sequence. It constructs the envelope center bias using the amplitude of the dominant shift dynamics mode, and generates vectors using the execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode. Weight adjustments are then performed in conjunction with the shift preparation stage, torque unloading stage, gear shifting stage, and torque reconstruction stage to generate a stage-sensitive Zonotope vector set, where:

[0122] The construction of the envelope center bias is as follows:

[0123] At each sampling time, the amplitude of the dominant shift dynamic mode corresponding to the lift state sequence is read. The amplitude of the dominant shift dynamic mode is used as the basic value, and the change in the amplitude of the dominant shift dynamic mode between adjacent sampling times is calculated. The amplitude of the dominant shift dynamic mode at the current sampling time is combined with the change to form a center bias vector. The center bias vectors corresponding to each sampling time are arranged to form an envelope center bias sequence.

[0124] The construction of the generated vector is as follows:

[0125] At each sampling time, the amplitude values ​​of the execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode corresponding to the boost state sequence are read. The amplitude values ​​of the three modes are combined to form a basic generation vector. The change in the amplitude values ​​of the three modes between adjacent sampling times is calculated. The change is concatenated with the basic generation vector to form an extended generation vector. The extended generation vectors corresponding to each sampling time are arranged to form a generation vector sequence.

[0126] The generation-sensitive Zonotope generated vector set is as follows:

[0127] At each sampling time, the corresponding shift stage label in the generated vector sequence and the phased shift state sequence is read. When the sampling time is in the shift preparation stage, the value corresponding to the execution hysteresis instability mode in the generated vector is amplified. When the sampling time is in the torque unloading stage, the value corresponding to the torque impact disturbance mode in the generated vector is amplified. When the sampling time is in the gear switching stage, the value corresponding to the speed mismatch oscillation mode in the generated vector is amplified. When the sampling time is in the torque reconstruction stage, the value corresponding to each mode in the generated vector is balanced and scaled. The adjusted generated vectors corresponding to each sampling time are arranged in the order of sampling time to form a phase-sensitive Zonotope generated vector group.

[0128] The state envelope construction unit constructs a Zonotope safe set corresponding to each sampling time based on the envelope center bias and the stage-sensitive Zonotope generation vector set. It then introduces homology theory to perform topological homology coding, generating homology loop labeling results to form the initial dynamic envelope sequence, where:

[0129] The construction of the Zonotope safe set corresponding to each sampling time is as follows:

[0130] The envelope center offset is used as the center vector, and the generated vectors in the stage-sensitive Zonotope generated vector group are arranged in column vector form to form a generation matrix. Starting from the center vector, each column of the generated vectors in the generation matrix is ​​linearly combined along the positive and negative directions, and the results of all linear combinations of the generated vectors are superimposed to obtain the envelope vertex set corresponding to each sampling time. The outer polyhedron boundary is calculated based on the envelope vertex set as the Zonotope safety set, where:

[0131] The boundary of the enclosing polyhedron is calculated based on the set of envelope vertices, specifically as follows:

[0132] At each sampling time, all vertex vectors in the envelope vertex set are read, and the coordinates of each vertex vector are extracted according to the state dimension. The maximum and minimum values ​​of the corresponding coordinate values ​​of all vertices are calculated in each dimension of the state space. The maximum and minimum values ​​of each dimension are used as the boundary range to construct the boundary interval of each dimension. The boundary intervals of each dimension are combined according to the state dimension to form a convex hull region containing all vertices. The polyhedral boundary surface is constructed according to the connection relationship between adjacent vertices in the convex hull region.

[0133] Homology theory refers to the algebraic representation method used to describe connected components, ring structures, and cavity structures in topology.

[0134] The execution of topological homology coding is as follows:

[0135] Calculate the adjacency relationship between vertices in the Zonotope safe set, construct a simple complex structure based on the adjacency relationship, count the number of connected components and the number of closed loop structures in the simple complex, combine the number of connected components and the number of closed loop structures to form a topological homology encoding vector, record the topological homology encoding vector corresponding to each sampling time, and generate homology loop marking results.

[0136] The initial dynamic envelope sequence is formed as follows:

[0137] The Zonotope safe sets at the same sampling time are associated with the corresponding homology ring labeling results and arranged sequentially to form the initial dynamic envelope sequence;

[0138] The dynamic envelope evolution unit, based on the initial dynamic envelope sequence, recursively updates the envelope center offset, generation vector change, and envelope boundary expansion between adjacent sampling times, and synchronously updates the homology loop labeling results, forming a stage-recursive safe envelope set, wherein:

[0139] The envelope center offset, generator vector change, and envelope boundary expansion between adjacent sampling times are recursively updated, specifically as follows:

[0140] At adjacent sampling times, the envelope center offset corresponding to the previous sampling time and the current sampling time is read, and the difference between the two is calculated as the envelope center offset. The generation vector group corresponding to the adjacent sampling time is read, and the difference between the generation vectors is calculated to form the generation vector change. The generation vector change is added to the generation vector group corresponding to the previous sampling time to form the updated generation vector. The envelope vertex set is recalculated based on the updated generation vector, and the distance difference between the updated envelope boundary and the envelope boundary at the previous sampling time is calculated to obtain the envelope boundary expansion. The envelope center offset, generation vector change, and envelope boundary expansion are superimposed on the envelope parameters at the previous sampling time to obtain the recursively updated envelope parameters at the current sampling time.

[0141] The safety boundary determination unit, based on the stage-recursive safety envelope set, performs state determination on the envelope boundary and homology loop marking results corresponding to each sampling time, and updates the determination results to the corresponding envelope boundary, forming a shifting dynamic safety envelope sequence, wherein:

[0142] The state determination is performed on the envelope boundary and homology loop marking results corresponding to each sampling time, specifically as follows:

[0143] At each sampling time, the corresponding envelope boundary and coherence loop marking results are read, the change in envelope boundary is calculated and compared with the change in envelope boundary at adjacent sampling times. When the change in envelope boundary increases continuously and the number of coherence loops increases, the current sampling time is determined to be an envelope expansion state. When the change in envelope boundary decreases and the number of coherence loops remains unchanged, the current sampling time is determined to be an envelope stable state. When the change in envelope boundary changes abruptly and the number of coherence loops decreases, the current sampling time is determined to be an envelope contraction state, and corresponding state marking results are generated.

[0144] The dynamic safety envelope sequence for gear shifting is formed as follows:

[0145] Read the envelope boundary, updated homology loop labeling result and state determination result corresponding to each sampling time in the sampling time sequence, and associate the three types of data in the time sequence to generate a shift dynamic safety envelope sequence.

[0146] In this embodiment, generating the envelope crossing detection result sequence includes:

[0147] Read the standard test data corresponding to each time window in the phased shift state sequence, and combine it with the dominant shift dynamic mode, execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode corresponding to each time window in the shift mode evolution sequence to determine the dominant mode type and corresponding disturbance channel for each time window, where:

[0148] The dominant mode type and corresponding disturbance channel for each time window are determined as follows:

[0149] Within each time window, read the mode amplitudes of the dominant shift dynamics mode, execution hysteresis instability mode, torque shock disturbance mode, and speed mismatch oscillation mode in the shift mode evolution sequence. Compare the amplitudes of the four types of modes according to their numerical values ​​and select the mode with the largest amplitude value as the dominant mode type.

[0150] Based on the correspondence between the dominant mode type and the standard test data, the execution hysteresis instability mode is mapped to the actuator displacement data channel, the shift solenoid valve drive current data channel and the clutch engagement pressure data channel; the torque impact disturbance mode is mapped to the transmission torque data channel; and the speed mismatch oscillation mode is mapped to the input shaft speed data channel and the output shaft speed data channel. The correspondence is recorded as the disturbance channel for each time window.

[0151] Based on the dominant mode type and disturbance channel, disturbance injection is performed. Hysteresis disturbances are injected into the actuator displacement data channel, shift solenoid valve drive current data channel, and clutch engagement pressure data channel within the time window corresponding to the hysteresis instability mode. Impact disturbances are injected into the transmission torque data channel within the time window corresponding to the torque impact disturbance mode. Offset disturbances are injected into the input shaft speed data channel and output shaft speed data channel within the time window corresponding to the speed mismatch oscillation mode, generating a disturbance test sequence, where:

[0152] The injection of hysteresis disturbances is specifically as follows:

[0153] Within the time window corresponding to the hysteresis instability mode, read the actuator displacement data, shift solenoid valve drive current data, and clutch engagement pressure data, calculate the data change between adjacent sampling times, weight and superimpose the data change at the current sampling time with the data change at the previous sampling time, and write the superposition result back to the data channel at the corresponding sampling time to form a data sequence after hysteresis disturbance.

[0154] The injection of shock disturbances is specifically as follows:

[0155] Read transmission torque data within the time window corresponding to the torque impact disturbance mode, select the position where the torque change amplitude is greater than the amplitude threshold at the corresponding sampling time, and superimpose the instantaneous amplitude increment. Gradually attenuate the amplitude increment in adjacent sampling times to form a transmission torque data sequence after the impact disturbance.

[0156] The injection of offset perturbation is specifically as follows:

[0157] Within the time window corresponding to the speed mismatch oscillation mode, the input shaft speed data and the output shaft speed data are read. A fixed offset is superimposed on the input shaft speed data and the output shaft speed data respectively, and the fixed offset is maintained within the continuous sampling time to form a speed data sequence after offset perturbation.

[0158] Generate the perturbation test sequence as follows:

[0159] Read the actuator displacement data, shift solenoid valve drive current data, and clutch engagement pressure data after injecting hysteresis disturbance, the transmission torque data after injecting impact disturbance, and the input shaft speed data and output shaft speed data after injecting offset disturbance in the order of time window, and align and arrange the data after various disturbances according to a unified time index to form a disturbance test sequence.

[0160] The actual shift response trajectory is extracted based on the phased shift state sequence, and the disturbance test trajectory is extracted based on the disturbance test sequence. The actual shift response trajectory and the disturbance test trajectory are mapped sequentially to the corresponding dynamic safety envelope boundary in the shift dynamic safety envelope sequence according to the time window, forming the trajectory mapping result, where:

[0161] The actual shift response trajectory is extracted based on a phased shift state sequence, specifically as follows:

[0162] According to the time window order, the various types of data in the standard test data sequence corresponding to the same sampling time are combined to form a state vector. The state vectors corresponding to each sampling time are arranged in time order to form a continuous state trajectory sequence, which is used as the real shift response trajectory.

[0163] The perturbation test trajectory is extracted based on the perturbation test sequence, specifically as follows:

[0164] Within each time window, the corresponding input shaft speed data, output shaft speed data, transmission torque data, actuator displacement data, shift solenoid valve drive current data, and clutch engagement pressure data in the disturbance test sequence are read. The data after various disturbances are aligned according to a unified time index, and the various disturbance data corresponding to the same sampling time are combined to form a disturbance state vector. The disturbance state vectors corresponding to each sampling time are arranged in chronological order to form a continuous disturbance state trajectory sequence, which serves as the disturbance test trajectory.

[0165] The trajectory mapping result is as follows:

[0166] Within each time window, read the state vectors of the corresponding sampling moments in the actual shift response trajectory and the disturbance test trajectory, and read the dynamic safety envelope boundary of the corresponding time window in the shift dynamic safety envelope sequence. Match the state vectors corresponding to each sampling moment in the actual shift response trajectory and the disturbance test trajectory with the dynamic safety envelope boundary of the corresponding time window, and record the correspondence between the trajectory points corresponding to each sampling moment and the dynamic safety envelope boundary in the order of the time window to form the trajectory mapping result.

[0167] Based on the trajectory mapping results, the distance between the corresponding trajectory points of the actual shift response trajectory and the disturbance test trajectory and the dynamic safety envelope boundary within each time window is calculated. When the trajectory point is inside the dynamic safety envelope boundary, it is marked as an in-envelope state; when the trajectory point exceeds the dynamic safety envelope boundary, it is marked as an envelope crossing state. The corresponding trajectory states are output in the order of the time windows to form an envelope crossing detection result sequence, where:

[0168] The distance between the corresponding trajectory points of the actual shift response trajectory and the disturbance test trajectory and the dynamic safety envelope boundary within each time window is calculated as follows:

[0169] Within each time window, read the trajectory point state vector and the set of vertices of the dynamic safety envelope boundary from the trajectory mapping results. Calculate the Euclidean distance between the trajectory point state vector and each vertex in the set of vertices of the dynamic safety envelope boundary. Calculate the differences between the trajectory point state vector and each vertex in the dimensions of input shaft speed, output shaft speed, transmission torque, and actuator displacement. Square the differences in each dimension and add them together. Then, calculate the square root of the sum to obtain the Euclidean distance between the trajectory point state vector and the corresponding vertex. Select the vertex with the smallest Euclidean distance as the distance value between the trajectory point and the dynamic safety envelope boundary. When the trajectory point is inside the dynamic safety envelope boundary, set the distance value to a positive value; when the trajectory point is outside the dynamic safety envelope boundary, set the distance value to a negative value.

[0170] The envelope crossing detection result sequence is formed as follows:

[0171] The distance values ​​corresponding to each sampling time are read sequentially according to the time window, and the corresponding trajectory status markers are generated according to the sign of the distance value. Sampling times with positive distance values ​​are marked as in-envelope states, and sampling times with negative distance values ​​are marked as envelope crossing states, thus forming an envelope crossing detection result sequence.

[0172] In this embodiment, generating the security control test report includes:

[0173] Based on the envelope crossing detection result sequence, the trajectory state corresponding to each time window is read, and the target time window corresponding to the envelope crossing state is extracted. When the target time window is the envelope crossing state, the shift mode evolution results corresponding to the target time window and the previous three time windows are read to form a backtracking mode window sequence.

[0174] Based on the backtracking modal window sequence, the modal amplitudes of the dominant shifting dynamics mode, execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode are extracted in each backtracking time window. The modal amplitudes of the same type of mode in each backtracking time window are accumulated to obtain the cumulative modal intensity of each type of mode.

[0175] Based on the cumulative modal intensities corresponding to various modes, the mode category with the largest cumulative modal intensity value is determined as the dominant instability mode, and the dominant instability modes corresponding to each time window are connected in order of time window to construct the risk propagation path;

[0176] Based on the dominant instability mode, risk propagation path, envelope crossing detection result sequence and corresponding shift stage, a safety control test report is generated.

[0177] Example 1: To verify the feasibility of this invention in practice, it was applied to the durability and safety testing of shift controllers at an automotive transmission manufacturer. Traditional testing methods mainly rely on fixed thresholds to judge parameters such as input shaft speed, output shaft speed, and transmission torque, which are prone to misjudgment or omission under complex shifting conditions. In the implementation of this invention, during bench testing of a batch of DCT-380 dual-clutch transmission shift controllers, slight shifting shocks occurred during continuous upshifting. However, because the shock amplitude did not exceed the fixed threshold range, the traditional testing system failed to identify the potential risk. Abnormal clutch wear occurred during the durability test, leading to premature termination of the test, reflecting the problem that existing technologies are unable to identify early dynamic risks.

[0178] In this test scenario, the method of this invention was deployed on an edge computing test node to collect data on input shaft speed, output shaft speed, transmission torque, actuator displacement, shift solenoid valve drive current, and clutch engagement pressure. The system segmented the single shift process into stages and constructed a shift mode evolution sequence. Simultaneously, it dynamically modeled the shift process based on the Zonotope dynamic safety envelope. In the continuous upshift test, the system analyzed the 1st-2nd gear upshift process. The test period was 0.85 seconds, the sampling frequency was 1000 Hz, and a total of 850 sets of data were collected. Within the time interval of 0.36 seconds to 0.42 seconds, the amplitude of the execution hysteresis instability mode increased from 0.18 to 0.46, while the amplitude of the torque impact disturbance mode increased from 0.12 to 0.39, and the dynamic safety envelope radius expanded from 0.52 to 0.81. By retrospective analysis, the dominant instability mode was identified as the execution hysteresis instability mode, and the risk propagation path was determined to be from the shift preparation stage to the torque unloading stage. As can be seen from the above embodiments, the method of the present invention can identify potential risks in advance under complex shift conditions, improve the accuracy and real-time performance of the safety control test of the shift controller, and verify the practical application effect and technical advantages of the present invention.

[0179] Table 1 Comparison of Safety Test Results for Shift Controllers Based on Dynamic Safety Envelope

[0180] Test number Shift type Peak torque ripple (N·m) Speed ​​mismatch amplitude (r / min) Envelope crossing situation Result of this invention Traditional method for determining results Durability test results 1 1→2 gear 18.6 132 yes risk normal minor impact 2 2 → 3 gear 11.3 87 no normal normal normal 3 3 to 4 gear 22.4 171 yes risk normal Clutch wear 4 4→5 gear 9.8 72 no normal normal normal 5 5→6 gear 23.7 182 yes risk abnormal Shift shock 6 1→2 gear 17.1 125 yes risk normal Slight shaking 7 2 → 3 gear 10.7 81 no normal normal normal 8 3 to 4 gear 21.9 165 yes risk normal Clutch wear 9 4→5 gear 12.5 93 no normal normal normal 10 5→6 gear 20.6 151 yes risk normal Shift shock

[0181] As shown in Table 1, out of 10 test samples of shift controllers, the method of this invention identified 6 units with potential risks, while the traditional method only identified 1 abnormal device. This indicates that the method of this invention has a higher risk identification capability under complex shifting conditions. From the durability test results, 5 out of 10 devices exhibited abnormalities such as minor impacts, clutch wear, or shifting shocks. All 5 of these were identified as risky states in advance by the method of this invention, while the traditional method only identified 1 device in advance. This demonstrates that the method of this invention can detect potential anomalies before failure occurs, exhibiting strong foresight and predictive capabilities.

[0182] From the overall data distribution, the devices that experienced anomalies generally exhibited high torque fluctuations and large speed mismatch amplitudes. Among the six devices with peak torque fluctuations exceeding 17 N·m, five showed varying degrees of anomalies during durability testing, indicating a strong correlation between dynamic disturbances during gear shifting and device instability. In summary, the method of this invention outperforms traditional methods in terms of risk identification accuracy, early warning capability, and testing efficiency, effectively improving the reliability and engineering application value of gear shift controller safety control testing.

[0183] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A safety control test method for a gear shift controller based on edge computing, characterized in that, include: The test data of the shift controller is collected and preprocessed at the edge computing node. The single shift process is segmented into stages to generate a staged shift state sequence. Based on the phased shift state sequence, the shift dynamic state feature parameters are extracted, position-type state components and momentum-type state components are constructed, a symplectic state vector sequence is generated, and a symplectic structure state matrix is ​​constructed. Improved symplectic geometric mode decomposition is performed at the edge computing node. The position-type state components and momentum-type state components are rewritten as symplectic state components in screw form. A screw extended symplectic basis is constructed, and the symplectic structure state matrix is ​​extended into a sparse state tensor. Tensorized symplectic decoupling is performed to generate a shift mode evolution sequence. Based on the shift mode evolution sequence, a dynamic envelope module based on Zonotope is constructed. The envelope center bias and generation vector are constructed based on the shift mode features. The weights are adjusted in combination with the shift stage to generate a Zonotope generation vector group sensitive to the generation stage. A Zonotope safety set is constructed and a shift dynamic safety envelope sequence is generated recursively. Perturbation injection is performed based on the phased shift state sequence and shift mode evolution sequence, and the actual shift response trajectory and the perturbation test trajectory are mapped to the shift dynamic safety envelope sequence to generate an envelope crossing detection result sequence. Based on the envelope crossing detection result sequence backtracking modal evolution results, the dominant unstable mode is identified and the risk propagation path is determined, and a safety control test report is generated.

2. The safety control test method for a shift controller based on edge computing according to claim 1, characterized in that, The test data for the shift controller specifically includes input shaft speed data, output shaft speed data, transmission torque data, actuator displacement data, shift solenoid valve drive current data, clutch engagement pressure data, and shift stage indicator signal data.

3. The safety control test method for a shift controller based on edge computing according to claim 1, characterized in that, The generation of the phased shift state sequence includes: The various data in the shift controller test data are arranged in time sequence according to a unified sampling time to form the original test data sequence. The original test data sequence is preprocessed, including time alignment, outlier removal and normalization, to generate a standard test data sequence. The single gear shift process is segmented into stages according to the sliding time window and the shift stage identifier signal. Within each time window, the corresponding shift stage label is determined according to the changing trend of the standard test data sequence, forming a staged shift state sequence.

4. The safety control test method for a shift controller based on edge computing according to claim 1, characterized in that, The construction of the symplectic structure state matrix includes: Based on the phased shift state sequence, the input shaft speed data, output shaft speed data, transmission torque data and actuator displacement data corresponding to each time window are read, and the difference between the data of two adjacent sampling times is calculated in chronological order to obtain the change in input shaft speed, change in output shaft speed, change in transmission torque and change in actuator displacement. Position-type state components are constructed based on the changes in input shaft speed and output shaft speed, and momentum-type state components are constructed based on the changes in transmission torque and actuator displacement. A single-time symplectic state vector is constructed based on position-type state components and momentum-type state components. The single-time symplectic state vectors corresponding to each sampling time are concatenated in chronological order to generate a symplectic state vector sequence. Based on the symplectic state vector sequence, the symplectic state vectors corresponding to each sampling time are used as column vectors of the symplectic structure state matrix, and arranged sequentially according to the sampling time to form the symplectic structure state matrix.

5. The safety control test method for a shift controller based on edge computing according to claim 1, characterized in that, The generation of the shift mode evolution sequence includes: Based on the symplectic structure state matrix, an improved symplectic geometric mode decomposition is performed. The position-class state components and momentum-class state components corresponding to each sampling time are read and rewritten as symplectic state components in spinor form. The Lie algebra quantity, which characterizes the angular momentum rotation relationship, is introduced. Heterogeneous symplectic-Lie algebra splicing is performed. The blocks are spliced ​​in the order of spinor symplectic state components first and Lie algebra quantity second to form a heterogeneous symplectic Lie state block sequence. Based on the heterogeneous syn-Li state block sequence, syn-base construction processing is performed on the state blocks corresponding to each sampling time to obtain the corresponding syn-base matrix. The main diagonal structure is extracted as the syn-base principal block. A syn-fractal basis recursive structure is introduced. According to the golden ratio recursive scaling method, the syn-base principal block corresponding to each sampling time is subjected to hierarchical expansion to form multi-layer self-similar syn-base sub-blocks, and a syn-fractal basis recursive sequence is constructed. Based on the symplectic fractal basis recursive sequence, spectral domain compression processing is performed on the symplectic structure state matrix, and the symplectic state components, symplectic principal blocks and symplectic sub-blocks corresponding to each sampling time are mapped to a holographic symplectic spectrum matrix. The principal spectrum components are extracted based on the state coupling relationship between adjacent sampling times, and a sparse state tensor is constructed. Tensorized symplectic decoupling is performed based on sparse state tensors. The main state component, temporal variation component, and coupling variation component are extracted along the state dimension, time dimension, and coupling dimension. They are then combined to obtain the modal coefficient sequence corresponding to each sampling time. The modal coefficient sequence is mapped to the main spectrum component in the holographic symplectic spectrum matrix to obtain the main modal component and hierarchical modal component corresponding to each sampling time. Based on the modal coefficient sequence, principal modal components, and hierarchical modal components, each modal component is classified according to the modal amplitude change, modal temporal evolution trend, and modal coupling change relationship to generate a shift modal evolution sequence.

6. The safety control test method for a shift controller based on edge computing according to claim 1, characterized in that, The recursive generation of the shift dynamic safety envelope sequence includes: A dynamic envelope module based on Zonotope is constructed. The dynamic envelope module consists of a modal state input unit, a vector generation construction unit, a state envelope construction unit, a dynamic envelope evolution unit, and a safety boundary determination unit. The modal state input unit reads the dominant shift dynamic mode, execution hysteresis instability mode, torque impact disturbance mode and speed mismatch oscillation mode corresponding to each sampling moment in the shift mode evolution sequence, extracts the mode amplitude, mode hierarchy relationship and mode coupling change results, forms the modal state input sequence, and introduces the Koopman operator to perform high-dimensional linear lifting to generate the lifting state sequence; The vector generation unit is based on the improved state sequence. It constructs the envelope center bias by using the amplitude of the dominant shift dynamics mode, and constructs the generation vector by using the execution hysteresis instability mode, torque impact disturbance mode and speed mismatch oscillation mode. It also performs weight adjustment in combination with the shift preparation stage, torque unloading stage, gear switching stage and torque reconstruction stage to generate a stage-sensitive Zonotope generation vector group. The state envelope construction unit constructs a Zonotope safe set corresponding to each sampling time based on the envelope center bias and the stage-sensitive Zonotope generation vector group, and introduces homology theory to perform topological homology coding processing to generate homology ring labeling results, forming the initial dynamic envelope sequence; The dynamic envelope evolution unit recursively updates the envelope center offset, the change in the generated vector, and the expansion of the envelope boundary between adjacent sampling times based on the initial dynamic envelope sequence, and synchronously updates the homology ring marking results to form a stage-recursive safe envelope set. The safety boundary determination unit performs state determination on the envelope boundary and the homology loop marking results corresponding to each sampling time based on the stage recursive safety envelope set, and updates the determination results to the corresponding envelope boundary to form a shift dynamic safety envelope sequence.

7. The safety control test method for a shift controller based on edge computing according to claim 1, characterized in that, The generated envelope crossing detection result sequence includes: Read the standard test data corresponding to each time window in the phased shift state sequence, and combine the dominant shift dynamic mode, execution hysteresis instability mode, torque impact disturbance mode and speed mismatch oscillation mode corresponding to each time window in the shift mode evolution sequence to determine the dominant mode type and corresponding disturbance channel corresponding to each time window; Based on the dominant mode type and disturbance channel, disturbance injection is performed. Hysteresis disturbances are injected into the actuator displacement data channel, shift solenoid valve drive current data channel and clutch engagement pressure data channel in the time window corresponding to the hysteresis instability mode. Impact disturbances are injected into the transmission torque data channel in the time window corresponding to the torque impact disturbance mode. Offset disturbances are injected into the input shaft speed data channel and output shaft speed data channel in the time window corresponding to the speed mismatch oscillation mode, generating a disturbance test sequence. The actual shift response trajectory is extracted based on the phased shift state sequence, and the disturbance test trajectory is extracted based on the disturbance test sequence. The actual shift response trajectory and the disturbance test trajectory are mapped to the corresponding dynamic safety envelope boundary in the shift dynamic safety envelope sequence according to the time window order to form the trajectory mapping result. Based on the trajectory mapping results, the distance between the corresponding trajectory points of the real shift response trajectory and the disturbance test trajectory and the dynamic safety envelope boundary within each time window is calculated. When the trajectory point is inside the dynamic safety envelope boundary, it is marked as the envelope-in-the-boundary state. When the trajectory point exceeds the dynamic safety envelope boundary, it is marked as the envelope-crossing state. The corresponding trajectory states are output in the order of the time windows to form an envelope-crossing detection result sequence.

8. The safety control test method for a shift controller based on edge computing according to claim 1, characterized in that, The generated security control test report includes: Based on the envelope crossing detection result sequence, the trajectory state corresponding to each time window is read, and the target time window corresponding to the envelope crossing state is extracted. When the target time window is the envelope crossing state, the shift mode evolution results corresponding to the target time window and the previous three time windows are read to form a backtracking mode window sequence. Based on the backtracking modal window sequence, the modal amplitudes of the dominant shifting dynamics mode, execution hysteresis instability mode, torque impact disturbance mode, and speed mismatch oscillation mode are extracted in each backtracking time window. The modal amplitudes of the same type of mode in each backtracking time window are accumulated to obtain the cumulative modal intensity of each type of mode. Based on the cumulative modal intensities corresponding to various modes, the mode category with the largest cumulative modal intensity value is determined as the dominant instability mode, and the dominant instability modes corresponding to each time window are connected in order of time window to construct the risk propagation path; Based on the dominant instability mode, risk propagation path, envelope crossing detection result sequence and corresponding shift stage, a safety control test report is generated.

Images