Simulation evaluation method and platform for vehicle multi-dimensional performance index extraction and adaptive fusion

Abstract

The invention relates to the technical field of vehicle engineering and computer simulation testing, and discloses a simulation evaluation method and platform for vehicle multi-dimensional performance index extraction and adaptive fusion, and the method comprises the following steps: generating a standard working condition prototype and road excitation based on a parameterized construction mode; a whole vehicle dynamic model containing sprung and unsprung mass is constructed, and time domain simulation solving is carried out; extracting multi-dimensional indexes such as maneuverability, economy, comfort and attitude control according to data solved by time domain simulation; and calculating a sensitivity jacobian matrix of the design variables to the indexes by using a complex method, calculating an adaptive weight based on sensitivity inverse proportion logic, and carrying out weighted fusion on the normalized indexes to obtain a comprehensive score. According to the method, through a self-adaptive fusion strategy of high-precision dynamic modeling and data driving, the defects that a traditional evaluation weight is high in subjectivity and low in calculation precision are overcome, and objective and accurate quantitative evaluation of the comprehensive driving performance of the vehicle is achieved.

Description

Technical Field

[0001] This invention relates to the fields of vehicle engineering and computer simulation testing technology, specifically to a simulation evaluation method and platform for extracting and adaptively fusing multi-dimensional performance indicators of vehicles. Background Technology

[0002] With the continuous deepening of the forward development process in the automotive industry, the simulation evaluation of the overall driving performance of vehicles has become a key link in chassis tuning and vehicle parameter optimization. In the early stages of vehicle development, the use of virtual prototyping technology to predict and evaluate the handling stability, driving smoothness and energy economy of the vehicle can shorten the R&D cycle and reduce testing costs.

[0003] However, traditional simulation evaluation techniques still have shortcomings in terms of model accuracy and multiphysics coupling analysis. Traditional simulation evaluation methods treat mobility, comfort, and economy as independent subsystems and solve them separately. They lack a unified vehicle dynamics model to simultaneously capture the nonlinear coupling effect of sprung mass and unsprung mass under complex road surface excitation, making it difficult to truly reflect the comprehensive performance of the vehicle under dynamic extreme conditions. At the same time, in the sensitivity analysis of the impact of design variables on performance indicators, existing technologies rely on the finite difference method. This method is extremely sensitive to the step size selection and is easily affected by both truncation error and rounding error, making it difficult to provide high-precision gradient information while ensuring computational efficiency.

[0004] Furthermore, existing technologies suffer from excessive subjectivity in the comprehensive evaluation and fusion of multi-dimensional performance indicators. Current comprehensive scoring systems employ the analytic hierarchy process (AHP) or expert scoring to assign weights to each performance indicator. This static weight allocation method, reliant on subjective experience, cannot be dynamically adjusted based on changes in design parameters or the actual physical characteristics of the vehicle system. Therefore, this invention proposes a simulation evaluation method and platform for extracting and adaptively fusing multi-dimensional vehicle performance indicators to address the shortcomings of existing technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a simulation evaluation method and platform for extracting and adaptively fusing multi-dimensional performance indicators of vehicles. This solves the problems in existing technologies, such as insufficient description of multi-physics coupling effects in whole-vehicle simulation models, numerical accuracy errors in sensitivity calculations based on finite differences, and the lack of objectivity and robustness in evaluation results due to reliance on subjective static weights in comprehensive multi-dimensional performance evaluation.

[0006] To achieve the above objectives, the present invention provides a simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of vehicles, comprising the following steps: S1. Based on the parametric construction method, corresponding standard working condition prototypes are generated for mobility evaluation, economic evaluation and comprehensive comfort evaluation, and simulation environment input data including wheel road surface roughness excitation input is output. S2. Construct a vehicle dynamics model that includes the sprung mass and unsprung mass of the vehicle, and use the input data of the simulation environment to perform time-domain simulation and solve the problem, and output the time-domain response data of the vehicle. S3. Receive the time-domain response data and extract performance indicators according to the preset evaluation dimensions. The performance indicators include at least maneuverability indicators, economy indicators, vertical comfort indicators, and attitude control indicators. S4. The extracted performance indicators are subjected to baseline normalization processing. The sensitivity Jacobian matrix of the vehicle design decision variables to each performance indicator is calculated. Based on the sensitivity Jacobian matrix, adaptive weights are calculated according to the logic that is inversely proportional to the sensitivity amplitude. The adaptive weights are then used to perform weighted summation on the normalized performance indicators to finally synthesize a scalar score of the vehicle's comprehensive driving performance.

[0007] Preferably, a simulation evaluation platform for extracting and adaptively fusing multi-dimensional performance indicators of vehicles is characterized by comprising: The working condition generation module is used to generate standard working condition prototypes based on parametric construction methods, and outputs simulation environment input data including wheel and road surface roughness excitation inputs. The vehicle dynamics module is used to construct a vehicle dynamics model, perform time-domain simulation solutions using the input data from the simulation environment, and output the vehicle's time-domain response data. The performance index calculation module is used to receive the time-domain response data and extract multi-dimensional performance indexes, including maneuverability indexes, economic indexes, vertical comfort indexes, and attitude control indexes. The adaptive fusion module is used to perform baseline normalization on the multidimensional performance indicators, calculate the sensitivity Jacobian matrix of the vehicle design decision variables to each performance indicator, allocate adaptive weights according to a strategy inversely proportional to the sensitivity amplitude, and output a comprehensive score.

[0008] This invention provides a simulation evaluation method and platform for extracting and adaptively fusing multi-dimensional performance indicators of vehicles. It has the following beneficial effects: 1. This invention employs parametric road reconstruction technology based on spatial frequency spectrum and constructs a vehicle dynamics model with fourteen degrees of freedom by combining Lagrange's mechanics principles. It describes the coupled motion of sprung mass and unsprung mass, as well as the interaction between nonlinear suspension and magic formula tire forces. This overcomes the limitation of traditional simplified models that can only simulate a single performance. It can simultaneously extract multi-dimensional indicators such as maneuverability, economy, vertical comfort, and attitude control in a single simulation, thereby improving the efficiency of simulation evaluation and the consistency between the results and the real physical environment.

[0009] 2. This invention introduces a complex step method to calculate the sensitivity Jacobian matrix. By applying a small, purely imaginary perturbation step size to the imaginary part of the design decision variables, the partial derivatives of the performance index are obtained. Compared with the traditional finite difference method, this avoids the truncation error caused by the difficulty in selecting the step size in the real domain difference operation and the rounding error caused by the cancellation of subtraction. It can improve the sensitivity calculation results without significantly increasing the computational cost, thus providing reliable gradient data support for subsequent adaptive fusion.

[0010] 3. After baseline normalization of multidimensional performance indicators, this invention automatically calculates weights based on the sensitivity Jacobian matrix, so that indicators that are too sensitive to or unstable to design variables receive smaller weights. It abandons the arbitrariness of subjective weight assignment based on expert experience in traditional evaluation methods, and automatically balances the competitive relationship between different performance indicators through a data-driven approach, ensuring that the final synthesized vehicle comprehensive driving performance scalar score has higher objectivity, scientificity, and robustness to changes in design parameters. Attached Figure Description

[0011] Figure 1 This is the overall flowchart of the present invention; Figure 2 This is a schematic diagram of the time-domain simulation solution of the present invention. Detailed Implementation

[0012] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 and Figure 2This invention provides a simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of vehicles. This method relies on a comprehensive vehicle driving performance evaluation system, which includes a working condition generation module, a vehicle dynamics module, a performance indicator calculation module, and an adaptive fusion module. The working condition generation module is used to generate standardized simulation environment input data according to preset parameters; the vehicle dynamics module is used to construct a vehicle physical model and perform time-domain simulation solutions; the performance indicator calculation module is used to process simulation data and extract specific performance indicators; and the adaptive fusion module is used to calculate indicator weights and output a comprehensive score.

[0014] First, the parameterized performance evaluation test case construction steps are executed. The test case generation module, based on a parameterized construction method, generates corresponding standard test case prototypes for mobility evaluation, economy evaluation, and comprehensive comfort evaluation. For mobility evaluation, the test case generation module constructs a mobility evaluation test case prototype, defined as a free acceleration straight road with zero initial speed, and its key parameter is the road length. Under this test case, the simulated vehicle is set to start from a standstill, accelerate continuously with maximum available acceleration until it reaches the end of the road.

[0015] For economic evaluation, the operating condition generation module constructs an economic evaluation operating condition prototype, which is defined as a driving cycle following a specific speed-time profile. In this embodiment, the China Light Vehicle Driving Cycle (CLTC-P) defined in the People's Republic of China National Standard GB / T38146.1-2019 is used as the speed strategy data source. The simulated vehicle is set to drive strictly according to the given speed strategy to evaluate the energy consumption efficiency of the vehicle under typical driving tasks.

[0016] For comfort evaluation, the working condition generation module constructs a comprehensive comfort evaluation working condition prototype. The comprehensive comfort evaluation working condition prototype is an evaluation sequence composed of multiple independent test segments, which correspond to the evaluation requirements of vertical comfort, longitudinal comfort and lateral comfort respectively.

[0017] In the first segment of the comprehensive comfort evaluation prototype, the test case generation module constructs a statistically representative uneven road surface to evaluate the vehicle's vertical vibration filtering performance. The generation of the statistically representative uneven road surface follows the ISO 8608 standard, and uses a spatial frequency spectrum model to represent the spatial power spectral density of the road surface unevenness. The calculation formula is as follows: ; in, This represents the spectral density value at a reference spatial frequency, used to describe the roughness of the road surface. Indicates spectral index; Indicates the current spatial frequency With reference space frequency The ratio of .

[0018] To utilize the spatial power spectrum of road surface roughness in time-domain simulations, the driving condition generation module discretizes the spatial power spectrum of road surface roughness, dividing it into frequency ranges. Discretized Each frequency point is calculated using the following formula: ; ; ; in, Indicates frequency interval; and These represent the minimum and maximum spatial frequencies, respectively. Indicates the number of discrete points; Indicates the first Discrete frequency points.

[0019] For each discrete frequency Its corresponding amplitude Calculate as follows: ; Based on the calculated amplitude, and using the left wheel as a reference, the road surface roughness of the left wheel is synthesized in the spatial domain using the inverse frequency spectrum method. The calculation formula is as follows: ; in, Indicates the first Road surface power spectral density values ​​at discrete frequency points; Indicates the longitudinal position of the road; The summation operator is used to represent the summation operation. At each discrete frequency point, the value increases from 1 to a value... Each result is summed up; This indicates the first result obtained from the aforementioned calculation. The amplitude of each frequency component; Indicates uniform distribution in the interval Random phase within.

[0020] Considering that the left and right wheels are not in the same position, the distance between them is the wheelbase. In the frequency domain, the power spectra of the left and right wheels satisfy a cross-correlation relationship, which is given by the coherence function. The description and calculation formula are as follows: ; in, Indicates spatial frequency Input the coherent function value for the road surface of the left and right wheels, with a value range of 0 to 1; This represents the operation of an exponential function with the natural constant e as the base. This indicates the wheelbase (width) between the left and right wheels of a vehicle. Represents spatial frequency 1.5 to the power of 1.

[0021] The operating condition generation module uses a complex spectrum linear combination method to synthesize the right wheel signal, first constructing the complex spectrum components of the left wheel. and independent auxiliary complex spectral components The calculation formula is as follows: ; ; in, The imaginary unit; For the left wheel random phase; For separately generated independent random phases; This indicates exponentiation.

[0022] Right-hand complex spectrum components The weighted superposition of the revolver complex spectrum components and the auxiliary complex spectrum components yields the following calculation formula: ; in, Indicates the first The coherence function values ​​at discrete frequency points; The weighting coefficients for incoherent components are calculated by subtracting the square root of the coherence function value from 1. Right wheel spatial domain unevenness The synthesis is performed by taking the real part of the right-hand complex spectrum component, and the calculation formula is as follows: ; in, It is the random initial phase of the right wheel.

[0023] Considering the spatial lag between the road surface unevenness of the front and rear wheels, the working condition generation module obtains the road surface unevenness of the rear wheels based on the generated unevenness of the left and right front wheels, and then uses the road surface unevenness of the rear wheels as a basis for calculation. For example, the calculation formula is as follows: ; in, Indicates the vehicle's wheelbase; The phase corresponding to the front wheel (left or right); Indicates the current vertical position Subtract vehicle wheelbase This is used to represent the time and space lag relationship between the front and rear wheels when they pass the same point.

[0024] Through the above steps, the working condition generation module generates road surface unevenness excitation inputs at the four wheels, and the simulated vehicle is set to operate at a preset constant speed. Passing through this section of road.

[0025] In the second segment of the comprehensive comfort evaluation prototype, the condition generation module sets up an absolutely flat road surface and defines parameterized braking-acceleration composite events on this road surface to evaluate longitudinal comfort. In the braking-acceleration composite events, the vehicle first travels steadily at an initial speed. Then, a target deceleration is applied to slow the vehicle down to a preset intermediate speed. Immediately afterwards, a target acceleration is applied to accelerate the vehicle back to its initial speed. This process is used to stimulate the vehicle's pitch attitude.

[0026] In the third segment of the comprehensive comfort evaluation prototype, the scenario generation module sets up an absolutely flat road surface and defines a parameterized two-lane change maneuver on this surface to evaluate lateral comfort. The road centerline trajectory of this maneuver conforms to the ISO 3888-2 standard in terms of geometric parameters. The simulated vehicle is set to enter at a constant speed. Enter and complete the two-lane change maneuver. This process is used to stimulate the vehicle's lateral tilt.

[0027] Reference Figure 1 and Figure 2 The vehicle dynamics module constructs a 14-DOF vehicle dynamics model by defining generalized coordinates, establishing energy functions, and solving the Lagrange equations. The module first defines a generalized coordinate system to describe the instantaneous motion of the vehicle system. To capture the translational and rotational attitudes of the vehicle body and the independent motion characteristics of the four wheels under complex dynamic conditions, this embodiment selects 14 independent generalized coordinates to form a state vector. State vector The specific form is defined as follows: ; in: , , These three degrees of freedom represent the longitudinal, lateral, and vertical displacements of the center of mass of the sprung mass (i.e., the vehicle body) in the inertial coordinate system (global coordinate system), respectively. They are used to describe the overall spatial translational position of the vehicle body. , , These represent the vehicle body's position around its own coordinate system. axis, shaft and The roll, pitch, and yaw angles of the rotation axis are three degrees of freedom used to describe the three-dimensional attitude changes of the vehicle body in space. , , These represent the vertical displacements of the four unsprung masses (i.e., the wheels and their fixed suspension components) relative to the inertial coordinate system, namely the left front, right front, left rear, and right rear. These four degrees of freedom are independent of the vehicle body motion and are used to describe the bouncing response of the wheels under the excitation of uneven road surfaces. They are key variables for evaluating vertical comfort and tire contact performance. , , , These represent the rotation angles of the four wheels—left front, right front, left rear, and right rear—around their rotation axes. These four degrees of freedom are used to calculate the wheel's angular velocity and tire longitudinal slip ratio, directly affecting the calculation results of driving force and braking force, and thus affecting the evaluation of maneuverability and economy. This represents the transpose operation of a matrix or vector, converting a row vector into a column vector.

[0028] The vehicle dynamics module, based on the generalized coordinate system defined above, further constructs the vehicle's local coordinate system and transformation matrix. The module internally includes a pre-defined rotation transformation matrix from the inertial coordinate system to the vehicle body coordinate system, which is derived from Euler angles. , , Constructed according to a predetermined rotation sequence (e.g., ZYX sequence), the vehicle dynamics module can project external forces acting on the vehicle body (such as air resistance) to the local coordinate system through this transformation matrix, or project the suspension connection point velocities and accelerations in the local coordinate system of the vehicle body back to the inertial coordinate system, thereby ensuring that the motion coupling relationship of the multibody system is handled within a unified mathematical framework.

[0029] Furthermore, for each of the four unsprung masses, the vehicle dynamics module establishes an independent vertical motion coordinate system. The vertical motion of each unsprung mass is linked to the corresponding corner point of the vehicle body through suspension geometric constraints. The vehicle dynamics module adjusts the motion based on the instantaneous displacement of the vehicle body. and posture Based on the vehicle's geometric parameters (such as front / rear track and front / rear wheelbase), the vertical position and velocity of the four corner points of the vehicle body (i.e., the suspension contact points) in the inertial coordinate system are calculated in real time. This position and velocity information, along with the corresponding unsprung mass vertical displacement... and speed The difference constitutes the compression amount and compression rate of the suspension spring and shock absorber, providing kinematic input for subsequent suspension force calculations.

[0030] For the rotational degrees of freedom of the wheel The vehicle dynamics module correlates this with the vehicle's longitudinal motion. This is achieved by considering the rotational degrees of freedom of the wheels. Differentiating the wheel's angular velocity yields the angular velocity. And combined with the wheel rolling radius The longitudinal slip ratio of the tire is calculated in real time based on the longitudinal velocity of the wheel center. This slip ratio is a core input parameter of tire dynamics models (such as the magic formula), directly determining the magnitude of the longitudinal force generated by the tire, thereby achieving two-way coupling between wheel rotational dynamics and vehicle translational dynamics.

[0031] Reference Figure 1 and Figure 2 After determining the system's degrees of freedom and generalized coordinate vectors, the vehicle dynamics module uses the second kind of Lagrange equations from analytical mechanics to derive a set of differential equations describing the vehicle's motion. This method establishes the system's dynamic characteristics from an energy perspective, avoiding the tedious process of separating rigid bodies and analyzing complex internal constraints required in the Newton-Euler method. This allows for efficient handling of the multi-degree-of-freedom coupling characteristics of the vehicle system. Specifically, the vehicle dynamics module first constructs the system's Lagrangian function. The Lagrangian function is the core scalar function describing the system's energy state, defined as the difference between the system's total kinetic energy and total potential energy. The calculation formula is as follows: ; in, The Lagrangian function representing the system is a generalized coordinate system. and generalized speed The function; This represents the total kinetic energy of the entire vehicle system, which consists of two parts: one part is the translational kinetic energy and rotational kinetic energy of the sprung mass (vehicle body) about its center of mass; the other part is the vertical kinetic energy of the four unsprung masses (wheels and suspension components) and the rotational kinetic energy of the wheels about their axes of rotation. The total kinetic energy is expressed in a generalized coordinate system. and generalized speed The function; It represents the total potential energy of the entire vehicle system.

[0032] Based on the Lagrangian function mentioned above, the vehicle dynamics module establishes dynamic equations that include dissipative forces and non-conservative forces, and the calculation formulas are as follows: ; in, This indicates calculating the physical quantity within the parentheses with respect to time. The total derivative; Indicates the sign of partial derivatives; This represents the generalized velocity vector, i.e., the generalized coordinate vector. The first derivative with respect to time includes the vehicle's translational velocity, angular velocity, suspension vertical velocity, and wheel speed; It represents the partial derivative of the Lagrange function with respect to the generalized velocity, and physically corresponds to the generalized momentum of the system; It represents the partial derivative of the Lagrange function with respect to generalized coordinates, which physically corresponds to the generalized force caused by the potential field (such as gravity); The vector represents the generalized force acting on the system. To facilitate numerical solution by computer, the vehicle dynamics module expands and rearranges the above differential equations into a standard matrix form. By expanding the differential terms on the left side, the terms containing generalized acceleration are separated, yielding the equation relating the generalized force acting on the system to the current state of motion. The calculation formula is as follows: ; in, The generalized mass matrix of the system has dimensions of . This matrix is ​​a generalized coordinate system. The function includes the vehicle mass, moment of inertia, and coupled inertia terms caused by coordinate system rotation transformation; This represents the generalized acceleration vector, i.e., the generalized coordinate vector. The second derivative with respect to time; This represents the nonlinear force vector of the system, with dimension . This vector contains all dynamic terms that do not include generalized acceleration, primarily the Coriolis force, centrifugal force, and terms derived from potential energy. The generalized gravitational force term.

[0033] Finally, in order to perform time-domain simulation, the vehicle dynamics module needs to calculate the system state at the next moment. By transforming the above matrix equations, the generalized acceleration is explicitly expressed, and the vehicle dynamics solution equations are obtained. The calculation formulas are as follows: ; in, Representing the generalized mass matrix In practical calculations, matrix factorization is used instead of direct inversion to improve computational efficiency and numerical stability.

[0034] By transforming the matrix equations, the vehicle dynamics module utilizes the current state. and external input generalized force vector Calculate the instantaneous acceleration Then, the complete dynamic response of the vehicle over time is obtained through numerical integration recursion.

[0035] Reference Figure 1 and Figure 2 After constructing the left-hand inertial term of the dynamic equations, the vehicle dynamics module constructs the generalized force vector on the right-hand side of the equations by calculating and mapping external forces. Furthermore, a numerical integration algorithm is used to advance the system's state in the time domain, and the generalized force vector is obtained. The construction of the system serves as a link between external physical excitations and the vehicle's internal dynamic response. The vehicle dynamics module calculates the interaction forces between the tires and the road surface, the elastic damping forces of the suspension system, and aerodynamic loads in real time, and projects these physical forces onto the aforementioned 14 generalized coordinate directions to synthesize the final generalized force vector. .

[0036] In terms of tire force calculation, the vehicle dynamics module uses the Magic Formula semi-empirical tire model to calculate the dynamic forces at the tire-road contact patch. Specifically, the vehicle dynamics module first calculates the longitudinal slip ratio of each wheel based on the current vehicle speed, yaw rate, and wheel rotational angular velocity. and lateral slip angle Simultaneously, based on the vertical displacement of the unsprung mass of the wheel at the current moment... (in represent , , , ) and road surface unevenness input The difference, combined with the tire's vertical stiffness, is used to calculate the instantaneous vertical load on the tire. .Will , and Substituting these parameters into the magic formula, the longitudinal force in the tire coordinate system is calculated. Lateral force and the restoring torque Subsequently, the vehicle dynamics module uses a coordinate transformation matrix to convert these tire forces from the local coordinate system to the vehicle coordinate system, and then distributes them to the corresponding generalized coordinates according to the principle of virtual work. For example, the longitudinal tire force affects not only the longitudinal displacement of the vehicle body but also... It contributes to the generalized force and also affects the wheel rotation angle. It generates a counter-torque contribution, thereby achieving the coupling of wheel rotational dynamics and vehicle translational dynamics.

[0037] In terms of suspension force calculation, the vehicle dynamics module calculates the suspension force based on the relative motion between the sprung mass (body) and the unsprung mass (wheel assembly). The vehicle dynamics module first utilizes generalized coordinates. , , Calculate the vertical displacement and velocity of the vehicle body at the four suspension connection points, and compare them with the corresponding unsprung mass vertical displacement. and speed By performing differential calculations, the dynamic travel and velocity of the suspension are obtained. Based on the nonlinear spring stiffness characteristic curve and the damper damping characteristic curve (FV curve), the vertical force generated by the suspension at each corner is calculated. This suspension force, as an internal force, acts on the generalized force components of the sprung mass in equal magnitude and opposite direction (affecting...). , , (acceleration) and generalized force components of unsprung mass (affecting) (acceleration).

[0038] In terms of equation solving and time-domain propagation, due to the established dynamic equations It is a set of second-order ordinary differential equations. Before numerical solution, the vehicle dynamics module reduces it to a first-order state-space equation system. The vehicle dynamics module defines a state vector with a dimension of 28. It contains 14 generalized coordinates and 14 generalized velocities: ; the corresponding state derivative vector for: ; The vehicle dynamics module uses a fourth-order Runge-Kutta numerical integration algorithm to discretize and solve the above state equations, setting the simulation step size to be... The vehicle dynamics module utilizes the current moment state vector Through the calculation of the intermediate slopes in four steps (corresponding to the fourth-order Runge-Kutta numerical integration algorithm respectively), , , , (coefficients), weighted to estimate the next moment. state vector This process iterates until the set simulation end time is reached, thereby generating a complete time-domain response data sequence containing the vehicle's displacement, velocity, and acceleration for all degrees of freedom. This time-domain data forms the basis for subsequent performance index extraction.

[0039] Reference Figure 1 and Figure 2 The performance index calculation module receives the time-domain simulation results from the vehicle dynamics module and extracts the corresponding quantitative indicators according to different evaluation dimensions. For the maneuverability evaluation, the performance index calculation module calculates the maneuverability index that reflects the longitudinal dynamic performance of the vehicle by analyzing the vehicle's motion trajectory data under extreme acceleration conditions.

[0040] Specifically, the performance index calculation module obtains the longitudinal displacement of the vehicle's center of gravity output by the vehicle dynamics module under the mobility evaluation condition (i.e., the zero initial speed free acceleration condition). Over time The changing data sequence, under this operating condition, the vehicle dynamics model simulates the dynamic process of the driver requesting torque at maximum accelerator pedal opening, the vehicle powertrain outputting ultimate driving force, and the tires being at the edge of their adhesion limit. The performance index calculation module is set to a preset road end position. The performance index calculation module iterates through the displacement sequence output by the simulation and identifies those that meet the conditions. Minimum time index Since the vehicle was stationary at the start of the simulation and The performance index calculation module directly defines the termination time as the mobility evaluation index. Mobility indicators The calculation is as follows: ; in, This indicates the time when the vehicle reaches the end of the road during the simulation. Indicates at time The absolute longitudinal displacement of the vehicle's center of gravity in the global coordinate system; This represents the target value for the total length of the road under the preset mobility evaluation conditions; : Indicates the value retrieval operation under specific conditions.

[0041] Reference Figure 1 and Figure 2 The performance index calculation module analyzes the powertrain system state data of the vehicle under the economic evaluation condition (i.e., the CLTC-P driving cycle) to calculate the economic index characterizing the vehicle's energy utilization efficiency. In this embodiment, this index is quantified as the total energy consumption of the electric drive system throughout the entire operating cycle. For the simulated vehicle with a dual-motor all-wheel drive architecture, the performance index calculation module first obtains the real-time operating state parameters of the front and rear drive motors at each simulation time step from the vehicle dynamics module. Specifically, it obtains the real-time required torque of the front drive motor. and rotational speed And the real-time torque demand of the rear drive motor. and rotational speed .

[0042] Based on the above operating parameters, the performance index calculation module calculates the instantaneous mechanical power at the output shaft ends of the front and rear motors. Front motor mechanical power and the mechanical power of the motor The calculation formula is as follows: ; ; in: This indicates the instantaneous mechanical power of the front motor output shaft, measured in kilowatts (kW). This indicates the real-time electromagnetic torque demand of the front motor, measured in Newton-meters (Nm). Positive values ​​indicate the driving state, while negative values ​​indicate the braking (energy recovery) state. : Indicates the real-time speed of the front motor, in revolutions per minute (rpm); 9550: Indicates the physical conversion constant for converting the product of torque and speed into kilowatts of power; This indicates the instantaneous mechanical power of the rear motor output shaft, expressed in kilowatts (kW). This indicates the real-time electromagnetic torque demand of the rear motor, expressed in Newton-meters (Nm). This indicates the real-time speed of the rear motor, in revolutions per minute (rpm).

[0043] Subsequently, the performance index calculation module uses the pre-stored motor system efficiency MAP model to query the real-time operating efficiency of the front and rear motors based on the current torque and speed. and Based on mechanical power and operating efficiency, the module calculates the real-time electrical power consumption of the motor at the DC bus. During this process, the module needs to distinguish between drive mode and energy recovery mode. For the front motor, its real-time electrical power... The calculation formula is as follows: ; For the rear motor, its real-time electrical power The calculation formula is as follows: ; in, This indicates the real-time electrical power at the DC input terminal of the front motor controller, in kilowatts (kW). A positive value indicates energy consumption from the battery, while a negative value indicates energy being fed back to the battery. This represents a piecewise function, with different calculation logic selected based on the sign of the mechanical power. This indicates that the front motor is in driving mode. At this time, the electrical power must be greater than the mechanical power (divided by the efficiency), reflecting the loss. This indicates that the front motor is in a regenerative braking state. At this time, the absolute value of the feedback electrical power is less than the absolute value of the mechanical power (multiplied by the efficiency), reflecting the loss. This represents the real-time electrical power at the DC input terminal of the rear motor controller, in kilowatts (kW). Finally, the performance index calculation module algebraically sums the real-time electrical power of all drive motor units in the system to obtain the instantaneous total power demand of the entire vehicle, and then performs a definite integral over time to calculate the total energy consumption for completing this driving cycle. The calculation formula is as follows: ; in: This refers to the economic performance indicator, which is the vehicle's total energy consumption under the entire evaluation condition. The unit is usually kilowatt-hour (kWh) or joule (J), depending on the time. Units; This represents a definite integral operation, with the integration interval from the simulation start time 0 to the simulation end time. .

[0044] Through this step, the performance index calculation module accurately quantifies the energy flow process of the vehicle under standard operating conditions and integrates the impact of motor efficiency characteristics and energy recovery strategies on the overall vehicle energy consumption. In this single indicator.

[0045] Reference Figure 1 and Figure 2 The performance index calculation module processes the vehicle's time-domain response data under a comprehensive comfort evaluation working condition sequence (including vertical uneven road surface, longitudinal acceleration and deceleration, and lateral lane change segments) and extracts multi-dimensional comfort indicators that characterize human vibration perception and vehicle posture control stability.

[0046] For vertical comfort evaluation, the performance index calculation module first acquires the time-domain signals of vertical acceleration at the driver's seat interface or the vehicle's center of gravity when the vehicle traverses randomly uneven road segments. To simulate the human body's varying sensitivity to vibrations at different frequencies, the performance index calculation module first applies a vertical frequency weighting filter conforming to the GB / T4970 standard (or ISO2631 standard) to the original vertical acceleration signal. This filter maintains or amplifies the gain in the 4Hz-8Hz frequency band, which is sensitive to human sensitivity, while attenuating the less sensitive frequency band. The filtered output signal is denoted as... .

[0047] Based on frequency-weighted acceleration signal The performance index calculation module first calculates the frequency-weighted root mean square value. This is used to evaluate the average energy level of vibrations. The calculation formula is as follows: ; in, The frequency-weighted root mean square acceleration index represents vertical comfort, measured in meters per second squared (m² / s²). ); This represents the coefficient used to normalize the total time and is used to calculate the average value. This represents the total duration of the simulation for the vertical comfort test section, expressed in seconds (s).

[0048] To further capture the impact of transient peak vibrations caused by road impacts on comfort (RMS values ​​are not sensitive to occasional impacts), the performance index calculation module further calculates the vibration dose value. Vibration dose value The calculation formula using the fourth power accumulation method is as follows: ; in, This indicates the vertical vibration dose value, expressed in meters per 1.75 power seconds (m² / s). ); This indicates that the fourth square root operation is performed on the integral result; Indicates time The fourth power of the absolute value of the frequency-weighted acceleration amplifies the proportion of large-amplitude signals in the integral result, thus making the indicator more reflective of the discomfort of impact vibration.

[0049] For longitudinal and lateral comfort evaluation, the performance index calculation module focuses on the vehicle's attitude stability during maneuvering, and analyzes the vehicle's pitch rate response under braking and acceleration conditions. And the roll rate response under two-lane change conditions. Excessive angular velocity can cause vestibular discomfort (such as motion sickness) and a sense of unease in passengers. The performance index calculation module extracts the root mean square value of the vehicle's pitch angular velocity. The calculation formula is as follows: ; in: This is an index representing the root mean square value of pitch angular velocity, with units of radians per second (rad / s). Indicates time The angular velocity of the vehicle's rotation about its horizontal axis (i.e., the pitch angle) (time derivative) This represents the square of the pitch rate; simultaneously, the performance index calculation module extracts the root mean square value of the vehicle roll rate. The calculation formula is as follows: ; in, This is an index representing the root mean square value of the roll rate, with units of radians per second (rad / s). This represents the angular velocity (i.e., roll angle) of the vehicle about its longitudinal axis at time t. (time derivative) : Represents the square of the roll rate.

[0050] Finally, the performance metric calculation module summarizes all the individual metrics calculated above and outputs a set of performance metrics containing 6 dimensions. This is used for subsequent adaptive fusion. .

[0051] Reference Figure 1 and Figure 2 The adaptive fusion module receives the multi-dimensional index set output by the performance index calculation module. Next, baseline normalization is performed. This step aims to address the heterogeneity of physical dimensions and differences in numerical magnitude among different performance indicators, providing a unified mathematical basis for subsequent weighted fusion.

[0052] Specifically, the adaptive fusion module identifies the original set of input metrics. It includes data items with different physical units: mobility index The unit is seconds, and the economic indicators are... The unit is kilowatt-hour or joule, vertical comfort index The unit is meters per second squared, while the attitude control index The unit is radians per second. If these values ​​are directly weighted and summed, the larger absolute magnitude of the indicators will mask the changes of the smaller absolute magnitude of the indicators, resulting in distorted evaluation results.

[0053] To eliminate the influence of these dimensions and magnitudes, the adaptive fusion module introduces the concept of a baseline vehicle model. The baseline vehicle model represents the initial state of the vehicle design, the parameter configuration of the previous generation model or a benchmark competitor model. The system pre-simulates the same operating conditions and dynamics to obtain the reference index set corresponding to the baseline vehicle model. And store it in the system database.

[0054] During real-time evaluation, the adaptive fusion module employs a ratio normalization method to normalize each performance index of the current design scheme. Mapped to dimensionless relative values The calculation formula is as follows: ; in: Indicates the first The dimensionless values ​​of the performance indicators after normalization represent the degree of performance deviation of the current design from the baseline design. This indicates that the current design's value for this indicator is lower than the baseline; if This indicates that the value is greater than the baseline; if This indicates that it is in line with the baseline; This indicates the number of simulation results obtained at the current time. The values ​​of the original physical performance indicators; This indicates that the value retrieved from the database corresponds to the first... The baseline reference value for each indicator.

[0055] Through this process, the adaptive fusion module uniformly transforms all performance metrics into a dimensionless scalar space that fluctuates around the value 1.0. This eliminates the limitations of physical units and makes the metrics numerically comparable, thereby ensuring that the subsequent sensitivity-based weight allocation can truly reflect the relative sensitivity of each metric to changes in design parameters, rather than being constrained by the magnitude of its original physical quantity.

[0056] Reference Figure 1 and Figure 2 After obtaining the normalized performance metrics, the adaptive fusion module further performs sensitivity analysis to quantify the specific impact of changes in design decision variables on each performance metric. The core of this process lies in constructing and solving the Jacobian matrix. The Jacobian matrix establishes a linear mapping relationship from the design parameter space to the performance index space.

[0057] Specifically, the adaptive fusion module first defines the vehicle's design decision variable vector. Design decision variable vector It includes adjustable physical parameters that affect vehicle performance, such as the spring stiffness coefficients of the front and rear suspensions, the damping coefficients of the shock absorbers, the torsional stiffness of the anti-roll bars, and the geometric coordinates of hard points, etc., in order to construct... Jacobian matrix of dimension The adaptive fusion module needs to calculate each metric. For each design variable The partial derivatives are used to adapt the adaptive fusion module to support two numerical differential calculation modes, adapting to different calculation accuracy requirements and simulator compatibility.

[0058] In the first calculation mode, the adaptive fusion module uses the central difference method for gradient approximation. The central difference method designs the decision variable vector at the current design point. Small positive and negative perturbation steps are applied to each dimension, and the tangent slope is estimated through quadratic simulation. For the nth element in the matrix... Line 1 Column elements The calculation formula is as follows: ; in: Represents the Jacobian matrix in the form of the first... Line 1 The element of the column, in physical terms, is the first... The performance metric relative to the first Sensitivity coefficients of each design variable; Represents the i-th index function For the j-th variable The partial derivatives; : Represents the function mapping for vehicle dynamics simulation and index extraction process, with input being a parameter vector and output being the first parameter vector. Individual performance index values; Indicates that for the first The step size of the small perturbation applied to each design variable, the selection of the small perturbation step size is determined by the variable. A certain percentage of the nominal value needs to be balanced between truncation error and rounding error; Indicates the first The unit basis vector in the i-th direction, that is, the vector in the i-th direction One element is 1, and the rest are 0; This represents the total span between positive and negative disturbance points.

[0059] In the second calculation mode, the adaptive fusion module adopts the complex step method. The complex step method utilizes the Taylor series expansion property of the analytic function to perform perturbation calculation in the complex domain, expanding the design variables into complex form. While keeping the real part unchanged, a small perturbation is applied to the imaginary part. The calculation formula is as follows: ; in: This represents the operation of taking the imaginary part of a complex number result; This indicates that the first element in the input variable vector will be... After adding pure imaginary perturbations to each element, the entire simulation calculation process is executed. This requires that the underlying dynamics solution library and index calculation library be overloaded to support data types that support complex number operations.

[0060] By iterating through all Each design variable is used to construct a complete Jacobian matrix column by column using the adaptive fusion module. The Jacobian matrix not only reveals the direction (sign) and intensity (absolute value) of the influence of parameters on each indicator, but also provides a mathematical basis for subsequent automatic weight allocation based on sensitivity.

[0061] Reference Figure 1 and Figure 2 The weight construction and scoring logic executed by the adaptive fusion module aims to address the imbalance in physical magnitude and design sensitivity of multi-dimensional performance indicators. Traditional fixed-weight methods cannot handle the differences in response to parameter changes among different indicators in the design space; that is, some indicators are extremely sensitive to parameters (easily changed), while others are extremely insensitive (difficult to optimize). To achieve balanced improvement across performance dimensions during multi-objective optimization, this invention employs a dynamic weight allocation strategy based on inverse sensitivity transformation, generating a Jacobian matrix... It is transformed into a scalar evaluation function to guide design optimization.

[0062] The adaptive fusion module first processes the Jacobian matrix. Perform row-wise statistical analysis to quantify the global activity of each performance metric at the current design point. For the first... Each performance metric is calculated by the adaptive fusion module relative to all... The average absolute sensitivity of each design decision variable This calculation process is not only an arithmetic mean of the sensitivity values, but also physically characterizes the plasticity or variability of this index in the local space of the current design parameters. The calculation formula is as follows: ; in, As a measure of A scalar measure of the activity level of a performance indicator; the larger the value, the more influenced the indicator is by design variables. The more pronounced the disturbance, the more significant the impact. The normalization coefficients eliminate the influence of the design variable dimensions; This indicates that the Jacobian matrix is... Iterate through all elements and sum them. Indicates taking the first Line 1 The absolute value of the column elements.

[0063] Based on the calculated average absolute sensitivity The adaptive fusion module constructs nonnormalized weights that are inversely proportional to the sensitivity amplitude. This constitutes the core mechanism of adaptive fusion: by assigning lower weights to high-sensitivity indicators, their dominance in the overall score is suppressed, preventing numerical oscillations caused by drastic fluctuations in high-sensitivity indicators during the optimization process; simultaneously, higher weights are assigned to low-sensitivity indicators, amplifying their signal strength in the overall score. This forces subsequent parameter optimization algorithms to invest more search power in performance dimensions that are difficult to improve. To ensure the robustness of numerical computation, a regularization factor is introduced. The calculation formula is as follows: ; in, For the first Initial reverse weights for each indicator; Set to a very small positive number (e.g.) As a regularization term, its technical function is to prevent the sensitivity of a certain indicator from being affected. When the denominator approaches zero, the overflow error caused by the denominator being zero ensures the stability of the algorithm under various extreme design points.

[0064] To map the calculated weights to relative importance coefficients that conform to probabilistic statistical meaning and to ensure the mathematical convexity of the comprehensive scoring function, the adaptive fusion module normalizes the initial weight vector. This normalization process maps the weights of all dimensions to... Given an interval where the sum of all constraint weights is strictly equal to 1, the calculation formula is as follows: ; in, Indicates the first The final normalized weight of the first performance metric directly determines the first... The contribution percentage of each indicator in the final score; For the first The initial weights of each indicator; It is for all systems The initial weights of the six evaluation dimensions (in this embodiment) are summed to form the normalized denominator.

[0065] Finally, the adaptive fusion module performs the synthesis operation of the comprehensive score. The synthesis operation combines the multi-dimensional performance index vectors that have been normalized to the baseline through a linear weighted combination. Compressed into a single scalar score Scalar score As the objective function, it can be directly called by external optimization algorithms (such as genetic algorithms, particle swarm optimization, or gradient descent) for minimization optimization. The calculation formula is as follows: ;in, This represents the linear superposition of all evaluation dimensions.

[0066] To achieve the aforementioned vehicle dynamics simulation, complex working condition construction, and multi-dimensional index calculation, especially to meet the computing power requirements for frequent calls to the simulation model for sensitivity analysis in adaptive weight optimization, this invention uses C++ to build a layered and modular high-performance computing architecture. Through memory management optimization, object-oriented polymorphic design, and efficient linear algebra library integration, it solves the technical bottlenecks of low execution efficiency and high memory overhead in traditional scripting languages ​​when performing large-scale iterative simulations.

[0067] Specifically, the software architecture of the evaluation platform mainly consists of three core layers from bottom to top: the basic mathematical operation layer, the dynamic component object layer, and the simulation control and application layer.

[0068] At the basic mathematical operation layer, considering that the aforementioned dynamic equations involve high-dimensional matrix multiplication, inversion, and decomposition operations, the platform integrates a high-performance linear algebra template library at the bottom layer. The basic mathematical operation layer encapsulates core data structures such as MatrixXd and VectorXd, and uses the SIMD instruction set to accelerate matrix operations at the hardware level. In addition, for the aforementioned numerical integration requirements, this layer implements a general integrator interface Integrator, which is concretized as the RungeKutta4 class. The RungeKutta4 class is designed as a general solver that does not depend on a specific physical model. It only accepts state vector pointers and differential equation callback functions, thereby decoupling the numerical algorithm from the physical model and facilitating the replacement of more efficient integration algorithms without breaking the model code structure.

[0069] At the dynamics component object layer, the platform adopts object-oriented programming principles, abstracting the physical vehicle system into a hierarchical class structure. The core class, VehicleSystem, serves as the container for the entire vehicle, internally combining sub-objects such as Chassis (body), Suspension, Tire, and Powertrain. To support different types of suspension or tire models (e.g., switching between linear tires and magic formula tires for different precision requirements), these subsystem classes all inherit from an abstract base class and implement polymorphism through virtual functions. For example, the Tire base class defines a pure virtual function, which specific subclasses override to implement concrete mechanical calculations.

[0070] This design allows the system to dynamically load different component models at runtime via configuration files without recompiling the main program.

[0071] In terms of data interaction, components pass state data through constant references or smart pointers, avoiding deep copying of large-scale data objects and reducing memory bus bandwidth pressure, which is crucial for ensuring real-time performance within millisecond-level simulation steps. At the simulation control and application layer, the core module is responsible for orchestrating the entire evaluation process.

[0072] Parameter Injection Mechanism: To support sensitivity analysis, the core module implements a parameter reflection mechanism based on hash mapping. This mechanism allows external optimization algorithms to directly index and modify the object member variables of the corresponding object in memory using string names. This enables efficient calculation of the Jacobian matrix. At that time, the program can quickly complete in memory. The parameters are perturbed and reset without re-instantiating the entire vehicle object, thus improving the efficiency of differential calculation.

[0073] Memory pool technology: For the large amount of time-domain data generated during simulation, this layer adopts a pre-allocated memory pool strategy. Before the simulation starts, according to the preset simulation duration and step size, the system allocates a contiguous memory block in the heap area at once to store the data. This avoids the frequent triggering of dynamic memory allocation by the operating system during the time step of the loop, eliminates the risk of memory fragmentation, and reduces system call latency.

[0074] Furthermore, the platform's input / output interfaces are designed to support multi-format parsing. The input end uses the ConfigParser class to parse JSON or XML configuration files, mapping vehicle hardpoint coordinates, stiffness and damping parameters, and operating condition definitions to C++ object properties. The output end uses the DataLogger class, employing double buffering technology to asynchronously write simulation results to binary or CSV files, ensuring that disk I / O operations do not block the main computing thread, thus guaranteeing the smoothness and time determinism of the simulation process. Through this in-depth architectural optimization, the C++ evaluation platform built by this invention can complete hundreds of full-vehicle dynamic operating condition simulations per second in a standard desktop computing environment, providing solid computing power support for the aforementioned multi-dimensional index extraction and adaptive weight optimization.

Claims

1. A simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of vehicles, characterized in that, Includes the following steps: S1. Based on the parametric construction method, corresponding standard working condition prototypes are generated for mobility evaluation, economic evaluation and comprehensive comfort evaluation, and simulation environment input data including wheel road surface roughness excitation input is output. S2. Construct a vehicle dynamics model that includes the sprung mass and unsprung mass of the vehicle, and use the input data of the simulation environment to perform time-domain simulation and solve the problem, and output the time-domain response data of the vehicle. S3. Receive the time-domain response data and extract performance indicators according to the preset evaluation dimensions. The performance indicators include at least maneuverability indicators, economy indicators, vertical comfort indicators, and attitude control indicators. S4. The extracted performance indicators are subjected to baseline normalization processing. The sensitivity Jacobian matrix of the vehicle design decision variables to each performance indicator is calculated. Based on the sensitivity Jacobian matrix, adaptive weights are calculated according to the logic that is inversely proportional to the sensitivity amplitude. The adaptive weights are then used to perform weighted summation on the normalized performance indicators to finally synthesize a scalar score of the vehicle's comprehensive driving performance.

2. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of vehicles according to claim 1, characterized in that, In step S1, the output, which includes the simulation environment input data containing the wheel-road surface roughness excitation input, specifically includes: The spatial power spectral density of road surface roughness is represented by a spatial frequency spectrum model. The spatial power spectral density is calculated using the spectral density value at a reference frequency, the ratio of the spatial frequency to the reference frequency, and the spectral index. The spatial power spectral density is discretized, and the amplitude corresponding to each discrete frequency point is calculated. Using the vehicle's left front wheel as a reference, the road surface unevenness of the left front wheel is synthesized using the amplitude and random phase in the inverse frequency spectrum method. The right wheel complex spectrum component is constructed based on the preset coherence function. The right front wheel road surface unevenness is synthesized by weighted superposition of the left wheel complex spectrum component and the auxiliary complex spectrum component, and the real part is taken. Based on the lag relationship between the wheelbase of the front and rear wheels of the vehicle, the road surface unevenness of the left front wheel and the road surface unevenness of the right front wheel are spatially delayed to generate the road surface unevenness of the left rear wheel and the road surface unevenness of the right rear wheel.

3. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of a vehicle according to claim 1, characterized in that, In step S2, the process of constructing the vehicle dynamics model includes: Fourteen independent generalized coordinates are selected to form a state vector, which includes six degrees of freedom describing the translational and rotational attitude of the vehicle body, four degrees of freedom describing the vertical displacement of the four unsprung masses, and four degrees of freedom describing the rotation angle of the four wheels. Based on the principles of Lagrange mechanics, the total kinetic energy and total potential energy of the vehicle system are calculated, and the Lagrange function is constructed. Based on the Lagrange function, establish the dynamic matrix equations containing the generalized mass matrix, nonlinear force vector, and generalized force vector; By inverting the dynamic matrix equation, a generalized acceleration vector is obtained. Then, a numerical integration algorithm is used to recursively derive the generalized acceleration vector to obtain the time-domain response data of the vehicle, which includes displacement, velocity, and acceleration.

4. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of a vehicle according to claim 3, characterized in that, The construction process of the generalized force vector includes: Based on the longitudinal slip ratio, lateral slip angle and vertical load of the wheel, the longitudinal force, lateral force and self-aligning torque in the tire coordinate system are calculated using the Magic Formula semi-empirical tire model, and the calculated tire forces are converted to the vehicle coordinate system. The vertical force generated by the suspension is calculated based on the difference between the vertical motion state of the vehicle body at the suspension connection point and the motion state of the unsprung mass, combined with the preset nonlinear spring stiffness characteristic curve and damper damping characteristic curve. The tire force and the vertical force generated by the suspension, which have been transformed to the vehicle coordinate system, are projected onto the direction corresponding to the generalized coordinates to synthesize the generalized force vector.

5. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of a vehicle according to claim 1, characterized in that, In step S3, extracting performance indicators based on preset evaluation dimensions specifically includes: For mobility evaluation, the minimum time required for a vehicle to reach a preset road endpoint under free acceleration conditions with zero initial speed is identified, and the minimum time is used as the mobility index. For economic evaluation, real-time operating status parameters of the front drive motor and the rear drive motor of the vehicle under standard driving cycle conditions are obtained, including speed and torque. Based on the pre-stored motor system efficiency spectrum, the real-time electric power of the front drive motor and the rear drive motor is calculated, and the energy consumption in the driving mode and the energy recovery in the braking mode are distinguished during the calculation process. The economic index characterizing the total energy consumption is obtained by performing a definite integral operation with respect to the sum of the real-time electrical power of the front drive motor and the rear drive motor over time.

6. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of a vehicle according to claim 1, characterized in that, In step S3, the specific methods for extracting the vertical comfort index and the attitude control index include: A frequency-weighted acceleration is obtained by applying a vertical frequency-weighted filter that conforms to human perception to the time-domain signal of the vertical acceleration of a vehicle passing through segments of a random uneven road surface. The root mean square value of the frequency-weighted acceleration is calculated, and the vibration dose value of the frequency-weighted acceleration is calculated using the fourth power cumulative method. The root mean square value and the vibration dose value are used as the vertical comfort index. The pitch rate response of the vehicle under braking and acceleration conditions and the roll rate response under two-lane change conditions are analyzed, and the root mean square value of pitch rate and the root mean square value of roll rate are calculated respectively, which are used as the attitude control index.

7. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of a vehicle according to claim 1, characterized in that, In step S4, the process of calculating the sensitivity Jacobian matrix employs a complex step method, which specifically includes: The vehicle design decision variables are extended to complex form, and while keeping the real part of the vehicle design decision variables unchanged, a small pure imaginary perturbation step size is applied to the imaginary part. The vehicle dynamics simulation and index extraction process is executed to obtain the performance index response in the complex domain; Extract the imaginary part of the performance index response in the complex domain, divide the imaginary part by the pure imaginary perturbation step size to obtain the partial derivative of the performance index with respect to the vehicle design decision variables, and traverse all vehicle design decision variables to construct the sensitivity Jacobian matrix.

8. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of a vehicle according to claim 1, characterized in that, In step S4, calculating the adaptive weights specifically includes: Perform row-oriented statistical analysis on the sensitivity Jacobian matrix to calculate the average absolute sensitivity of each performance index relative to all vehicle design decision variables; Construct an initial inverse weight that is inversely proportional to the average absolute sensitivity value, and introduce a regularization factor into the denominator to prevent numerical overflow; The initial inverse weights of all evaluation dimensions are normalized so that the sum of all weights equals one, thus obtaining the final adaptive weights.

9. The simulation evaluation method for extracting and adaptively fusing multi-dimensional performance indicators of a vehicle according to claim 1, characterized in that, In step S4, the baseline normalization process specifically includes: Select a known vehicle model as the baseline model; The baseline model is controlled to run under the same standard evaluation conditions as the vehicle model to be evaluated, and a set of performance indicators is calculated as baseline reference values. The performance index of the vehicle model to be evaluated, output by the performance index calculation module, is divided by the corresponding baseline reference value to eliminate the difference in dimensions and obtain dimensionless performance indexes for weighted summation by the adaptive weights.

10. A simulation evaluation platform for extracting and adaptively fusing multi-dimensional performance indicators of vehicles, characterized in that, The simulation evaluation method for extracting and adaptively fusing multidimensional performance indicators of a vehicle, as described in claims 1-9, includes: The working condition generation module is used to generate standard working condition prototypes based on parametric construction methods, and outputs simulation environment input data including wheel and road surface roughness excitation inputs. The vehicle dynamics module is used to construct a vehicle dynamics model, perform time-domain simulation solutions using the input data from the simulation environment, and output the vehicle's time-domain response data. The performance index calculation module is used to receive the time-domain response data and extract multi-dimensional performance indexes, including maneuverability indexes, economic indexes, vertical comfort indexes, and attitude control indexes. The adaptive fusion module is used to perform baseline normalization on the multidimensional performance indicators, calculate the sensitivity Jacobian matrix of the vehicle design decision variables to each performance indicator, allocate adaptive weights according to a strategy inversely proportional to the sensitivity amplitude, and output a comprehensive score.

Images