Iterative test verification method and system for engine ecu

Abstract

The present application relates to a kind of iterative test verification method and system of engine ECU, comprising the following steps, collection engine operating parameter set, condition classification is carried out and extraction condition characteristic sequence, according to this, preset virtual engine model is configured, and the virtual model after configuration is generated;Using the model, closed-loop simulation test is carried out to engine ECU, obtains simulation test response sequence, and based on the sequence, control logic verification is carried out, and the logic verification result containing amplitude margin and time margin is output;If the result does not satisfy "amplitude margin and time margin are not less than zero" passing condition, it is judged as unstable oscillation identification, and ECU control logic needs to be updated and closed-loop simulation test is re-executed until verification result satisfies steady convergence identification, solve the problem that existing virtual simulation verification method exists model and real condition adaptation, and verification lacks closed-loop iterative correction mechanism, lead to the technical problem of low verification accuracy and correction efficiency of engine ECU control logic.

Description

Technical Field

[0001] This invention relates to the field of engine technology, and in particular to an iterative testing and verification method and system for engine ECUs. Background Technology

[0002] In current virtual simulation verification of engine ECU control logic, there is a common problem of insufficient model adaptability to real operating conditions, making it difficult to accurately reproduce complex and ever-changing actual operating environments. At the same time, existing verification methods lack an effective closed-loop iterative correction mechanism, which cannot efficiently optimize the control strategy based on test feedback. This results in limited accuracy of verification results and long problem correction cycles, which seriously affect the efficiency and reliability of ECU development. Summary of the Invention

[0003] The main objective of this invention is to provide an iterative testing and verification method for engine ECUs, which solves the problems of insufficient model adaptation to real working conditions and lack of closed-loop iterative correction mechanism in existing virtual simulation verification methods, resulting in low verification accuracy and correction efficiency of engine ECU control logic.

[0004] To achieve the above objectives, the present invention provides an iterative testing and verification method for an engine ECU, comprising the following steps:

[0005] Collect the set of engine operating parameters, and classify the operating parameters according to the operating conditions to obtain the operating condition classification results;

[0006] Based on the operating condition classification results, feature extraction is performed on the set of operating parameters to obtain an operating condition feature sequence.

[0007] Based on the operating condition feature sequence, the parameters of the preset virtual engine model are configured to obtain the configured virtual model;

[0008] The engine ECU is subjected to closed-loop simulation test using the configured virtual model to obtain the simulation test response sequence. The control logic of the engine ECU is then verified based on the simulation test response sequence to obtain the logic verification result.

[0009] Based on the logic verification result, it is determined whether the control logic meets the preset pass conditions. If not, the control logic of the engine ECU is updated, and the process returns to the step of performing closed-loop simulation test on the engine ECU using the virtual model after the configuration is passed, until the logic verification result meets the pass conditions.

[0010] Furthermore, the process involves collecting a set of engine operating parameters and classifying the set of operating parameters according to their operating conditions to obtain the operating condition classification results, including:

[0011] Multi-channel synchronous sampling is performed on the physical sensors of a real engine to obtain raw waveform data. The raw waveform data is then timestamped and high-frequency noise is filtered out to obtain a set of operating parameters, which includes speed time-series curves and torque time-series curves.

[0012] Based on a preset time sliding window, the transient rate of change of the set of operating parameters is calculated to obtain the fluctuation characteristic matrix;

[0013] Based on the fluctuation feature matrix, data distribution density clustering is performed to obtain the operating condition classification result, wherein the operating condition classification result includes steady-state operating condition label and sudden change operating condition label.

[0014] Furthermore, the step of performing data distribution density clustering based on the fluctuation feature matrix to obtain the working condition classification result includes:

[0015] Calculate the Euclidean distance between each row vector in the wave feature matrix to obtain the feature distance matrix;

[0016] Based on the feature distance matrix, the number of neighboring vectors of each row vector within a preset neighborhood radius is counted to obtain the local density value;

[0017] The row vectors whose local density values ​​are greater than a preset density threshold are connected and merged to obtain a high-density cluster.

[0018] Row vectors not classified into the high-density clusters are extracted as discrete feature points;

[0019] The time sliding window corresponding to the high-density cluster is marked as the steady-state operating condition label, and the time sliding window corresponding to the discrete feature point is marked as the abrupt operating condition label. The steady-state operating condition label and the abrupt operating condition label are merged to obtain the operating condition classification result.

[0020] Furthermore, the step of extracting features from the set of operating parameters based on the operating condition classification results to obtain an operating condition feature sequence includes:

[0021] Based on the operating condition classification results, the set of operating parameters is mapped on a time axis and data is extracted to obtain segmented time series curves, wherein the segmented time series curves include multiple steady-state data segments and multiple transient data segments.

[0022] The mean and variance of each steady-state data segment are calculated to obtain a steady-state amplitude feature sequence, and the peak value and response time of each transient data segment are calculated to obtain a transient response feature sequence.

[0023] The operating condition feature sequence is obtained by splicing the steady-state amplitude feature sequence and the transient response feature sequence along the time axis.

[0024] Furthermore, the step of configuring parameters of the preset virtual engine model based on the operating condition feature sequence to obtain the configured virtual model includes:

[0025] The operating condition feature sequence is analyzed to separate steady-state feature components and transient feature components. The steady-state feature components and the transient feature components are then matched with the physical equation nodes in the preset virtual engine model to obtain a parameter mapping matrix. The parameter mapping matrix includes steady-state coefficient coordinates and transient delay coordinates. The transient delay coordinates correspond to the control pulse duty cycle dimension of the engine ECU.

[0026] The steady-state default static parameters in the physical equation nodes are extracted based on the steady-state coefficient coordinates, and the transient default static parameters in the physical equation nodes are extracted based on the transient delay coordinates.

[0027] Calculate the amplitude difference between the steady-state characteristic component and the steady-state default static parameter, and calculate the time phase difference between the transient characteristic component and the default static parameter. Then, interpolate and fit the amplitude difference and the time phase difference along the time axis to obtain a dynamic bias curve, wherein the dynamic bias curve includes a torque bias waveform and a speed hysteresis waveform.

[0028] The amplitude difference is interpolated and fitted along the steady-state coefficient coordinate to obtain the torque bias waveform; the time phase difference between the transient characteristic component and the transient default static parameter is calculated, and the time phase difference is interpolated and fitted along the transient delay coordinate to obtain the speed hysteresis waveform;

[0029] The torque bias waveform is input to the corresponding physical equation node in the preset virtual engine model according to the steady-state coefficient coordinates to obtain the nonlinear response module; the speed hysteresis waveform is input to the corresponding physical equation node in the preset virtual engine model according to the transient delay coordinates to obtain the transient hysteresis module; the nonlinear response module and the transient hysteresis module are combined to obtain the configured virtual model.

[0030] Furthermore, the closed-loop simulation test of the engine ECU using the configured virtual model to obtain the simulation test response sequence includes:

[0031] The control output terminal of the engine ECU is connected to the configured virtual model for communication, and the control pulse waveforms emitted by the engine ECU are continuously acquired.

[0032] The control pulse waveform is input to the nonlinear response module and transient hysteresis module in the configured virtual model for physical state calculation to obtain the control feedback waveform, wherein the control feedback waveform includes the fuel injection pulse width curve and the speed feedback curve.

[0033] Based on a preset sampling clock, the injection pulse width curve and the speed feedback curve in the control feedback waveform are synchronized and aligned on the time axis, and the time difference and response change between the step change point of the injection pulse width curve and the initial response point of the speed feedback curve are extracted.

[0034] The time difference and the response change are sequentially concatenated to obtain the simulation test response sequence.

[0035] Furthermore, the step of verifying the control logic of the engine ECU based on the simulation test response sequence to obtain the logic verification result includes:

[0036] Extract the time difference and response change from the simulation test response sequence, calculate the slope of the response change within the time difference, and obtain the dynamic gradient sequence.

[0037] Extreme points are extracted from the response change and a line is fitted to obtain a decay envelope curve. The maximum amplitude is extracted from the decay envelope curve to obtain the overshoot value. The time difference values ​​are accumulated in chronological order to obtain a cumulative time series. The starting position where the value is continuously less than a preset gradient threshold is found in the dynamic gradient series, and the cumulative time value corresponding to the starting position is extracted from the cumulative time series to obtain the convergence period.

[0038] The difference between the overshoot amplitude and the preset safety boundary threshold is calculated to obtain the amplitude margin, and the difference between the convergence period and the preset response cutoff time is calculated to obtain the time margin.

[0039] The magnitude margin and the time margin are numerically concatenated to obtain the logic verification result.

[0040] This invention also provides an iterative testing and verification system for an engine ECU, comprising:

[0041] The acquisition module is used to acquire a set of engine operating parameters and classify the set of operating parameters according to operating conditions to obtain the operating condition classification results.

[0042] The extraction module is used to extract features from the set of operating parameters based on the operating condition classification results to obtain an operating condition feature sequence;

[0043] The configuration module is used to configure the parameters of the preset virtual engine model based on the working condition feature sequence to obtain the configured virtual model.

[0044] The verification module is used to perform closed-loop simulation testing on the engine ECU through the configured virtual model, obtain the simulation test response sequence, and perform control logic verification on the engine ECU based on the simulation test response sequence to obtain the logic verification result.

[0045] The update module is used to determine whether the control logic meets the preset pass conditions based on the logic verification result. If it does not meet the pass conditions, the control logic of the engine ECU is updated, and the process returns to the step of performing closed-loop simulation test on the engine ECU using the virtual model after the configuration is passed, until the logic verification result meets the pass conditions.

[0046] This invention provides an iterative testing and verification method for an engine ECU, comprising the following steps: collecting a set of engine operating parameters and classifying the operating parameter set according to operating conditions to obtain operating condition classification results; extracting features from the operating parameter set based on the operating condition classification results to obtain an operating condition feature sequence; configuring parameters for a preset virtual engine model based on the operating condition feature sequence to obtain a configured virtual model; performing closed-loop simulation testing on the engine ECU using the configured virtual model to obtain a simulation test response sequence, and verifying the control logic of the engine ECU based on the simulation test response sequence to obtain a logic verification result; and performing closed-loop simulation testing on the engine ECU using the configured virtual model to obtain a simulation test response sequence, and verifying the control logic of the engine ECU based on the simulation test response sequence to obtain a logic verification result; and performing closed-loop simulation testing on the engine ECU based on the simulation test response sequence to obtain a logic verification result. The logic verification result determines whether the control logic meets the preset pass conditions. If not, the control logic of the engine ECU is updated, and the process returns to the step of performing closed-loop simulation testing on the engine ECU using the configured virtual model, until the logic verification result meets the pass conditions. This solves the technical problems of insufficient model-to-real-world operating conditions and lack of a closed-loop iterative correction mechanism in existing virtual simulation verification methods, which lead to low verification accuracy and correction efficiency of engine ECU control logic. It achieves high-fidelity operating condition-driven closed-loop iterative verification, effectively improving the accuracy of ECU control logic verification and development iteration efficiency. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1This is a schematic diagram of the steps of an iterative testing and verification method for an engine ECU in one embodiment of the present invention;

[0049] Figure 2 This is a structural block diagram of an iterative testing and verification device for an engine ECU in one embodiment of the present invention;

[0050] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0051] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0052] The following describes in detail, with reference to the accompanying drawings, an iterative testing and verification method for an engine ECU according to an embodiment of the present invention. First, the iterative testing and verification method for an engine ECU according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

[0053] like Figure 1 As shown, Figure 1 This invention provides an iterative testing and verification method for an engine ECU, comprising the following steps:

[0054] Step S1: Collect the set of engine operating parameters and classify the set of operating parameters according to operating conditions to obtain the operating condition classification result.

[0055] Specifically, in this step, multi-dimensional parameters of the engine during operation are first collected in real time using onboard sensors or a bench testing system. These parameters include, but are not limited to, engine speed, intake air volume, throttle opening, coolant temperature, and exhaust back pressure, forming a complete set of operating parameters. Then, based on preset operating condition classification rules (e.g., according to the two-dimensional mapping relationship between engine speed and load), a clustering algorithm or rule engine is used to classify the parameter set, dividing the continuous operating data into several typical operating condition intervals, such as idling, low speed and low load, and high speed and high load, thereby obtaining a structured operating condition classification result. This classification process ensures that the parameter distribution within each operating condition category is highly similar, while significant differences exist between different categories, providing clear boundaries for subsequent feature extraction. In this embodiment, the above scheme effectively improves the accuracy and stability of operating condition identification, laying a data foundation for constructing a high-fidelity virtual engine model.

[0056] Step S2: Based on the operating condition classification results, feature extraction is performed on the set of operating parameters to obtain an operating condition feature sequence, and the parameters of the preset virtual engine model are configured based on the operating condition feature sequence to obtain the configured virtual model.

[0057] Specifically, after obtaining the operating condition classification results, representative dynamic and static features are extracted for the subset of operating parameters corresponding to each operating condition. These features include, for example, the standard deviation of engine speed fluctuation, average intake pressure, stable air-fuel ratio range, and transient response delay time under that operating condition, forming a sequence of operating condition features arranged by time or operating condition order. Subsequently, this feature sequence is mapped to the corresponding configurable parameter interfaces in a preset virtual engine model, such as the friction loss coefficient, charging efficiency curve, and combustion heat release rate model. Based on the magnitude and trend of the feature values, the internal parameters of the model are dynamically assigned and calibrated, thereby generating a configured virtual model that highly matches the actual engine operating state. In this embodiment, the above scheme significantly improves the parameter adaptation accuracy of the virtual engine model under multiple operating conditions, providing a reliable model foundation for subsequent high-fidelity closed-loop simulation testing.

[0058] Step S3: Perform closed-loop simulation test on the engine ECU using the configured virtual model to obtain the simulation test response sequence, and verify the control logic of the engine ECU based on the simulation test response sequence to obtain the logic verification result.

[0059] Specifically, the configured virtual model and the ECU of the engine under test are connected to a real-time simulation platform to construct a closed-loop circuit: the ECU outputs injection pulse width, ignition advance angle, and throttle opening commands based on the state quantities such as engine speed, intake pressure, and knock signal fed back by the virtual model. The virtual model dynamically updates its internal state according to these control commands and returns new response quantities. This cyclical interaction forms a complete simulation test response sequence. Subsequently, this response sequence is compared point by point with preset control targets (such as maintaining the air-fuel ratio at 14.7±0.3 and the transient speed overshoot not exceeding 150 rpm) to determine whether the ECU output meets the timing logic and performance boundary requirements under various operating conditions, thereby generating structured logic verification results. In this embodiment, the above scheme realizes automated and full-coverage verification of ECU control logic under high-fidelity operating conditions, significantly improving the defect detection rate and verification efficiency.

[0060] Step S4: Based on the logic verification result, determine whether the control logic meets the preset pass conditions. If not, update the control logic of the engine ECU and return to the step of performing closed-loop simulation test on the engine ECU using the virtual model after configuration, until the logic verification result meets the pass conditions.

[0061] Specifically, based on the logic verification results, the system automatically compares the simulation test response sequence with preset pass conditions. These conditions include key control indicators such as idle speed fluctuation not exceeding ±30 rpm and the time for the air-fuel ratio to recover to the target value during acceleration being less than 200 ms. If any indicator exceeds the allowable range, the current control logic is deemed to have failed verification, triggering a partial update mechanism for the ECU control strategy. This includes adjusting the ignition angle correction table or gain scheduling parameters, while retaining the configured virtual model, and re-executing the closed-loop simulation test process. This iterative process continues until the verification indicators under all operating conditions meet the pass conditions. In this embodiment, the above scheme achieves automatic closed-loop optimization of the control logic, effectively shortening the ECU development cycle and ensuring that the final control strategy is robust and compliant across all operating conditions.

[0062] In a specific embodiment, the process of collecting the engine's operating parameter set and classifying the operating parameter set according to operating conditions to obtain the operating condition classification result includes:

[0063] Multi-channel synchronous sampling is performed on the physical sensors of a real engine to obtain raw waveform data. The raw waveform data is then timestamped and high-frequency noise is filtered out to obtain a set of operating parameters, which includes speed time-series curves and torque time-series curves.

[0064] Based on a preset time sliding window, the transient rate of change of the set of operating parameters is calculated to obtain the fluctuation characteristic matrix;

[0065] Based on the fluctuation feature matrix, data distribution density clustering is performed to obtain the operating condition classification result, wherein the operating condition classification result includes steady-state operating condition label and sudden change operating condition label.

[0066] Specifically, in actual operation, firstly, physical sensing units such as crankshaft position sensors, torque flanges, and intake pressure sensors deployed on the actual engine are used to acquire each signal in parallel at a sampling frequency of no less than 10 kHz using a multi-channel synchronous data acquisition card, forming raw waveform data containing the original voltage or digital pulse sequence. Subsequently, a unified timestamp is applied to the data of each channel based on a high-precision global clock source, and high-frequency electromagnetic interference and mechanical vibration noise are suppressed using wavelet threshold denoising or low-pass Butterworth filters, retaining effective dynamic information. Finally, a set of operating parameters with strict timing consistency is reconstructed, among which the core parameters include the curve of engine speed changing with time and the timing response curve of output torque.

[0067] Based on this, a time sliding window with a width of 500 ms and a step size of 50 ms is set, and the set of operating parameters is traversed along the time axis. The mean absolute value of the first difference between the rotational speed and torque within the window is calculated to characterize the degree of transient change, thereby constructing a fluctuation feature matrix in which each row is a time window and each column is a parameter change rate.

[0068] The matrix is ​​input into the DBSCAN clustering algorithm based on the density reachability principle. Using preset neighborhood radii (e.g., Euclidean distance of rate of change not exceeding 8 rpm / s or 5 N·m / s) and minimum sample size (e.g., 5 consecutive windows) as clustering hyperparameters, the algorithm automatically identifies high-density regions and sparse boundaries. High-density regions are labeled as steady-state operating conditions, corresponding to the engine operating at constant speed or constant load. Data points located in sparse regions or at cluster edges are classified as abrupt operating conditions, reflecting transient processes such as acceleration, deceleration, or load jumps. The entire process ensures that operating condition classification does not rely on human experience thresholds and can adapt to the dynamic characteristics under different driving cycles.

[0069] In this embodiment, the above-mentioned scheme realizes a fully automatic and highly robust mapping from raw sensor signals to semantic chemical condition labels, which significantly improves the accuracy and generalization ability of subsequent feature extraction and model configuration.

[0070] In a specific embodiment, the step of performing data distribution density clustering based on the fluctuation feature matrix to obtain the working condition classification result includes:

[0071] Calculate the Euclidean distance between each row vector in the wave feature matrix to obtain the feature distance matrix;

[0072] Based on the feature distance matrix, the number of neighboring vectors of each row vector within a preset neighborhood radius is counted to obtain the local density value;

[0073] The row vectors whose local density values ​​are greater than a preset density threshold are connected and merged to obtain a high-density cluster.

[0074] Row vectors not classified into the high-density clusters are extracted as discrete feature points;

[0075] The time sliding window corresponding to the high-density cluster is marked as the steady-state operating condition label, and the time sliding window corresponding to the discrete feature point is marked as the abrupt operating condition label. The steady-state operating condition label and the abrupt operating condition label are merged to obtain the operating condition classification result.

[0076] Specifically, after obtaining the fluctuation feature matrix, the Euclidean distance between any two row vectors in the matrix is ​​first calculated. Each row vector corresponds to a time sliding window, and its elements contain the quantified features of the engine speed change rate and torque change rate within that window. By calculating the Euclidean distance between all row vectors pairwise, a symmetric feature distance matrix is ​​constructed to characterize the similarity of dynamic behavior between different time windows.

[0077] Based on the aforementioned feature distance matrix, a preset neighborhood radius is set (e.g., 8.5 units, a reasonable threshold determined in the dimensionless feature space after normalizing the rate of change of rotational speed [rpm / s] and the rate of change of torque [N·m / s]). Each row vector is traversed, and the number of other row vectors it contains within that neighborhood radius is counted, serving as the local density value corresponding to that time window. This local density value reflects the degree of clustering of the dynamic state within that window in the overall data distribution.

[0078] Next, row vectors with local density values ​​greater than a preset density threshold (e.g., containing at least 5 adjacent windows) are identified as core points. Based on the density reachability principle, interconnected core points are merged to form high-density clusters. These high-density clusters represent typical steady-state behavior patterns that recur and change gradually during engine operation.

[0079] For isolated row vectors that are not included in any high-density clusters, they are extracted as discrete feature points. These points usually correspond to transient processes with abrupt changes in dynamic characteristics that are difficult to cluster, such as non-stationary conditions like rapid acceleration, rapid deceleration, or load step changes.

[0080] The time-sliding windows corresponding to high-density clusters are uniformly labeled as steady-state operating conditions, while the time-sliding windows corresponding to discrete feature points are labeled as abrupt operating conditions. The two types of labels are merged in their original time order to form a complete operating condition classification result. In this embodiment, the above scheme achieves adaptive and unsupervised partitioning of engine operating states through a rigorous density clustering mechanism, effectively distinguishing between steady-state and transient operating conditions, and providing a reliable semantic foundation for subsequent high-precision modeling and control verification.

[0081] In a specific embodiment, the step of extracting features from the set of operating parameters based on the operating condition classification results to obtain an operating condition feature sequence includes:

[0082] Based on the operating condition classification results, the set of operating parameters is mapped on a time axis and data is extracted to obtain segmented time series curves, wherein the segmented time series curves include multiple steady-state data segments and multiple transient data segments.

[0083] The mean and variance of each steady-state data segment are calculated to obtain a steady-state amplitude feature sequence, and the peak value and response time of each transient data segment are calculated to obtain a transient response feature sequence.

[0084] The operating condition feature sequence is obtained by splicing the steady-state amplitude feature sequence and the transient response feature sequence along the time axis.

[0085] Specifically, after obtaining the operating condition classification results, the results are first precisely aligned with the original set of operating parameters on the time axis. Based on the steady-state operating condition label or abrupt operating condition label corresponding to each time sliding window, the speed time series curve and torque time series curve are segmented to generate several continuous segmented time series curves. Among them, the continuous windows marked with the steady-state operating condition label are merged into a steady-state data segment, reflecting the engine's operating state under approximately constant load and speed, while the windows marked with the abrupt operating condition label are combined into a transient data segment, corresponding to dynamic processes such as acceleration, deceleration, or load switching.

[0086] For each steady-state data segment, the arithmetic mean and standard deviation of its speed and torque within that time period are calculated as steady-state amplitude features characterizing the operating point and fluctuation degree of the steady-state condition. All steady-state data segments are arranged in chronological order to form a steady-state amplitude feature sequence. At the same time, for each transient data segment, the maximum deviation (i.e., peak value) during the speed or torque change process and the time required from the start of the disturbance to reach the peak value (i.e., response time) are identified to quantify the intensity and dynamic response speed of the transient process, thereby constructing a transient response feature sequence sorted by time.

[0087] The steady-state amplitude feature sequence and the transient response feature sequence are concatenated according to their chronological order on the original time axis to ensure temporal logical consistency between the features, forming a complete operating condition feature sequence. This sequence retains the core operating point information of steady-state operation while embedding the dynamic characteristics of key transient events, providing a structured, high-information-density input basis for subsequent virtual model configuration and control logic verification. In this embodiment, the above scheme achieves efficient and accurate extraction from raw operating parameters to semantic operating condition features, significantly improving the realism and coverage of scenario reproduction in engine ECU simulation testing.

[0088] In a specific embodiment, configuring the parameters of a preset virtual engine model based on the operating condition feature sequence to obtain a configured virtual model includes:

[0089] The operating condition feature sequence is analyzed to separate steady-state feature components and transient feature components. The steady-state feature components and the transient feature components are then matched with the physical equation nodes in the preset virtual engine model to obtain a parameter mapping matrix. The parameter mapping matrix includes steady-state coefficient coordinates and transient delay coordinates. The transient delay coordinates correspond to the control pulse duty cycle dimension of the engine ECU.

[0090] The steady-state default static parameters in the physical equation nodes are extracted based on the steady-state coefficient coordinates, and the transient default static parameters in the physical equation nodes are extracted based on the transient delay coordinates.

[0091] Calculate the amplitude difference between the steady-state characteristic component and the steady-state default static parameter, and calculate the time phase difference between the transient characteristic component and the default static parameter. Then, interpolate and fit the amplitude difference and the time phase difference along the time axis to obtain a dynamic bias curve, wherein the dynamic bias curve includes a torque bias waveform and a speed hysteresis waveform.

[0092] The amplitude difference is interpolated and fitted along the steady-state coefficient coordinate to obtain the torque bias waveform; the time phase difference between the transient characteristic component and the transient default static parameter is calculated, and the time phase difference is interpolated and fitted along the transient delay coordinate to obtain the speed hysteresis waveform;

[0093] The torque bias waveform is input to the corresponding physical equation node in the preset virtual engine model according to the steady-state coefficient coordinates to obtain the nonlinear response module; the speed hysteresis waveform is input to the corresponding physical equation node in the preset virtual engine model according to the transient delay coordinates to obtain the transient hysteresis module; the nonlinear response module and the transient hysteresis module are combined to obtain the configured virtual model.

[0094] Specifically, after obtaining the operating condition feature sequence, the sequence is first subjected to structured analysis to distinguish the different dynamic characteristic components contained within it. Specifically, the mean and variance features from the steady-state data segment are extracted as steady-state feature components, which reflect the engine's operating point and fluctuation range under approximately constant speed and load conditions. Simultaneously, the peak value and response time features from the transient data segment are separated as transient feature components to characterize the intensity and speed of the system response during non-steady-state processes such as rapid acceleration, rapid deceleration, or load step changes. This analysis process strictly follows the time axis alignment principle, ensuring that each feature corresponds one-to-one with its original occurrence time, providing an accurate time series reference for subsequent model parameter mapping.

[0095] The aforementioned steady-state and transient characteristic components are respectively matched with the physical equation nodes in the preset virtual engine model through variable addressing. The virtual engine model contains multiple differential or algebraic equation nodes describing combustion efficiency, friction loss, intake flow, turbocharger response, and ECU control logic. Each node has a standardized parameter interface. By comparing the physical meaning of the characteristic components with the domain of the node variables, the steady-state characteristic components are mapped to the positions of equation parameters related to static output characteristics (such as the torque coefficient under rated operating conditions), forming steady-state coefficient coordinates. Simultaneously, the transient characteristic components are mapped to nodes related to dynamic delay characteristics (such as the intake manifold charging time constant and the ECU injection pulse duty cycle response delay), generating transient delay coordinates. This constructs a complete parameter mapping matrix, which not only establishes the topological relationship between measured characteristics and model parameters but also clarifies the spatial dimension for subsequent parameter correction.

[0096] Default static parameters corresponding to the steady-state coefficient coordinates and transient delay coordinates are extracted from the virtual engine model. The steady-state default static parameters are typically obtained through bench calibration and represent the ideal output characteristics under new engine conditions or standard environments; the transient default static parameters reflect the dynamic response capability under theoretical design. Next, the amplitude difference between the steady-state characteristic components and the corresponding steady-state default static parameters is calculated. This difference reflects the static output offset caused by factors such as aging, carbon buildup, and changes in ambient temperature and humidity during actual operation. Simultaneously, the time phase difference between the transient characteristic components and the corresponding transient default static parameters is calculated. This difference quantifies the degree of lag in control execution and physical response of the real system. All differences retain their original timestamps to ensure strict temporal consistency in subsequent fitting processes.

[0097] The amplitude difference is fitted along the steady-state coefficient coordinate using high-order spline interpolation to generate a continuous and smooth torque bias waveform. This waveform is used to dynamically compensate for the output deviation of the virtual model under different steady-state operating conditions. Simultaneously, the time phase difference is fitted along the transient delay coordinate in the same way to form a speed hysteresis waveform. This waveform accurately depicts the delay characteristics of the real engine's speed response relative to the command signal under abrupt changes in operating conditions. It is worth noting that the construction of the speed hysteresis waveform is specifically related to the ECU control pulse duty cycle dimension, ensuring that the virtual model can reproduce the adjustment rhythm and execution inertia of the actual controller under different load abrupt changes.

[0098] The torque bias waveform is injected into the nonlinear response module related to combustion torque output in the virtual engine model, replacing the original fixed static parameters and enabling it to adaptively adjust according to operating conditions. The speed hysteresis waveform is embedded into the transient hysteresis module coupled with intake and exhaust dynamics and turbine response, reconstructing the delay terms in its differential equations. By collaboratively integrating the nonlinear response module and the transient hysteresis module, a configured virtual model is finally formed. In this embodiment, the above scheme achieves refined and dynamic parameter reconstruction of the virtual engine model based on measured operating condition characteristics, significantly improving the accuracy of reproducing the nonlinear and hysteresis characteristics of the real engine in hardware-in-the-loop simulation, and providing a high-fidelity digital twin foundation for ECU control strategy verification.

[0099] In a specific embodiment, the step of performing closed-loop simulation testing on the engine ECU using the configured virtual model to obtain a simulation test response sequence includes:

[0100] The control output terminal of the engine ECU is connected to the configured virtual model for communication, and the control pulse waveforms emitted by the engine ECU are continuously acquired.

[0101] The control pulse waveform is input to the nonlinear response module and transient hysteresis module in the configured virtual model for physical state calculation to obtain the control feedback waveform, wherein the control feedback waveform includes the fuel injection pulse width curve and the speed feedback curve.

[0102] Based on a preset sampling clock, the injection pulse width curve and the speed feedback curve in the control feedback waveform are synchronized and aligned on the time axis, and the time difference and response change between the step change point of the injection pulse width curve and the initial response point of the speed feedback curve are extracted.

[0103] The time difference and the response change are sequentially concatenated to obtain the simulation test response sequence.

[0104] Specifically, after the virtual model is constructed, it is connected in a closed loop with the ECU of the engine under test for simulation testing. In practice, the control output of the engine ECU is first connected to the input port of the virtual model via a high-speed CAN or FlexRay communication interface to establish a real-time data link. The control pulse waveform emitted by the ECU is continuously acquired at a sampling frequency of no less than 10 kHz. This waveform mainly represents the pulse width modulation sequence of the injector drive signal, and its duty cycle directly reflects the torque command strength of the ECU for the current operating condition. Subsequently, the acquired control pulse waveform is synchronously input to the nonlinear response module and transient hysteresis module in the virtual model. These two modules jointly perform physical state calculations: the nonlinear response module performs static mapping of the injection pulse width based on the current torque bias waveform and outputs the theoretical injection quantity; the transient hysteresis module, combined with the speed hysteresis waveform, simulates factors such as intake charge delay, turbine inertia, and combustion propagation time to calculate the dynamic response process of the actual speed evolution over time, ultimately generating a control feedback waveform containing the injection pulse width curve and the speed feedback curve.

[0105] To ensure timing consistency in subsequent analyses, a high-precision sampling clock synchronized with the ECU's internal master clock (e.g., based on the IEEE 1588 precision time protocol) is used to align the injection pulse width curve and the speed feedback curve along their time axes. Based on this, the rising or falling edge of the injection pulse width curve caused by sudden load command changes is identified, and its starting time is marked as the trigger point of the control command. Simultaneously, the initial response point where the speed begins to deviate from its steady-state value is located on the speed feedback curve; this point is typically defined as the moment when the rate of change of speed first exceeds a preset sensitivity threshold (e.g., ±2 rpm / s). Then, the time difference between the step change point and the initial response point is calculated; this difference characterizes the total delay from receiving the control command to generating a measurable physical response. Simultaneously, the actual change in speed during this response process, i.e., the speed increment or decrement at the initial response point, is extracted to quantify the dynamic gain characteristics of the system.

[0106] All extracted time differences and response changes are serialized according to their occurrence order on the original timeline and encapsulated in a fixed-length data frame structure to form a simulation test response sequence. This sequence not only preserves the causal relationship of each control-response event but also fully records the dynamic behavior characteristics of the virtual engine under different operating conditions, including key indicators such as response speed, hysteresis, and output amplitude deviation. The entire closed-loop simulation process strictly maintains real-time constraints, ensuring that the ECU always operates in a virtual environment with the same dynamic characteristics as its actual controlled object, thereby effectively exposing potential control logic defects or parameter mismatch problems.

[0107] In this embodiment, the above-mentioned scheme realizes a refined evaluation of the control performance of the engine ECU under complex operating condition switching through a high-fidelity closed-loop simulation architecture. It significantly improves the ability of hardware-in-the-loop testing to reproduce real physical delays and nonlinear responses, and provides a high-confidence quantitative basis for the verification and optimization of ECU control strategies.

[0108] In a specific embodiment, the step of inputting the control pulse waveform into the nonlinear response module and transient hysteresis module in the configured virtual model for physical state calculation to obtain the control feedback waveform includes:

[0109] The control pulse waveform is input to the transient hysteresis module in the configured virtual model, the level transition edge of the control pulse waveform is extracted, and the speed hysteresis waveform in the transient hysteresis module is extracted. The pulse duty cycle of the control pulse waveform is calculated based on the level transition edge.

[0110] Using the pulse duty cycle as a lookup index, the corresponding delay time value is matched in the speed hysteresis waveform, and the control pulse waveform is shifted backward along the time axis according to the delay time value to obtain the hysteresis pulse sequence.

[0111] The hysteresis pulse sequence is input into the nonlinear response module in the configured virtual model, and the torque bias waveform in the nonlinear response module is extracted.

[0112] The hysteresis pulse sequence and the torque bias waveform are multiplied and added in the time domain to obtain the instantaneous torque value. The instantaneous torque value is then integrated over time to obtain the speed feedback curve.

[0113] The pulse width values ​​of the hysteresis pulse sequence are extracted as the injection pulse width curve, and the injection pulse width curve is combined with the speed feedback curve to obtain the control feedback waveform.

[0114] Specifically, the control pulse waveform output by the engine ECU is first sent to the transient hysteresis module in the configured virtual model. This module detects the edge position of the transition from low to high or high to low level in the control pulse waveform in real time and uses this as the start time of the control command taking effect. At the same time, a pre-embedded speed hysteresis waveform is called from within the transient hysteresis module. This waveform is constructed with the pulse duty cycle as the independent variable and the system response delay time as the dependent variable.

[0115] Based on the extracted level transition edges, the time width between adjacent rising and falling edges is calculated, and combined with the control cycle length, the pulse duty cycle of the current control pulse waveform is determined. Subsequently, using this pulse duty cycle as a lookup index, the corresponding delay time value is retrieved from the speed hysteresis waveform. This value reflects the time lag required for the physical system to generate a measurable speed response from receiving the control command under the current duty cycle. Based on this delay time value, the original control pulse waveform is shifted backward on the time axis to generate a hysteresis pulse sequence. This sequence accurately simulates the control execution lag caused by factors such as intake inertia, combustion delay, and turbine response in a real engine.

[0116] The hysteresis pulse sequence is input to the nonlinear response module, from which the configured torque bias waveform is read. This waveform characterizes the static torque output deviation caused by aging, environmental factors, or carbon buildup under different steady-state conditions. The amplitude of the hysteresis pulse sequence (usually normalized to 0-1) is multiplied and added point-by-point in the time domain with the torque bias waveform value at the corresponding moment to obtain the instantaneous torque value at each sampling moment. Subsequently, based on the basic principles of dynamics, the instantaneous torque value is numerically integrated along the time axis, and combined with the engine's moment of inertia parameters, the real-time speed change trend is derived, ultimately forming a continuous speed feedback curve. Simultaneously, the width value of each pulse in the hysteresis pulse sequence is directly extracted as a representation of the actual injector opening time, forming the injection pulse width curve. Finally, the injection pulse width curve and the speed feedback curve are combined on the same time base to form a complete control feedback waveform for subsequent closed-loop test analysis.

[0117] In this embodiment, the above scheme achieves high-precision reproduction of time delay and amplitude nonlinearity in the real engine control-response link by performing staged physical calculations on the control pulse waveform through the transient hysteresis module and the nonlinear response module in sequence, which significantly enhances the fidelity of the virtual model in hardware-in-the-loop simulation of the ECU's dynamic interactive behavior.

[0118] In a specific embodiment, the step of using the pulse duty cycle as a lookup index to match the corresponding delay time value in the speed hysteresis waveform includes:

[0119] Extract the discrete nodes of the speed hysteresis waveform, and perform coordinate axis projection separation on the discrete nodes of the waveform to obtain the duty cycle coordinates and the reference delay value;

[0120] The pulse duty cycle is input into the duty cycle coordinate system for numerical comparison, and the adjacent coordinate intervals that enclose the pulse duty cycle are located in the duty cycle coordinate system.

[0121] Calculate the difference between the duty cycle of the pulse and the duty cycle of the starting point in the adjacent coordinate interval, and divide the difference between the duty cycle and the span of the adjacent coordinate interval to obtain the span offset ratio.

[0122] Based on the adjacent coordinate intervals, the corresponding boundary delay values ​​are extracted from the baseline delay values. The span offset ratio is multiplied and added to the boundary delay values ​​to obtain the delay time value.

[0123] Specifically, during the matching of delay time values, the speed hysteresis waveform is first discretized, extracting a series of pre-stored waveform discrete nodes. These nodes are generated during the model configuration phase by fitting measured transient response data. Each node contains a strictly corresponding pair of values: the horizontal axis represents the pulse duty cycle, and the vertical axis represents the corresponding reference delay value at that duty cycle, in milliseconds. Subsequently, all waveform discrete nodes are projected and separated along the horizontal and vertical axes to form independent duty cycle coordinate sequences and reference delay value sequences, maintaining a one-to-one correspondence and forming a one-dimensional lookup table structure. When the pulse duty cycle calculated from the current control pulse waveform is obtained in real time, it is input into the duty cycle coordinate sequence for numerical comparison. Through binary search or linear scan, two adjacent coordinate points that can encompass the pulse duty cycle are quickly located, namely a left endpoint and a right endpoint, which together define a local coordinate interval.

[0124] Based on this, the difference between the current pulse duty cycle and the duty cycle of the left endpoint of the coordinate interval is calculated. This difference reflects the relative offset of the lookup value within the interval. Simultaneously, the span of the coordinate interval is calculated, i.e., the difference between the duty cycles of the right and left endpoints. Dividing the duty cycle difference by the span of the coordinate interval yields a span offset ratio between 0 and 1, which precisely describes the interpolation weight of the current pulse duty cycle between adjacent nodes. Next, boundary delay values ​​corresponding to the two duty cycle endpoints are extracted from the baseline delay value sequence and denoted as the left and right boundary delay values, respectively. Finally, using the span offset ratio as the weight, linear interpolation is performed on the left and right boundary delay values: specifically, the span offset ratio is multiplied by the right boundary delay value, and then added to the product of (1 minus the span offset ratio) and the left boundary delay value, resulting in a continuous and smooth delay time value. This entire process avoids the response discontinuity problem caused by step-by-step table lookup, ensuring the physical rationality and numerical stability of the delay time under different duty cycle conditions.

[0125] In this embodiment, the above scheme achieves high-precision and continuous mapping from pulse duty cycle to delay time value through coordinate projection and linear interpolation mechanism based on discrete nodes. It effectively overcomes the extrapolation distortion problem of traditional lookup table method under non-calibrated working conditions and significantly improves the accuracy of virtual engine model in reproducing ECU control delay characteristics in transient working condition simulation.

[0126] In a specific embodiment, the step of verifying the control logic of the engine ECU based on the simulation test response sequence to obtain the logic verification result includes:

[0127] Extract the time difference and response change from the simulation test response sequence, calculate the slope of the response change within the time difference, and obtain the dynamic gradient sequence.

[0128] Extreme points are extracted from the response change and a line is fitted to obtain a decay envelope curve. The maximum amplitude is extracted from the decay envelope curve to obtain the overshoot value. The time difference values ​​are accumulated in chronological order to obtain a cumulative time series. The starting position where the value is continuously less than a preset gradient threshold is found in the dynamic gradient series, and the cumulative time value corresponding to the starting position is extracted from the cumulative time series to obtain the convergence period.

[0129] The difference between the overshoot amplitude and the preset safety boundary threshold is calculated to obtain the amplitude margin, and the difference between the convergence period and the preset response cutoff time is calculated to obtain the time margin.

[0130] The magnitude margin and the time margin are numerically concatenated to obtain the logic verification result.

[0131] Specifically, after generating the simulation test response sequence, quantitative verification of the engine ECU control logic was immediately carried out. First, the time difference and corresponding response change were extracted frame by frame from the sequence. The time difference represents the delay from the issuance of the control command to the start of the speed response, while the response change reflects the actual amplitude of the speed change during the response. Based on these two variables, the ratio of the response change to the corresponding time difference, i.e., the speed change rate per unit time, was calculated to form a dynamic gradient sequence. This sequence fully characterizes the acceleration or deceleration intensity of the system during the transient process. Subsequently, extreme point detection was performed on the response change to identify all local maximum points. These extreme points were then connected using high-order splines or piecewise linear methods to fit an upper envelope curve. This curve exhibits a monotonically decreasing trend, reflecting the decay characteristics of the system's oscillating energy. The first peak value extracted from this decay envelope curve was taken as the maximum amplitude, defined as the overshoot value, used to measure the overshoot degree of the system response under ECU control.

[0132] Simultaneously, the time differences in the simulated test response sequence, arranged chronologically, are sequentially accumulated to construct a cumulative time series. This series records the total time elapsed from the start of the test to the end of the current response event. A preset gradient threshold (e.g., 5 rpm / s², determined based on the engine's typical stable operating noise level) is set in the dynamic gradient sequence. The position where the gradient value first consistently falls below this threshold and remains below it for at least three consecutive sampling points is searched, and this position is determined as the starting point of the system entering the steady-state convergence phase. The cumulative time value corresponding to this starting position is then located in the cumulative time series, which is the convergence period, used to characterize the dynamic response speed of the ECU control loop. Furthermore, the overshoot amplitude is subtracted from a preset safety boundary threshold (e.g., ±3% of the rated speed), and the resulting positive value is the amplitude margin, reflecting the safety margin of the current control strategy in avoiding excessive overshoot. Similarly, the convergence period is subtracted from a preset response cutoff time (e.g., 200 milliseconds, conforming to the typical transient response specification of automotive engine ECUs) to obtain the time margin, characterizing the system's ability to complete adjustment within a specified time limit. Finally, the magnitude margin and time margin are numerically concatenated according to a fixed data structure to form a logical verification result containing two-dimensional performance indicators, which can be used for subsequent control parameter optimization or fault diagnosis.

[0133] In this embodiment, the above scheme, by constructing a dynamic gradient sequence and decay envelope curve, and combining it with an engineering-reasonable threshold system, achieves quantitative evaluation of the engine ECU control logic in two key dimensions: overshoot suppression capability and response convergence speed. This significantly improves the objectivity and traceability of control performance verification in hardware-in-the-loop testing.

[0134] The preset pass condition is to determine whether the amplitude margin and the time margin in the logic verification result are both not less than zero. If they are, the logic verification result is determined to be a steady-state convergence indicator. If not, the logic verification result is determined to be an unstable oscillation indicator. The steady-state convergence indicator is passed, and the unstable oscillation indicator is failed.

[0135] Specifically, after obtaining the logic verification results, the system automatically extracts two numerical indicators: amplitude margin and time margin, and sequentially checks whether both are greater than or equal to zero. If the amplitude margin is not less than zero, it indicates that the overshoot amplitude has not exceeded the preset safety boundary threshold; simultaneously, if the time margin is not less than zero, it indicates that the convergence period has not exceeded the response cutoff time. When both conditions are met, the current control process is deemed to have sufficient stability and timely response, and the system marks the logic verification result as a steady-state convergence indicator, with the corresponding test conclusion being "passed." Conversely, if either the amplitude margin or the time margin is less than zero, the control performance is considered substandard, and the system generates an instability oscillation indicator, with the corresponding test conclusion being "failed." This criterion strictly follows the engineering safety margin principle to ensure that the ECU maintains controllable convergence behavior under all transient conditions.

[0136] In this embodiment, the above scheme realizes the automated and quantitative determination of the engine ECU control stability through the dual-margin joint criterion, effectively avoiding the risk of misjudgment caused by subjective experience evaluation, and significantly improving the reliability and consistency of the logic verification link in closed-loop simulation test.

[0137] The iterative testing and verification method for the engine ECU in this embodiment of the invention has been described above. The iterative testing and verification apparatus for the engine ECU in this embodiment of the invention is described below. Please refer to [link / reference]. Figure 2 One embodiment of the iterative testing and verification device for the engine ECU in this invention includes:

[0138] The acquisition module 21 is used to acquire a set of engine operating parameters and classify the set of operating parameters according to operating conditions to obtain the operating condition classification result.

[0139] Extraction module 22 is used to extract features from the set of operating parameters based on the operating condition classification results to obtain an operating condition feature sequence;

[0140] Configuration module 23 is used to configure the parameters of a preset virtual engine model based on the working condition feature sequence to obtain a configured virtual model;

[0141] The verification module 24 is used to perform closed-loop simulation testing on the engine ECU through the configured virtual model, obtain the simulation test response sequence, and perform control logic verification on the engine ECU based on the simulation test response sequence to obtain the logic verification result.

[0142] The update module 25 is used to determine whether the control logic meets the preset pass conditions based on the logic verification result. If it does not meet the pass conditions, the control logic of the engine ECU is updated, and the process returns to the step of performing closed-loop simulation test on the engine ECU using the virtual model after the configuration is passed, until the logic verification result meets the pass conditions.

[0143] In this embodiment, the specific implementation of each unit in the above device embodiment is described in the above method embodiment, and will not be repeated here.

Claims

1. An iterative testing and verification method for an engine ECU, characterized in that, Includes the following steps: Collect the set of engine operating parameters, and classify the operating parameters according to the operating conditions to obtain the operating condition classification results; Based on the operating condition classification results, feature extraction is performed on the set of operating parameters to obtain an operating condition feature sequence. Based on the operating condition feature sequence, the parameters of the preset virtual engine model are configured to obtain the configured virtual model; The engine ECU is subjected to closed-loop simulation test using the configured virtual model to obtain the simulation test response sequence. The control logic of the engine ECU is then verified based on the simulation test response sequence to obtain the logic verification result. Based on the logic verification result, it is determined whether the control logic meets the preset pass conditions. If not, the control logic of the engine ECU is updated, and the process returns to the step of performing closed-loop simulation test on the engine ECU using the virtual model after the configuration is passed, until the logic verification result meets the pass conditions.

2. The iterative testing and verification method for an engine ECU according to claim 1, characterized in that, The process involves collecting a set of engine operating parameters, classifying the set of operating parameters according to their operating conditions, and obtaining the operating condition classification results, including: Multi-channel synchronous sampling is performed on the physical sensors of a real engine to obtain raw waveform data. The raw waveform data is then timestamped and high-frequency noise is filtered out to obtain a set of operating parameters, which includes speed time-series curves and torque time-series curves. Based on a preset time sliding window, the transient rate of change of the set of operating parameters is calculated to obtain the fluctuation characteristic matrix; Based on the fluctuation feature matrix, data distribution density clustering is performed to obtain the operating condition classification result, wherein the operating condition classification result includes steady-state operating condition label and sudden change operating condition label.

3. The iterative testing and verification method for the engine ECU according to claim 2, characterized in that, The data distribution density clustering based on the fluctuation feature matrix to obtain the working condition classification result includes: Calculate the Euclidean distance between each row vector in the wave feature matrix to obtain the feature distance matrix; Based on the feature distance matrix, the number of neighboring vectors of each row vector within a preset neighborhood radius is counted to obtain the local density value; The row vectors whose local density values ​​are greater than a preset density threshold are connected and merged to obtain a high-density cluster. Row vectors not classified into the high-density clusters are extracted as discrete feature points; The time sliding window corresponding to the high-density cluster is marked as the steady-state operating condition label, and the time sliding window corresponding to the discrete feature point is marked as the abrupt operating condition label. The steady-state operating condition label and the abrupt operating condition label are merged to obtain the operating condition classification result.

4. The iterative testing and verification method for an engine ECU according to claim 1, characterized in that, The step of extracting features from the set of operating parameters based on the operating condition classification results to obtain an operating condition feature sequence includes: Based on the operating condition classification results, the set of operating parameters is mapped on a time axis and data is extracted to obtain segmented time series curves, wherein the segmented time series curves include multiple steady-state data segments and multiple transient data segments. The mean and variance of each steady-state data segment are calculated to obtain a steady-state amplitude feature sequence, and the peak value and response time of each transient data segment are calculated to obtain a transient response feature sequence. The operating condition feature sequence is obtained by splicing the steady-state amplitude feature sequence and the transient response feature sequence along the time axis.

5. The iterative testing and verification method for an engine ECU according to claim 1, characterized in that, The step of configuring parameters for a preset virtual engine model based on the operating condition feature sequence to obtain a configured virtual model includes: The operating condition feature sequence is analyzed to separate steady-state feature components and transient feature components. The steady-state feature components and the transient feature components are then matched with the physical equation nodes in the preset virtual engine model to obtain a parameter mapping matrix. The parameter mapping matrix includes steady-state coefficient coordinates and transient delay coordinates. The transient delay coordinates correspond to the control pulse duty cycle dimension of the engine ECU. The steady-state default static parameters in the physical equation nodes are extracted based on the steady-state coefficient coordinates, and the transient default static parameters in the physical equation nodes are extracted based on the transient delay coordinates. Calculate the amplitude difference between the steady-state characteristic component and the steady-state default static parameter, and calculate the time phase difference between the transient characteristic component and the default static parameter. Then, interpolate and fit the amplitude difference and the time phase difference along the time axis to obtain a dynamic bias curve, wherein the dynamic bias curve includes a torque bias waveform and a speed hysteresis waveform. The amplitude difference is interpolated and fitted along the steady-state coefficient coordinate to obtain the torque bias waveform; the time phase difference between the transient characteristic component and the transient default static parameter is calculated, and the time phase difference is interpolated and fitted along the transient delay coordinate to obtain the speed hysteresis waveform; The torque bias waveform is input to the corresponding physical equation node in the preset virtual engine model according to the steady-state coefficient coordinates to obtain the nonlinear response module; the speed hysteresis waveform is input to the corresponding physical equation node in the preset virtual engine model according to the transient delay coordinates to obtain the transient hysteresis module; the nonlinear response module and the transient hysteresis module are combined to obtain the configured virtual model.

6. The iterative testing and verification method for an engine ECU according to claim 1, characterized in that, The closed-loop simulation test of the engine ECU using the configured virtual model yields a simulation test response sequence, including: The control output terminal of the engine ECU is connected to the configured virtual model for communication, and the control pulse waveforms emitted by the engine ECU are continuously acquired. The control pulse waveform is input to the nonlinear response module and transient hysteresis module in the configured virtual model for physical state calculation to obtain the control feedback waveform, wherein the control feedback waveform includes the fuel injection pulse width curve and the speed feedback curve. Based on a preset sampling clock, the injection pulse width curve and the speed feedback curve in the control feedback waveform are synchronized and aligned on the time axis, and the time difference and response change between the step change point of the injection pulse width curve and the initial response point of the speed feedback curve are extracted. The time difference and the response change are sequentially concatenated to obtain the simulation test response sequence.

7. The iterative testing and verification method for an engine ECU according to claim 1, characterized in that, The control logic verification of the engine ECU based on the simulation test response sequence, to obtain the logic verification result, includes: Extract the time difference and response change from the simulation test response sequence, calculate the slope of the response change within the time difference, and obtain the dynamic gradient sequence. Extreme points are extracted from the response change and a line is fitted to obtain a decay envelope curve. The maximum amplitude is extracted from the decay envelope curve to obtain the overshoot value. The time difference values ​​are accumulated in chronological order to obtain a cumulative time series. The starting position where the value is continuously less than a preset gradient threshold is found in the dynamic gradient series, and the cumulative time value corresponding to the starting position is extracted from the cumulative time series to obtain the convergence period. The difference between the overshoot amplitude and the preset safety boundary threshold is calculated to obtain the amplitude margin, and the difference between the convergence period and the preset response cutoff time is calculated to obtain the time margin. The magnitude margin and the time margin are numerically concatenated to obtain the logic verification result.

8. An iterative testing and verification system for an engine ECU, characterized in that, include: The acquisition module is used to acquire a set of engine operating parameters and classify the set of operating parameters according to operating conditions to obtain the operating condition classification results. The extraction module is used to extract features from the set of operating parameters based on the operating condition classification results to obtain an operating condition feature sequence; The configuration module is used to configure the parameters of the preset virtual engine model based on the working condition feature sequence to obtain the configured virtual model. The verification module is used to perform closed-loop simulation testing on the engine ECU through the configured virtual model, obtain the simulation test response sequence, and perform control logic verification on the engine ECU based on the simulation test response sequence to obtain the logic verification result. The update module is used to determine whether the control logic meets the preset pass conditions based on the logic verification result. If it does not meet the pass conditions, the control logic of the engine ECU is updated, and the process returns to the step of performing closed-loop simulation test on the engine ECU using the virtual model after the configuration is passed, until the logic verification result meets the pass conditions.

Images