Method and system for detecting braking performance of electric vehicle based on multi-dimensional sensing

Abstract

The application provides a method and system for detecting braking performance of an electric vehicle based on multi-dimensional sensing, and relates to the field of electric vehicles.The method comprises the following steps: determining a plurality of braking vibration detection components and braking temperature detection components through a dynamic twin group and a parameter self-adaptive bat algorithm; determining a plurality of braking performance detection scenes based on the plurality of braking vibration detection components and the braking temperature detection components; collecting a multi-dimensional sensing data set of the electric vehicle to be detected based on the plurality of braking vibration detection components, the braking temperature detection components and the braking performance detection scenes, wherein the multi-dimensional sensing data set comprises a vibration sequence of the plurality of braking vibration detection components, a temperature sequence of the plurality of braking temperature detection components and a braking distance under each braking performance detection scene; and generating a braking performance evaluation result based on the multi-dimensional sensing data set of the electric vehicle to be detected, which has the advantages of improving the efficiency and accuracy of the detection of the braking performance of the electric vehicle.

Description

Technical Field

[0001] The present invention relates to the field of electric vehicles, and in particular to a multi-dimensional sensing-based electric vehicle braking performance detection method and system. Background Art

[0002] With the rapid development of the new energy vehicle industry, the ownership of electric vehicles has shown explosive growth. As the core indicator of active vehicle safety, the accuracy and real-time performance of braking performance detection technologies have attracted increasing attention from the industry. The braking system of electric vehicles not only involves traditional mechanical braking, but also integrates multiple architectures such as regenerative braking and electronic booster braking (e.g., Nbooster decoupled electronic booster). During the braking process, it is necessary to coordinate the motor feedback braking force and the hydraulic braking force, and the system complexity far exceeds that of traditional fuel vehicles. Therefore, how to achieve comprehensive and accurate detection of braking performance through multi-dimensional sensing means has become a technical bottleneck that needs to be broken urgently at present.

[0003] Existing braking performance detection methods mainly include two types: the braking distance method and the braking skid mark method. The braking distance method evaluates braking performance based on the distance the vehicle travels from emergency braking at a specified vehicle speed to a complete stop. On a dry and flat road surface, the braking performance is qualified only if the no-load braking distance is ≤19m or the full-load braking distance is ≤20m without deviation or sideslip. The braking skid mark method judges by observing the skid marks left by tires on the road surface during emergency braking, which requires that all wheels have skid marks without deviation. However, the above two methods can only obtain macroscopic indicators such as braking distance and deceleration, and cannot reflect the dynamic changes of the vehicle body attitude during the braking process, such as key information including braking nosedive, roll, and yaw; for vehicles equipped with ABS or regenerative braking systems, the skid mark method is no longer applicable, and the detection dimensions are severely insufficient.

[0004] Therefore, it is necessary to provide a multi-dimensional sensing-based electric vehicle braking performance detection method and system, so as to improve the efficiency and accuracy of electric vehicle braking performance detection. Summary of the Invention

[0005] The present invention provides a multi-dimensional sensing-based electric vehicle braking performance detection method, comprising: determining a plurality of braking vibration detection components and braking temperature detection components through a bat algorithm with dynamic double subgroups and adaptive parameters; determining a plurality of braking performance detection scenarios based on the plurality of braking vibration detection components and the braking temperature detection components; collecting a multi-dimensional sensing dataset of the electric vehicle to be detected based on the plurality of braking vibration detection components, the braking temperature detection components and the braking performance detection scenarios, wherein the multi-dimensional sensing dataset includes the vibration sequences of the plurality of braking vibration detection components, the temperature sequences of the plurality of braking temperature detection components and the braking distance under each braking performance detection scenario; generating a braking performance evaluation result based on the multi-dimensional sensing dataset of the electric vehicle to be detected.

[0006] Furthermore, multiple braking vibration detection components and braking temperature detection components are determined using a dynamic bisubgroup and parameter-adaptive bat algorithm, including: determining multiple vibration detection components and temperature detection components; collecting a first test dataset of multiple test vehicles under a calibrated braking performance testing scenario, wherein the first test dataset includes vibration sequences of multiple vibration detection components and temperature sequences of temperature detection components; and determining multiple braking vibration detection components and braking temperature detection components based on the first test dataset of multiple test vehicles under a calibrated braking performance testing scenario using a dynamic bisubgroup and parameter-adaptive bat algorithm.

[0007] Furthermore, using a dynamic biswarm and parameter-adaptive bat algorithm, based on a first test dataset of multiple test vehicles in a calibrated braking performance testing scenario, multiple braking vibration detection components and braking temperature detection components are determined, including: S11, initializing a first population and a second population based on the first test dataset of multiple test vehicles in a calibrated braking performance testing scenario, wherein each bat in the first population and the second population represents a combination of at least one vibration detection component and at least one temperature detection component; S12, constructing a fitness function; S13, iteratively optimizing the first population and the second population until... The first termination condition is met, wherein, for each bat in the first group, in each iteration, the adaptive frequency and adaptive pulse rate of the bat are calculated by a clustering algorithm; for each bat in the second group, in each iteration, the bats to be swapped are determined, and the bats are updated based on the swapped bats; S14, determine whether the first group and the second group meet the second termination condition. If yes, determine multiple braking vibration detection components and braking temperature detection components based on the current global optimal solution of the first group and the second group. If no, proceed to S15; S15, swap the information of the first group and the second group, and proceed to S13.

[0008] Furthermore, based on the first test dataset of multiple test vehicles in a calibrated braking performance testing scenario, the first and second populations are initialized, including: for any two vibration detection components, calculating the response similarity of the two vibration detection components based on the first test dataset of multiple test vehicles in a calibrated braking performance testing scenario; for any two temperature detection components, calculating the response similarity of the two temperature detection components based on the first test dataset of multiple test vehicles in a calibrated braking performance testing scenario; and initializing the first and second populations based on the response similarity of any two vibration detection components and the response similarity of any two temperature detection components.

[0009] Furthermore, for each bat in the first group, in each iteration, the adaptive frequency and adaptive pulse rate of the bat are calculated using a clustering algorithm, including: calculating the similarity between any two bats; clustering the bats in the first group according to the similarity between any two bats using a clustering algorithm; and for each bat in the first group, calculating the adaptive frequency and adaptive pulse rate of the bat according to the number of bats included in the cluster to which the bat belongs.

[0010] Furthermore, based on multiple braking vibration detection components and braking temperature detection components, multiple braking performance testing scenarios are determined, including: collecting a second test dataset of multiple test vehicles under multiple performance testing scenarios based on multiple braking vibration detection components and braking temperature detection components, wherein the second test dataset includes vibration sequences of multiple vibration detection components, temperature sequences of braking temperature detection components, and braking distances; calculating the effective value of each performance testing scenario based on the second test dataset of multiple test vehicles under multiple performance testing scenarios; and determining multiple braking performance testing scenarios based on the effective value of each performance testing scenario.

[0011] Furthermore, based on the second test dataset of multiple test vehicles under multiple performance testing scenarios, the effective value of each performance testing scenario is calculated, including: for each braking performance testing scenario, based on the second test dataset of multiple test vehicles under the performance testing scenario, the response correlation coefficient corresponding to the braking performance testing scenario is calculated, and based on the response correlation coefficient corresponding to the braking performance testing scenario, the effective value of each performance testing scenario is calculated.

[0012] Furthermore, based on a second test dataset of multiple test vehicles in performance testing scenarios, the response correlation coefficients corresponding to the braking performance testing scenarios are calculated, including: for any two test vehicles, calculating the braking distance difference, vibration response similarity, and temperature response similarity of the corresponding performance testing scenarios for the two test vehicles; based on the braking distance difference, vibration response similarity, and temperature response similarity of the corresponding performance testing scenarios for any two test vehicles, calculating the vibration response correlation coefficient and temperature response correlation coefficient; and based on the vibration response correlation coefficient and temperature response correlation coefficient, calculating the response correlation coefficients corresponding to the braking performance testing scenarios.

[0013] Furthermore, based on the multidimensional sensor dataset of the electric vehicle to be detected, braking performance evaluation results are generated, including: acquiring multidimensional sensor datasets of multiple calibrated electric vehicles and braking performance evaluation results; identifying similar calibrated electric vehicles based on the multidimensional sensor dataset of the electric vehicle to be detected and the multidimensional sensor datasets of multiple calibrated electric vehicles; and generating braking performance evaluation results based on the similar calibrated electric vehicles.

[0014] This invention provides a multi-dimensional sensing-based electric vehicle braking performance testing system, comprising: a detection optimization module, used to determine multiple braking vibration detection components and braking temperature detection components through a dynamic bisubgroup and parameter-adaptive bat algorithm; the detection optimization module is also used to determine multiple braking performance testing scenarios based on the multiple braking vibration detection components and braking temperature detection components; a data acquisition module, used to acquire a multi-dimensional sensing dataset of the electric vehicle to be tested based on the multiple braking vibration detection components, braking temperature detection components, and braking performance testing scenarios, wherein the multi-dimensional sensing dataset includes vibration sequences of multiple braking vibration detection components, temperature sequences of multiple braking temperature detection components, and braking distances under each braking performance testing scenario; and a performance evaluation module, used to generate braking performance evaluation results based on the multi-dimensional sensing dataset of the electric vehicle to be tested.

[0015] Compared with existing technologies, the electric vehicle braking performance detection method and system based on multi-dimensional sensing provided by this invention has at least the following beneficial effects: The braking vibration detection component and temperature detection component are determined by a dynamic bi-subgroup and parameter-adaptive bat algorithm. Clustering algorithms are used to dynamically adjust the search frequency and pulse rate of each "bat," allowing bats in the dense solution space to undertake global exploration and bats in the sparse solution space to undertake local exploration. Simultaneously, a second subgroup maintains information flow through a bat-exchanging mechanism. The two subgroups alternately exchange information to achieve complementary and integrated search strategies, accurately selecting the optimal component combination with high information coverage and low redundancy while avoiding premature convergence. Secondly, based on the selected components, braking performance detection scenarios are further determined. The strong correlation between vibration response, temperature response, and braking distance is verified by calculating response correlation coefficients, ensuring that the selected scenarios reliably reflect braking performance differences in both sensing dimensions, thus selecting high-quality detection scenarios with true discriminative power and repeatability. Finally, the vehicle under test only needs to collect multi-dimensional sensor data in the calibrated effective scenarios. Combined with historical evaluation results of similar vehicles in the calibrated vehicle database, rapid inference is performed, obtaining accurate braking performance evaluation without undergoing a complete calibration process, significantly reducing detection costs and time. Attached Figure Description

[0016] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein: Figure 1 This is a flowchart illustrating a method for detecting the braking performance of an electric vehicle based on multidimensional sensing, according to some embodiments of this specification. Figure 2 This is a flowchart illustrating the dynamic twin group and parameter adaptive bat algorithm according to some embodiments of this specification; Figure 3 This is a schematic diagram of a multi-dimensional sensing-based electric vehicle braking performance detection system according to some embodiments of this specification. Detailed Implementation

[0017] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0018] Figure 1 This is a flowchart illustrating a method for detecting the braking performance of an electric vehicle based on multidimensional sensing, as shown in some embodiments of this specification. Figure 1 As shown, the method for detecting the braking performance of electric vehicles based on multidimensional sensing may include the following steps.

[0019] S1. Multiple braking vibration detection components and braking temperature detection components are determined through dynamic twin group and parameter adaptive bat algorithm.

[0020] In some embodiments, S1 includes: Identify multiple vibration detection components and temperature detection components; The first test dataset is collected from multiple test vehicles under a calibrated braking performance testing scenario. The first test dataset includes vibration sequences from multiple vibration detection components and temperature sequences from multiple temperature detection components. The vibration sequences are composed of the vibration amplitudes of the vibration detection components at multiple time points, and the temperature sequences are composed of the temperatures of the temperature detection components at multiple time points. Using a dynamic twin group and parameter-adaptive bat algorithm, multiple braking vibration detection components and braking temperature detection components are identified based on the first test dataset of multiple test vehicles in a calibrated braking performance testing scenario.

[0021] Vibration detection components refer to vehicle parts that may generate measurable vibration responses due to braking force during the braking process of an electric vehicle. Examples include: brake discs, where high-frequency vibrations are generated when the friction pads contact the disc surface during braking, and the amplitude and frequency of these vibrations can reflect the degree of disc wear and thermal deformation; brake calipers, where fluctuations in clamping force cause low-frequency vibrations, reflecting the uniformity of braking force output; brake pedals and pedal brackets, where the pulsation of brake pedal force is transmitted to these components, generating vibrations that reflect pedal feel and braking coordination; front suspension tower tops and shock absorber upper supports, where braking dive causes longitudinal impact vibrations, reflecting the vehicle's attitude stability; wheel hubs and bearings, where imbalances or a tendency to lock up during braking will generate abnormal vibrations, reflecting the wheel's operating status; and motor and reducer housings, where the motor reverses during regenerative braking, generating electromagnetic and mechanical vibrations that reflect the operating status of the electric drive system.

[0022] Temperature detection components refer to vehicle parts that may experience significant temperature rises during braking due to energy conversion and require close monitoring. Examples include: brake disc friction surfaces, the main area bearing braking heat, where temperature directly reflects braking load and the risk of heat fade; brake caliper pistons and brake fluid, the area generating heat in hydraulic transmission, where excessive temperature can lead to brake fluid vapor lock; drive motor stator windings and rotor, where electrical energy is fed back during regenerative braking, where temperature reflects the motor's thermal load; reducer gears and bearings, where regenerative braking torque is transmitted through the reducer, where temperature rise reflects the thermal state of the transmission chain; and the power battery pack, where the braking energy recovery current passes through the battery, where temperature rise reflects the battery's charging and discharging thermal effects.

[0023] The calibrated braking performance test scenario can include multiple pre-defined test scenario parameters, such as initial vehicle speed and road surface roughness.

[0024] Figure 2 This is a flowchart illustrating the dynamic twin group and parameter adaptive bat algorithm according to some embodiments of this specification, such as... Figure 2 As shown, in some embodiments, the dynamic twin group and parameter adaptive bat algorithm may include the following steps: S11. Based on the first test dataset of multiple test vehicles under the calibrated braking performance test scenario, initialize the first population and the second population, wherein each bat in the first population and the second population represents a combination of at least one vibration detection component and at least one temperature detection component. S12. Construct the fitness function; S13. Iteratively optimize the first and second populations until the first termination condition is met. For each bat in the first population, in each iteration, the adaptive frequency and adaptive pulse rate of the bat are calculated by a clustering algorithm. For each bat in the second population, in each iteration, the bats to be swapped are determined, and the bats are updated based on the swapped bats. S14. Determine whether the first group and the second group satisfy the second termination condition. If yes, determine multiple braking vibration detection components and braking temperature detection components based on the current global optimal solution of the first group and the second group. For example, select the vibration detection components and temperature detection components included in the current global optimal solution with the largest fitness value as multiple braking vibration detection components and braking temperature detection components. If not, execute S15. S15. Exchange information between the first and second populations, and execute S13.

[0025] In some embodiments, based on a first test dataset of multiple test vehicles in a calibrated braking performance testing scenario, a first population and a second population are initialized, including: For any two vibration detection components, based on the first test dataset of multiple test vehicles in the calibrated braking performance test scenario, the response similarity of the two vibration detection components is calculated. Specifically, the cosine similarity of the vibration sequence of each test vehicle corresponding to the two vibration detection components in the calibrated braking performance test scenario is calculated, and the mean of the cosine similarity of the vibration sequences of all test vehicles corresponding to the two vibration detection components in the calibrated braking performance test scenario is calculated as the response similarity of the two vibration detection components. For any two temperature detection components, based on the first test dataset of multiple test vehicles in the calibrated braking performance test scenario, the response similarity of the two temperature detection components is calculated. Specifically, the cosine similarity of the temperature sequence of each test vehicle corresponding to the two braking temperature detection components in the calibrated braking performance test scenario is calculated, and the average of the cosine similarity of the temperature sequence of all test vehicles corresponding to the two braking temperature detection components in the calibrated braking performance test scenario is taken as the response similarity of the two braking temperature detection components. The first and second populations are initialized based on the response similarity of any two vibration detection components and any two temperature detection components.

[0026] For example, initializing the first and second groups based on the response similarity of any two vibration detection components and any two temperature detection components can include the following process: S21. For each vibration detection component, based on the response similarity between any two vibration detection components, determine an optional set of vibration detection components, wherein the optional set includes a set of vibration detection components whose response similarity with the vibration detection component is less than a response similarity threshold (e.g., 0.5). S22. For each vibration detection component, based on the test dataset of multiple test vehicles in the calibrated braking performance test scenario, calculate the response difference value of the vibration detection component. Specifically, calculate the Euclidean distance between the vibration sequences of any two test vehicles corresponding to the vibration detection component in the calibrated braking performance test scenario, and average the Euclidean distances of all the vibration sequences of any two test vehicles corresponding to the vibration detection component in the calibrated braking performance test scenario to obtain the response difference value of the vibration detection component. S23. For each temperature detection component, based on the response similarity between any two temperature detection components, determine an optional set of temperature detection components, wherein the optional set includes a set of temperature detection components whose response similarity with the temperature detection component is less than a response similarity threshold (e.g., 0.5). S24. For each temperature detection component, based on the test dataset of multiple test vehicles in the calibrated braking performance test scenario, calculate the response difference value of the vibration detection component. Specifically, calculate the Euclidean distance between the temperature sequences of any two test vehicles corresponding to the temperature detection component in the calibrated braking performance test scenario, and average the Euclidean distances of the temperature sequences of all any two test vehicles corresponding to the temperature detection component in the calibrated braking performance test scenario to obtain the response difference value of the temperature detection component. S25. When generating each bat, first randomly determine the number of vibration detection components and the number of temperature detection components included in the bat. Randomly select one vibration detection component and one temperature detection component. From the selectable set of vibration detection components, select the vibration detection component with the largest response difference value. From the selectable set of temperature detection components, select the temperature detection component with the largest response difference value. If the number of selected vibration detection components is greater than 1, then select the vibration detection component with the largest response difference value from the intersection of the selectable sets of all selected vibration detection components. If the number of selected temperature detection components is greater than 1, then select the temperature detection component with the largest response difference value from the intersection of the selectable sets of all selected temperature detection components. Continue until the number of vibration detection components and the number of temperature detection components included in the bat are satisfied.

[0027] In some embodiments, for each bat in the first swarm, in each iteration, the adaptive frequency and adaptive pulse rate of the bat are calculated using a clustering algorithm, including: Calculate the similarity between any two bats. Specifically, calculate the overlap of the vibration detection components and the temperature detection components of the two bats, and take their average as the similarity between the two bats. Using clustering algorithms (such as K-means clustering), the bats in the first group are clustered based on the similarity between any two bats. For each bat in the first group, the adaptive frequency and adaptive pulse rate of the bat are calculated based on the number of bats in the cluster to which the bat belongs.

[0028] For example, the adaptive frequency and adaptive pulse rate of a bat can be calculated using the following formula:

[0029] in, f i Let i be the adaptive frequency of the i-th bat. f min The preset minimum frequency, f max The preset maximum frequency, N i Let be the number of bats in the cluster to which the i-th bat belongs. N min The minimum number of bats included in all clusters. r i Let i be the adaptive pulse rate of the i-th bat. r o This is the preset pulse rate.

[0030] Understandably, by clustering the bats in the first group into several clusters according to the similarity of the solutions, the more similar the component combination schemes represented by the bats in the cluster, the denser the solutions in that region, and the more likely the bats are to get trapped in a local optimum; conversely, the fewer the number of bats in the cluster, the more unique the solution is, and the surrounding area has not been fully explored.

[0031] The design logic for adaptive frequency is: when the number of bats in a cluster is small, Approaching 1, high frequencies correspond to small-step perturbations, allowing bats to perform fine-grained searches near their own solutions, fully exploring the local optimum potential of this unique component combination. When the number of bats in a cluster is large, low frequencies correspond to large-step perturbations, causing bats to jump out of existing dense areas for global searches, avoiding wasting iterations in already fully explored solution spaces. The design logic of the adaptive pulse rate is as follows: the fewer bats in a cluster, the larger the pulse rate, and bats perform local random walks with a higher probability, deeply exploring their own unique solutions; the more bats in a cluster, the smaller the pulse rate, and bats are more inclined to follow the global optimum, using collective intelligence to accelerate convergence. The synergy of these two approaches allows bats in dense clusters to undertake global exploration tasks, while bats in sparse clusters undertake local exploration tasks. This avoids the population converging prematurely to local optima while ensuring full exploration of unique component combination schemes, significantly improving search diversity and convergence efficiency, thus determining the optimal combination of vibration and temperature detection components faster and more accurately.

[0032] In the first type of group, iterative updates may include the following steps: For each bat, generate a random number in the range [0, 1]. When the random number is greater than the adaptive pulse rate, a new candidate solution is generated near the current global optimal solution. For example, a small number (e.g., 1, 2, etc.) of the vibration detection components and temperature detection components in the current global optimal solution are replaced as candidate solutions. When the random number is less than or equal to the adaptive pulse rate, the bat's position can be updated based on the bat's adaptive frequency. For example, the number of vibration detection components and temperature detection components that the bat needs to update can be calculated based on the bat's adaptive frequency.

[0033] in, Let be the number of vibration detection components that need to be updated for the i-th bat. N i,v Let i be the number of vibration detection components included in the i-th bat. Let be the number of temperature detection components that need to be updated for the i-th bat. N i,T Let be the number of temperature detection components included in the i-th bat, and [ ] be the rounding operation. Based on the number of vibration detection components that the bat needs to update, randomly select vibration detection components and replace them with vibration detection components that are not included. Based on the number of temperature detection components that the bat needs to update, randomly select temperature detection components and replace them with temperature detection components that are not included, as candidate solutions. Calculate the fitness value of the candidate solution. If the fitness value of the candidate solution is greater than the fitness value of the bat, then accept the candidate solution and replace the bat.

[0034] Calculate the overlap of vibration detection components and temperature detection components in two bats, and take their average as the similarity between the two bats.

[0035] For each bat, the bat with the lowest similarity is selected as the swap bat for that bat.

[0036] In the second group, iterative updates may include the following steps: For each bat, generate a random number in the range [0, 1]. When the random number is greater than the adaptive pulse rate, a new candidate solution is generated near the current global optimal solution. For example, a small number (e.g., 1, 2, etc.) of the vibration detection components and temperature detection components in the current global optimal solution are replaced as candidate solutions. When the random number is less than or equal to the adaptive pulse rate, the random swap bat includes the vibration detection component and temperature detection component (e.g., 1, 2, etc.) as candidate solutions. Calculate the fitness value of the candidate solution. If the fitness value of the candidate solution is greater than the fitness value of the bat, then accept the candidate solution and replace the bat.

[0037] The first termination condition can be reaching the maximum number of iterations, fitness convergence, etc.

[0038] The second termination condition can be multiple consecutive times (e.g., 3 times, 4 times, etc.), where the similarity between the current global optimal solution of the first group and the current global optimal solution of the second group is greater than the similarity threshold (e.g., 0.7), or the number of judgments is greater than the preset number threshold, etc.

[0039] Exchanging information between the first and second populations can include exchanging the current global optimal solution of the first population and the current global optimal solution of the second population.

[0040] The fitness function is related to the response similarity between any two vibration detection components included in the bat, the response difference between each vibration detection component, the response similarity between any two temperature detection components, and the response difference between each temperature detection component.

[0041] For example, the fitness function is:

[0042] in, F For fitness value, Let g be the response difference value of the g-th vibration detection component included in the bat. G The total number of vibration detection components included in the bat. D T,h The response difference value of the h-th temperature detection component included in the bat. H This refers to the total number of braking temperature detection components included in the bat. This represents the mean of the response similarity between any two vibration detection components included in the bat. Let w1, w2, w3, and w4 be the mean of the response similarities between any two temperature detection components included in the bat, with w1, w2, w3, and w4 as weights, where w1, w2, w3, and w4 are greater than 0. .

[0043] The response difference value measures the dispersion of a single component's response across multiple tests. A larger response difference value indicates a stronger ability to distinguish between different braking conditions (such as different initial velocities, different pedal forces, and different regenerative braking ratios), and contains richer effective information. Therefore, this factor encourages the selection of components that can capture rich braking features, ensuring that each selected component contributes independently. A higher mean response similarity between any two components indicates a high degree of overlap and functional redundancy between the selected components. This factor penalizes high similarity by taking the reciprocal; higher similarity results in lower fitness, thus forcing the algorithm to favor combinations of components with large differences in their response features, avoiding redundancy. These four factors work synergistically to achieve two goals simultaneously in the fitness function: first, ensuring sufficient information for each selected component (large difference value); and second, ensuring no information overlap between components (low similarity). The final output component combination covers the most braking performance dimensions with the fewest sensors while minimizing redundancy, improving the cost-effectiveness and diagnostic accuracy of the detection system.

[0044] The dual-swarm division of labor effectively balances global exploration and local exploitation. The first swarm dynamically calculates the adaptive frequency and impulse rate of each bat using a clustering algorithm. Bats in dense clusters perform global search tasks by hopping across the available solution space with low frequency and large steps, while bats in sparse clusters perform local exploitation tasks by meticulously exploring the vicinity of their own solutions with high frequency and small steps. This avoids the problem of uniform search behavior caused by all individuals using the same parameters in traditional bat algorithms. The second swarm uses a bat-exchanging mechanism. Each bat chooses an exchange partner and updates its information based on the partner's solution. This directed information exchange between individuals ensures continuous information flow within the swarm, preventing any bat from prematurely falling into a local optimum. The randomness introduced by the exchange mechanism further enhances swarm diversity. The dual-swarm information exchange mechanism achieves a complementary fusion of the two search strategies. When the first termination condition is not met, the two populations exchange information. This means that the high-frequency, fine-grained search strategy optimized by clustering in the first population and the diversified exploration strategy optimized by exchange in the second population mutually permeate each other. The first population gains the global exploration capability of the second population, and the second population gains the local exploration accuracy of the first population. The alternating action of the two strategies ensures that the algorithm maintains high search vitality throughout the iteration process, avoiding stagnation due to overly fine-grained search or divergence due to excessive random walks. The dual termination condition provides flexible convergence control. The first termination condition controls the optimization depth within a single iteration, ensuring that each population searches sufficiently in this iteration; the second termination condition controls the convergence timing of the overall algorithm, outputting the final result only when both populations reach the preset convergence criteria, avoiding the output of suboptimal solutions caused by premature convergence of a single population.

[0045] S2. Based on multiple braking vibration detection components and braking temperature detection components, multiple braking performance detection scenarios are determined.

[0046] In some embodiments, S2 specifically includes: Based on multiple braking vibration detection components and braking temperature detection components, a second test dataset is collected for multiple test vehicles under multiple performance testing scenarios. The second test dataset includes vibration sequences from multiple vibration detection components, temperature sequences from braking temperature detection components, and braking distances. Based on a second test dataset of multiple test vehicles in multiple performance testing scenarios, the effective value for each performance testing scenario is calculated. Based on the effective values ​​of each performance testing scenario, multiple braking performance testing scenarios are determined.

[0047] Among them, at least one detection scenario parameter (e.g., initial vehicle speed, road surface roughness, etc.) differs between any two performance detection scenarios.

[0048] In some embodiments, based on a second test dataset of multiple test vehicles in multiple performance testing scenarios, the effective value for each performance testing scenario is calculated, including: For each braking performance test scenario, based on the second test dataset of multiple test vehicles in the performance test scenario, the response correlation coefficient corresponding to the braking performance test scenario is calculated, and based on the response correlation coefficient corresponding to the braking performance test scenario, the effective value of each performance test scenario is calculated.

[0049] In some embodiments, based on a second test dataset of multiple test vehicles in a performance testing scenario, the response correlation coefficient corresponding to the braking performance testing scenario is calculated, including: For any two test vehicles, calculate the difference in braking distance, similarity of vibration response, and similarity of temperature response for the corresponding performance testing scenarios of the two test vehicles; Based on the braking distance difference, vibration response similarity, and temperature response similarity of any two test vehicles corresponding to the performance testing scenarios, calculate the vibration response correlation coefficient and the temperature response correlation coefficient. For example, take the braking distance difference and vibration response similarity of any two test vehicles corresponding to the performance testing scenarios as two variables, substitute them into the correlation coefficient calculation formula, and obtain the vibration response correlation coefficient. The calculation method of the temperature response correlation coefficient is similar and will not be repeated here. Based on the vibration response correlation coefficient and the temperature response correlation coefficient, the response correlation coefficient corresponding to the braking performance testing scenario is calculated. For example, the average value of the vibration response correlation coefficient and the temperature response correlation coefficient is used to obtain the response correlation coefficient corresponding to the braking performance testing scenario.

[0050] The larger the response correlation coefficient of the braking performance testing scenario, the larger the effective value of the performance testing scenario.

[0051] The effective values ​​of each performance test scenario are sorted from largest to smallest, and the top n (e.g., 3, 5, etc.) performance test scenarios are selected as braking performance test scenarios.

[0052] The fundamental purpose of calculating the correlation coefficients of vibration response and temperature response is to verify whether there is a strong correlation between vibration response and temperature response and the macroscopic performance indicator of braking distance under the same braking performance testing scenario. This ensures that the scenario can accurately reflect braking performance through both vibration and temperature sensing channels. Specifically, the calculation process of the vibration response correlation coefficient is as follows: the difference in braking distance between any two test vehicles under the scenario and the similarity of their vibration responses are substituted into the correlation coefficient formula as two variables. If the correlation coefficient is high, it indicates that the trend of vibration response change is highly consistent with the trend of braking distance change under the scenario, that is, the vibration characteristics can reliably map the difference in braking distance. Similarly, a high temperature response correlation coefficient indicates that the temperature characteristics can also reliably map the difference in braking distance. Only when both correlation coefficients are high does it indicate that a stable three-dimensional correlation has been formed among braking distance, vibration response, and temperature response under the scenario. At this point, the accuracy of comprehensively judging braking performance by fusing the two sensing dimensions of vibration and temperature is higher. If a scenario exhibits a high correlation coefficient only in vibration and a low correlation coefficient in temperature, it indicates insufficient excitation of braking thermal characteristics. Relying solely on vibration data and braking distance to determine braking performance will have blind spots; conversely, the same applies. Therefore, calculating the correlation coefficients of both dimensions separately and then averaging them to obtain the comprehensive response correlation coefficient essentially filters out high-quality detection scenarios where both vibration and temperature effectively reflect differences in braking performance. This ensures that the component combinations selected based on the bat algorithm can simultaneously acquire effective information from both sensor channels during actual testing, achieving multi-dimensional cross-validation of braking performance and improving the reliability and accuracy of the final evaluation results.

[0053] S3. Based on multiple braking vibration detection components, braking temperature detection components, and braking performance detection scenarios, collect multi-dimensional sensor datasets of the electric vehicle to be tested.

[0054] The multidimensional sensing dataset includes vibration sequences of multiple braking vibration detection components, temperature sequences of multiple braking temperature detection components, and braking distance for each braking performance detection scenario.

[0055] S4. Based on the multidimensional sensor dataset of the electric vehicle to be detected, generate braking performance evaluation results.

[0056] In some embodiments, S4 specifically includes: Acquire multiple calibrated multidimensional sensor datasets of electric vehicles and braking performance evaluation results; Based on the multidimensional sensor dataset of the electric vehicle to be detected and the multidimensional sensor dataset of multiple labeled electric vehicles, similar labeled electric vehicles are identified. Based on similar calibrated electric vehicles, braking performance evaluation results are generated.

[0057] Specifically, a pre-established calibrated vehicle database is created. Each calibrated vehicle possesses a complete multi-dimensional sensor dataset and corresponding braking performance evaluation results. This calibration data covers braking response characteristics under different vehicle models, braking system configurations, and mileages, constituting a standard answer set for braking performance evaluation. When the multi-dimensional sensor data of the vehicle to be tested is input, the database first retrieves several calibrated vehicles whose sensor features are most similar to those of the vehicle to be tested. The similarity is determined based on the overall matching degree of multi-dimensional sensor data such as vibration sequences and temperature sequences. Subsequently, based on the braking performance evaluation results of these similar calibrated vehicles, the braking performance of the vehicle to be tested is inferred. For example, braking performance evaluation results for the electric vehicle to be tested are generated through weighted averaging or nearest neighbor voting.

[0058] Figure 3 This is a schematic diagram of a multi-dimensional sensing-based electric vehicle braking performance detection system according to some embodiments of this specification, such as... Figure 3 As shown, a multi-dimensional sensing-based electric vehicle braking performance testing system may include: The detection optimization module is used to determine multiple braking vibration detection components and braking temperature detection components through dynamic twin group and parameter adaptive bat algorithm; The detection optimization module is also used to determine multiple braking performance testing scenarios based on multiple braking vibration detection components and braking temperature detection components; The data acquisition module is used to collect multidimensional sensor datasets of the electric vehicle to be tested based on multiple braking vibration detection components, braking temperature detection components, and braking performance testing scenarios. The multidimensional sensor datasets include vibration sequences of multiple braking vibration detection components, temperature sequences of multiple braking temperature detection components, and braking distances under each braking performance testing scenario. The performance evaluation module is used to generate braking performance evaluation results based on the multidimensional sensor dataset of the electric vehicle to be tested.

[0059] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for detecting the braking performance of electric vehicles based on multi-dimensional sensing, characterized in that, include: Multiple braking vibration detection components and temperature detection components were determined using a dynamic twin group and parameter-adaptive bat algorithm. Based on multiple braking vibration detection components and braking temperature detection components, multiple braking performance testing scenarios are determined; Based on multiple braking vibration detection components, braking temperature detection components, and braking performance testing scenarios, a multidimensional sensing dataset of the electric vehicle to be tested is collected. The multidimensional sensing dataset includes vibration sequences of multiple braking vibration detection components, temperature sequences of multiple braking temperature detection components, and braking distance under each braking performance testing scenario. Based on the multidimensional sensor dataset of the electric vehicle to be tested, braking performance evaluation results are generated.

2. The method for detecting the braking performance of electric vehicles based on multi-dimensional sensing according to claim 1, characterized in that, Multiple braking vibration detection components and braking temperature detection components were identified using a dynamic twin group and parameter-adaptive bat algorithm, including: Identify multiple vibration detection components and temperature detection components; A first test dataset is collected from multiple test vehicles under a calibrated braking performance testing scenario. The first test dataset includes vibration sequences from multiple vibration detection components and temperature sequences from multiple temperature detection components. Using a dynamic twin group and parameter-adaptive bat algorithm, multiple braking vibration detection components and braking temperature detection components are identified based on the first test dataset of multiple test vehicles in a calibrated braking performance testing scenario.

3. The method for detecting the braking performance of electric vehicles based on multi-dimensional sensing according to claim 2, characterized in that, Using a dynamic twin-subgroup and parameter-adaptive bat algorithm, based on the first test dataset of multiple test vehicles in a calibrated braking performance testing scenario, multiple braking vibration detection components and braking temperature detection components are identified, including: S11. Based on the first test dataset of multiple test vehicles under the calibrated braking performance test scenario, initialize the first population and the second population, wherein each bat in the first population and the second population represents a combination of at least one vibration detection component and at least one temperature detection component. S12. Construct the fitness function; S13. Iteratively optimize the first and second populations until the first termination condition is met. For each bat in the first population, in each iteration, the adaptive frequency and adaptive pulse rate of the bat are calculated by a clustering algorithm. For each bat in the second population, in each iteration, the bats to be swapped are determined, and the bats are updated based on the swapped bats. S14. Determine whether the first group and the second group satisfy the second termination condition. If yes, determine multiple braking vibration detection components and braking temperature detection components based on the current global optimal solution of the first group and the second group. If no, execute S15. S15. Exchange information between the first and second populations, and execute S13.

4. The method for detecting the braking performance of electric vehicles based on multi-dimensional sensing according to claim 3, characterized in that, Based on the first test dataset of multiple test vehicles under a calibrated braking performance testing scenario, the first and second populations are initialized, including: For any two vibration detection components, the response similarity between the two vibration detection components is calculated based on the first test dataset of multiple test vehicles under the calibrated braking performance test scenario. For any two temperature detection components, the response similarity between the two temperature detection components is calculated based on the first test dataset of multiple test vehicles under the calibrated braking performance test scenario. The first and second populations are initialized based on the response similarity of any two vibration detection components and any two temperature detection components.

5. The method for detecting the braking performance of electric vehicles based on multi-dimensional sensing according to claim 3, characterized in that, For each bat in the first group, in each iteration, the adaptive frequency and adaptive pulse rate of the bat are calculated using a clustering algorithm, including: Calculate the similarity between any two bats; Using a clustering algorithm, the bats in the first group are clustered based on the similarity between any two bats; For each bat in the first group, the adaptive frequency and adaptive pulse rate of the bat are calculated based on the number of bats in the cluster to which the bat belongs.

6. The method for detecting the braking performance of electric vehicles based on multi-dimensional sensing according to any one of claims 1-5, characterized in that, Based on multiple braking vibration detection components and braking temperature detection components, several braking performance testing scenarios are identified, including: Based on multiple braking vibration detection components and braking temperature detection components, a second test dataset is collected for multiple test vehicles under multiple performance testing scenarios. The second test dataset includes vibration sequences from multiple vibration detection components, temperature sequences from braking temperature detection components, and braking distances. Based on a second test dataset of multiple test vehicles in multiple performance testing scenarios, the effective value for each performance testing scenario is calculated. Based on the effective values ​​of each performance testing scenario, multiple braking performance testing scenarios are determined.

7. The method for detecting the braking performance of electric vehicles based on multi-dimensional sensing according to claim 6, characterized in that, Based on a second test dataset consisting of multiple test vehicles under multiple performance testing scenarios, valid values ​​for each performance testing scenario are calculated, including: For each braking performance test scenario, based on the second test dataset of multiple test vehicles in the performance test scenario, the response correlation coefficient corresponding to the braking performance test scenario is calculated, and based on the response correlation coefficient corresponding to the braking performance test scenario, the effective value of each performance test scenario is calculated.

8. The method for detecting the braking performance of electric vehicles based on multi-dimensional sensing according to claim 7, characterized in that, Based on a second test dataset of multiple test vehicles in performance testing scenarios, the response correlation coefficients corresponding to the braking performance testing scenarios are calculated, including: For any two test vehicles, calculate the difference in braking distance, similarity of vibration response, and similarity of temperature response for the corresponding performance testing scenarios of the two test vehicles; Based on the braking distance difference, vibration response similarity, and temperature response similarity of any two test vehicles corresponding to the performance testing scenarios, calculate the vibration response correlation coefficient and temperature response correlation coefficient. Based on the vibration response correlation coefficient and the temperature response correlation coefficient, the response correlation coefficient corresponding to the braking performance testing scenario is calculated.

9. The method for detecting the braking performance of an electric vehicle based on multidimensional sensing according to any one of claims 1-5, characterized in that, Based on the multidimensional sensor dataset of the electric vehicle to be tested, braking performance evaluation results are generated, including: Acquire multiple calibrated multidimensional sensor datasets of electric vehicles and braking performance evaluation results; Based on the multidimensional sensor dataset of the electric vehicle to be detected and the multidimensional sensor dataset of multiple labeled electric vehicles, similar labeled electric vehicles are identified. Based on similar calibrated electric vehicles, braking performance evaluation results are generated.

10. A multi-dimensional sensing-based electric vehicle braking performance testing system, characterized in that, The method for detecting the braking performance of an electric vehicle based on multidimensional sensing, according to any one of claims 1-9, includes: The detection optimization module is used to determine multiple braking vibration detection components and braking temperature detection components through dynamic twin group and parameter adaptive bat algorithm; The detection optimization module is also used to determine multiple braking performance detection scenarios based on multiple braking vibration detection components and braking temperature detection components; The data acquisition module is used to acquire a multi-dimensional sensor dataset of the electric vehicle to be tested based on multiple braking vibration detection components, braking temperature detection components, and braking performance testing scenarios. The multi-dimensional sensor dataset includes the vibration sequence of multiple braking vibration detection components, the temperature sequence of multiple braking temperature detection components, and the braking distance under each braking performance testing scenario. The performance evaluation module is used to generate braking performance evaluation results based on the multidimensional sensor dataset of the electric vehicle to be tested.

Images