Heavy truck accessory vibration evaluation method and system, terminal and medium

Abstract

The invention provides a heavy truck accessory vibration evaluation method and system, a terminal and a medium, and belongs to the technical field of commercial vehicle NVH. The method comprises the steps that vibration signals of accessories such as a fender and a bumper are collected and preprocessed under various preset working conditions, and a vibration acceleration root-mean-square value, a kurtosis value, a peak frequency and a speed root-mean-square value are calculated to serve as objective evaluation indexes; a subjective and objective consistency model is trained based on a historical data set, the historical data set is composed of single-working-condition objective indexes and corresponding single-working-condition subjective scores, and the model learns complex nonlinear mapping between the indexes by adopting a BP neural network; and for a to-be-evaluated accessory, inputting each working condition objective index into the trained model to obtain a prediction score, finally carrying out weighted calculation in combination with the working condition weight, and outputting a final vibration evaluation result. According to the invention, the efficiency and accuracy of counterweight truck accessory vibration evaluation are improved.

Description

Technical Field

[0001] This invention relates to the field of NVH technology for commercial vehicles, specifically to a method, system, terminal, and medium for evaluating the vibration of heavy-duty truck accessories. Background Technology

[0002] The vibration characteristics of the cab and body accessories of heavy-duty trucks, such as fenders, bumpers, rearview mirrors, and foot pedals, directly affect the vehicle's noise, vibration, and harshness (NVH) performance, thus impacting the driver's long-term driving comfort and fatigue levels. During vehicle operation, these accessories are subjected to complex vibrations from various excitation sources, including the road surface, engine, and transmission system. Especially under different operating conditions (such as idling, acceleration, constant speed, and braking), the vibration modes and energy distribution of these accessories undergo significant changes.

[0003] Currently, vibration evaluation methods for heavy-duty truck accessories have significant shortcomings. First, most methods rely on testing under single operating conditions or evaluating a single physical index, making it difficult to comprehensively and accurately reflect the overall vibration performance of accessories under complex and variable actual driving conditions. Second, the evaluation process often emphasizes the measurement of objective physical quantities while neglecting the driver's actual subjective feelings, leading to a discrepancy between objective test data and human comfort experience. This makes it difficult to directly use the evaluation results to guide design optimization aimed at improving user experience. Furthermore, the lack of a systematic method to scientifically correlate multi-condition, multi-dimensional objective indicators with subjective feelings results in a lack of unified and quantitative standards for evaluating the vibration comfort of accessories, making it difficult to objectively compare the vibration levels of different vehicle models or design schemes. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides a method, system, terminal, and medium for evaluating the vibration of heavy-duty truck accessories. Through systematic testing under multiple operating conditions and multiple indicators, and by utilizing a trained subjective-objective consistency model for intelligent prediction, the efficiency and accuracy of evaluating the vibration of heavy-duty truck accessories are improved.

[0005] The technical solution of the present invention provides a vibration evaluation method for heavy truck accessories, comprising the following steps: Vibration acceleration signals of target accessories are collected under several preset working conditions, and the collected vibration acceleration signals are preprocessed. Based on the preprocessed vibration acceleration signal, several objective evaluation indicators are calculated. Obtain subjective assessment scores of the vibration levels of the corresponding target accessories by evaluators under the corresponding single preset working conditions; A model that maintains consistency between subjective and objective evaluation is trained based on a historical dataset. Each sample in the historical dataset includes an objective evaluation index obtained under a preset working condition and a subjective feeling score corresponding to that preset working condition. The model outputs a predicted vibration score of the surrounding structures under that working condition based on the input objective evaluation index. For the attachment to be evaluated, the objective evaluation indicators calculated under several preset working conditions are input into the trained objective consistency model to obtain the predicted score for each working condition. Based on the preset weight coefficients of each working condition, the predicted scores for each working condition are weighted and calculated to obtain the final vibration evaluation result of the attachment.

[0006] As can be seen from the above technical solutions, this application has the following advantages: By systematically covering various typical working conditions (such as starting, idling, acceleration, constant speed, and braking), and comprehensively evaluating objective indicators such as the root mean square value of vibration acceleration, kurtosis value, peak frequency, and root mean square value of velocity, it can comprehensively and accurately characterize the vibration characteristics of the accessory, overcoming the limitations of single working condition and single indicator evaluation; by introducing a subjective-objective consistency model, through training and learning the complex nonlinear relationship between objective indicators and subjective scores for single working conditions in historical data, this model can automatically predict scores that conform to human subjective feelings based on objective test data, effectively integrating physical data and human experience, making the evaluation results more in line with actual driving experience; once the model training is complete, new accessories can be quickly evaluated, reducing the reliance on repetitive and large-scale subjective evaluations and improving evaluation efficiency. Attached Figure Description

[0007] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram of a vibration evaluation method for heavy truck accessories provided in an embodiment of the present invention.

[0009] Figure 2 This is a schematic block diagram of a vibration evaluation system for heavy truck accessories provided in an embodiment of the present invention.

[0010] Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0011] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0013] Figure 1 This is a schematic flowchart of a vibration evaluation method for heavy-duty truck accessories provided in an embodiment of the present invention. Figure 1 The executing entity can be a heavy-duty truck accessory vibration evaluation system. The heavy-duty truck accessory vibration evaluation method provided in this embodiment of the invention is executed by a computer device; correspondingly, the heavy-duty truck accessory vibration evaluation system runs on the computer device. Depending on different requirements, the order of the steps in this flowchart can be changed, and some steps can be omitted.

[0014] like Figure 1 As shown, the method includes the following steps.

[0015] S1, under several preset working conditions, collect vibration acceleration signals of the target attachment and preprocess the collected vibration acceleration signals.

[0016] In some alternative implementations, the target accessories include left and right fenders, left and right bumpers, left and right foot pedals, left and right rearview mirrors, and left and right taillight brackets. Preset operating conditions include at least two of the following: start / stop condition, idle condition, idle with air conditioning on condition, full throttle acceleration condition, constant speed condition, and braking condition.

[0017] Specifically, the selection of target accessories is based on key components that exhibit significant vibration response during the operation of the heavy-duty truck, directly affect the subjective experience of the driver and passengers, and are strongly related to the vehicle's appearance and functional safety. These include, but are not limited to: Left and right fenders: As key coverings on the sides of heavy trucks, their vibrations are easily observed visually and transmitted through the vehicle body, affecting the driving experience. Furthermore, long-term vibrations can easily cause the connection structure between the fender and the vehicle body to loosen. Therefore, they are included in the target accessories. Left and right bumpers: divided into front bumpers and rear bumpers. The front bumper is close to the engine and chassis vibration sources, while the rear bumper is significantly affected by the chassis transmission system and road surface excitation. The vibration of both not only affects the appearance stability, but may also be related to the functional reliability of the accessories on the bumper, and belongs to the core evaluation accessories. Left and right foot pedals: including the driver's foot pedals for getting in and out of the vehicle and the auxiliary foot pedals under the cab. The driver comes into direct contact with these pedals when getting in and out of the vehicle, and their vibrations are transmitted to the human body through the feet, which directly affects driving comfort. They are accessories with a very high degree of subjective feeling. Left and right rearview mirrors: As key components for drivers to observe road conditions to the side and rear of the vehicle, vibration of the rearview mirror housing and lens can cause blurred vision, affecting driving safety. Moreover, the amplitude and frequency of lens vibration are directly related to the driver's visual fatigue level, which needs to be evaluated in detail. Left and right rear taillight brackets: The rear taillights are fixed to the rear of the vehicle by brackets. Vibration of the brackets will be directly transmitted to the taillight body. Long-term high-frequency vibration can easily cause the taillight wiring to loosen and the lens to crack, affecting the lighting and signal indication functions. At the same time, the vibration of the brackets also reflects the vibration characteristics of the rear of the vehicle, which is necessary for evaluation.

[0018] The preset operating conditions are based on typical driving scenarios in the actual operation of heavy trucks, covering the entire life cycle of vehicle operation, including starting, idling, acceleration, constant speed, and braking, to ensure that the evaluation results can truly reflect the vibration characteristics of the accessories in actual use. The specific operating condition definitions and selection criteria are as follows.

[0019] (1) Start-up / stop operation The starting condition refers to the process of a heavy-duty truck engine starting from a shut-off state until it reaches a stable idle speed; the stopping condition refers to the process of the engine shutting off from a stable idle state until it completely stops rotating.

[0020] During the start / stop process, the rotational speed of rotating components such as the engine crankshaft and flywheel changes drastically, which will generate instantaneous impact vibration. This vibration is transmitted to various accessories through the engine mounts and chassis frame, which can easily lead to instantaneous stress concentration in the accessory connection structure. Moreover, the driver is highly sensitive to the vibration during the start / stop phase, so it is included in the preset operating conditions.

[0021] (2) Idle condition This refers to the stable operation of a heavy-duty truck engine under no-load conditions. The specific conditions are: the engine speed is maintained within the manufacturer-specified idle speed range, the vehicle is parked, and there is no additional load.

[0022] Idling is a common operating condition for heavy trucks, such as waiting to load or unload goods or stopping at traffic lights. Under this condition, the periodic combustion vibration of the engine is transmitted to various accessories through the chassis, forming continuous low-frequency vibration. Long-term idling vibration can easily lead to fatigue damage of accessories. Moreover, the driver is stationary during idling, so the vibration is more noticeable. This is one of the key operating conditions for evaluating the vibration comfort of accessories.

[0023] (3) Idle speed with air conditioning on Based on the above idling conditions, when the vehicle's air conditioning system is turned on, the engine speed may fluctuate slightly due to the increased air conditioning load; the other conditions are the same as the idling conditions.

[0024] When the air conditioning system is working, the periodic operation of the compressor will generate additional vibration excitation. This excitation, when superimposed with the engine idling vibration, will change the vibration characteristics of the accessories. Especially in summer or winter, heavy trucks use the air conditioning while idling very frequently. The vibration performance of the accessories under this condition directly affects the driver's comfort, so it needs to be set as a separate preset condition.

[0025] (4) Full throttle acceleration condition This refers to the acceleration process of a heavy truck, starting from a certain stable speed and pressing the accelerator pedal to the maximum opening until the speed reaches 80km / h. During acceleration, the gears are automatically or manually switched according to the vehicle's power system to ensure that the engine is in the effective power output range.

[0026] Under full throttle acceleration, the engine output torque increases sharply, and the load on the transmission system increases significantly, which easily generates torsional and impact vibrations. At the same time, the road excitation and power system vibration are superimposed during vehicle operation, which leads to complex multi-frequency vibrations in the accessories. The vibration level of the accessories under this condition directly reflects the matching degree between the vehicle's power performance and vibration control level, and has important evaluation significance.

[0027] (5) Uniform speed condition The evaluation speed refers to the condition in which a heavy truck travels at a constant speed on a flat road. Typical speeds within the commonly used driving speed range of heavy trucks are selected as the evaluation speeds, specifically including three sub-conditions: 40 km / h (common speed on urban roads), 60 km / h (common speed on national highways), and 80 km / h (common speed on expressways). The vehicle travel time in each sub-condition is no less than 120 seconds to ensure the stability of vibration signal acquisition.

[0028] The constant speed condition is the main condition in the long-distance transportation of heavy trucks, accounting for more than 60% of the total driving time. Under this condition, the vibration of the vehicle mainly comes from road excitation and steady-state vibration of the engine and transmission system. The vibration of the accessories under this condition has the characteristics of continuity and stability. Its vibration level directly determines the driver's comfort during long-distance driving and is one of the conditions with the highest weight among the preset conditions.

[0029] (6) Braking conditions This refers to the process from when a heavy truck starts traveling at a stable and constant speed, to when the brake pedal is pressed and the vehicle comes to a complete stop. During the braking process, the opening of the brake pedal is kept constant to avoid extreme impacts caused by sudden braking.

[0030] During braking, the friction between the brake pads and the brake disc generates vibration excitation. This excitation is transmitted to the frame and various accessories through the brake lines and axle. At the same time, the forward shift of the vehicle's center of gravity increases the load on the front axle, and the vibration response of the front accessories will change significantly. Since braking is a necessary condition for vehicle operation, the vibration performance of the accessories under this condition is related to driving stability and comfort during braking. Therefore, it is included in the preset conditions.

[0031] For the aforementioned target accessories and preset working conditions, high-precision triaxial accelerometers are selected. The sensors are positioned on the flat surface of the connecting brackets between each accessory and the vehicle frame, as follows: Left and right fenders: The sensors are attached to the brackets on the inside of the fenders that connect to the longitudinal beams of the vehicle body, with one sensor on each fender; Left and right bumpers: The front bumper sensor is attached to the bracket connecting the bumper crossbeam and the front frame, and the rear bumper sensor is attached to the connection point between the bumper bracket and the rear frame. One sensor is installed on each bumper. Left and right foot pedals: The sensors are attached to the connection between the foot pedal bracket and the vehicle body pillar, with one sensor installed on each foot pedal; Left and right rearview mirrors: The sensors are attached to the connecting flange surface between the rearview mirror base and the door pillar, with one sensor installed in each rearview mirror; Left and right rear taillight brackets: The sensor is attached to the connection position between the taillight bracket and the rear beam of the vehicle body, with one sensor installed in each taillight bracket.

[0032] The sampling frequency of the data acquisition device is set to 2000Hz, and the sampling duration is set according to the operating conditions: the sampling duration for start / stop conditions is 60s (covering 30s for start + 30s for stop), the sampling duration for idling conditions and idling with the air conditioning on is 120s, the sampling duration for full throttle acceleration conditions is determined according to the acceleration time, the sampling duration for each sub-condition of constant speed conditions is 120s, and the sampling duration for braking conditions is 30s (covering the entire process from the start of braking to the vehicle stopping).

[0033] The acquired vibration acceleration signals are preprocessed to eliminate environmental noise, interference signals, and data anomalies, and to extract effective vibration information. This includes: using the 3σ criterion to detect and remove outliers from the acquired raw vibration acceleration signals; using the db4 wavelet basis function for wavelet transform denoising; performing EMD decomposition on the wavelet transform-denoised signal; and performing dynamic filtering, dynamically adjusting the filter parameters according to the frequency characteristics of the vibration signals under different preset working conditions.

[0034] For start-up / stop conditions, the signal contains low-frequency components (0.5-10Hz) and sudden impulse signals. A combination of a second-order Butterworth high-pass filter (cutoff frequency 0.5Hz) and a second-order Butterworth low-pass filter (cutoff frequency 50Hz) is used to retain effective vibration components and suppress low-frequency environmental noise and high-frequency electromagnetic interference.

[0035] For idling and idling with the air conditioning on, the signal is mainly low-frequency noise (5-50Hz, mainly engine vibration) with fewer high-frequency components. A combination of a second-order Butterworth high-pass filter (cutoff frequency 5Hz) and a second-order Butterworth low-pass filter (cutoff frequency 100Hz) is used to filter out low-frequency ground vibration and high-frequency unrelated noise.

[0036] For constant speed operation, the signal is relatively smooth with less high-frequency noise (100-200Hz). A combination of a second-order Butterworth high-pass filter (cutoff frequency 1Hz) and a second-order Butterworth low-pass filter (cutoff frequency 200Hz) is used to balance low-frequency vibration and useful mid-to-high frequency signals.

[0037] For full-throttle acceleration and braking conditions: there are many high-frequency components (100-500Hz) in the signal. A combination of a 4th-order Butterworth high-pass filter (cutoff frequency 5Hz) and a 4th-order Butterworth low-pass filter (cutoff frequency 500Hz) is used to ensure that the high-frequency impact vibration signal is not lost.

[0038] The filtered vibration acceleration signal is standardized to normalize the signal amplitude to the range of [-1,1]. The standardized signal can eliminate the influence of the difference in signal amplitude under different accessories and working conditions on the subsequent index calculation, thus ensuring the consistency of the evaluation.

[0039] S2, based on the preprocessed vibration acceleration signal, calculate several objective evaluation indicators.

[0040] Based on the preprocessed vibration acceleration signal, four core objective evaluation indicators that can comprehensively characterize the vibration energy, impact characteristics, dominant excitation source, and human perception correlation of heavy truck accessories are selected, namely, root mean square value of vibration acceleration (Q), vibration kurtosis (W), vibration peak frequency (E), and root mean square value of vibration velocity (V).

[0041] (1) Root mean square value of vibration acceleration (Q) The root mean square value (Q) of vibration acceleration is an important indicator for characterizing the effective value of the vibration signal of an accessory. It reflects the energy accumulation level of the vibration acceleration signal of the accessory in the frequency range of 0-200Hz. Its value is directly related to the overall intensity of the accessory vibration. The larger the value, the stronger the vibration energy of the accessory and the more intense the vibration.

[0042] Based on the preprocessed vibration acceleration time-domain signal The root mean square value of vibration acceleration (Q) is calculated through the following steps.

[0043] Step 1.1: Determine the calculation time window. Set the window length according to the duration of different preset operating conditions. The time window length for start / stop and braking conditions is consistent with the actual duration of the operating conditions. For example, the start / stop condition is set to 30-60s, the braking condition is set to 10-20s, and the time window length for idling, idling with air conditioning on, and constant speed conditions is set to 60s to ensure that enough vibration cycles are covered and to reduce random errors.

[0044] Step 1.2: Analyze the vibration acceleration time-domain signal within the time window. Discretization is performed, and the discretization sampling interval is set to 1. The discretized signal sequence is obtained. ;in, , where is the number of discretized data points.

[0045] Step 1.3: Calculate Q using the root mean square value calculation formula, as follows:

[0046] The root mean square value of vibration acceleration can comprehensively reflect the energy distribution of vibration signals, avoiding the one-sided influence of instantaneous peak values ​​on evaluation results. Moreover, this index is directly related to the structural fatigue life of heavy truck accessories. At the same time, this index is a classic indicator for evaluating vibration intensity in the NVH field, and can provide a unified quantitative benchmark for comparing the vibration levels of different vehicle models and accessories.

[0047] (2) Vibration kurtosis value (W) The vibration kurtosis (W) is a statistical index characterizing the intensity of the impact or transient components in an accessory vibration signal. It reflects the "steepness" of the probability density distribution of the vibration acceleration signal in the 0-200Hz frequency range. When an impact pulse is present in the signal, the kurtosis value increases significantly; if the signal is a stationary sinusoidal or random vibration, the kurtosis value is close to 3 (the kurtosis value of a normally distributed signal is 3).

[0048] Based on the preprocessed time-domain discrete signal sequence of vibration acceleration Calculate the mean value of the vibration acceleration signal. and the standard deviation of the vibration acceleration signal Furthermore, the fourth central moment of the vibration acceleration signal Calculate the vibration kurtosis value according to the kurtosis value calculation formula. .

[0049] Heavy truck accessories are susceptible to transient impact vibrations under conditions such as full-throttle acceleration and braking. Although these impact vibrations are short in duration, they can easily lead to malfunctions such as loosening of accessory connecting bolts and cracking of plastic parts. The root mean square value of vibration acceleration cannot effectively identify such impact components. Vibration kurtosis value can accurately capture transient impacts in the signal, supplement the characterization of the impact characteristics of accessory vibration, complement the root mean square value, and comprehensively reflect the complex characteristics of accessory vibration.

[0050] (3) Peak vibration frequency (E) Peak frequency (E) is a key indicator characterizing the dominant excitation source or natural frequency of accessory vibration. It refers to the frequency value corresponding to the frequency component with the largest amplitude in the vibration acceleration spectrum obtained through spectral analysis within the 0-200Hz frequency range. This indicator is directly related to the root cause of accessory vibration. If the peak frequency coincides with a certain harmonic frequency of the engine, it indicates that the accessory vibration is mainly caused by engine excitation; if the peak frequency coincides with the accessory's own natural frequency, there is a risk of resonance.

[0051] Based on the preprocessed vibration acceleration time-domain signal, a Fast Fourier Transform (FFT) is used for spectrum analysis, and the peak frequency of vibration is calculated by combining the peak identification algorithm. The calculation steps are as follows.

[0052] Step 3.1, setting spectrum analysis parameters, including frequency resolution. and FFT points .

[0053] Step 3.2, discrete time-domain signal of vibration acceleration Perform an FFT transform to obtain the spectrum in complex form, and calculate each frequency point based on the real and imaginary parts of this spectrum. vibration acceleration amplitude .

[0054] Step 3.3, determine the frequency axis, the first The actual frequency corresponding to each frequency point The calculation formula is as follows:

[0055] The sampling frequency.

[0056] Step 3.4: Filter within the 0-200Hz frequency range. The highest frequency point corresponds to , which is the peak frequency of vibration, E.

[0057] Vibration peak frequency is strongly correlated with human comfort perception. According to human vibration physiology research, the human body is most sensitive to vibrations of 5-15Hz (vertical direction) and 20-50Hz (horizontal direction). If the peak frequency of accessory vibration falls within this sensitive range, even if the root mean square value of vibration acceleration is small, it will cause obvious discomfort to the driver. At the same time, peak frequency can provide direction for accessory design optimization and has important guiding significance.

[0058] (4) Root mean square value of vibration velocity (V) The root mean square (RMS) value of dynamic velocity (V) is a key indicator characterizing the low-frequency vibration characteristics of an accessory, reflecting the energy accumulation level of the accessory's vibration velocity signal within the 0-200Hz frequency range. Compared to vibration acceleration, vibration velocity is more consistent with the human body's subjective perception of low-frequency vibration, and the human body's perception of low-frequency vibration depends more on changes in vibration velocity.

[0059] The vibration velocity signal is obtained by integrating the preprocessed vibration acceleration signal, and then the root mean square value is calculated based on the velocity signal. The specific steps are as follows.

[0060] Step 4.1: Integrate the acceleration signal to obtain the velocity signal, and then discretely analyze the vibration acceleration signal in the time domain. Numerical integration is performed to obtain the discrete vibration velocity signal. .

[0061] Step 4.2, speed signal smoothing processing, using a 5-point moving average filter. Smoothing is performed to eliminate high-frequency noise introduced during integration, resulting in a smoothed velocity signal. .

[0062] Step 4.3: Based on the smoothed velocity signal, calculate the root mean square value V of the vibration velocity according to the following formula:

[0063] In the low-frequency vibration range, the correlation between vibration acceleration and the rate of change of vibration displacement is weak, while vibration velocity can more directly reflect the dynamic change amplitude of accessory vibration. For example, the greater the low-frequency vibration velocity of the left and right rearview mirrors, the greater the swing amplitude of the mirrors, and the more severe the blurring of the driver's rearward vision.

[0064] To eliminate the impact of differences in the scale of different indicators on the subsequent training of the subjective-objective consistency model, the above four types of objective evaluation indicators are standardized, and the Min-Max standardization method is used to map the indicator values ​​to the [0,1] interval.

[0065] S3, obtain the subjective perception score of the vibration level of the corresponding target accessory by the evaluator under the corresponding single preset working condition.

[0066] To quantify vibration perception, subjective ratings were divided into 5 levels, each with a clear semantic interpretation and perceptual characteristics, as shown in the table below.

[0067] Table 1: Subjective Rating Level Classification

[0068] Evaluators need to select the corresponding scoring range and fill in the specific score based on the above rules and their own actual perception of the target attachment under a single preset working condition.

[0069] S4. A subjective-objective consistency model is trained based on a historical dataset. Each sample in the historical dataset includes an objective evaluation index obtained under a preset working condition and a subjective feeling score corresponding to that preset working condition. The model outputs a predicted vibration score of the surrounding area under that working condition based on the input objective evaluation index.

[0070] The objective-subject consistency model constructed in this embodiment aims to establish a mapping bridge from physical vibration signals to human subjective perception. In the vibration evaluation of heavy-duty truck accessories, the key problem this model needs to address is that vehicle vibration is broadband, multi-source, and non-stationary, while human perception is non-linear, fuzzy, and comprehensive. Traditional linear regression models struggle to capture this complex relationship. Therefore, a model architecture based on backpropagation neural networks (BPNN) is adopted, whose non-linear fitting capabilities are well-suited for this scenario.

[0071] This BP neural network model employs a hierarchical feedforward structure that includes an input layer, two hidden layers, and an output layer.

[0072] The input layer receives a standardized feature vector composed of input features and distributes the feature vector to the hidden layer. It consists of 5 nodes, each corresponding to an input feature: root mean square value of vibration acceleration, vibration kurtosis value, peak frequency of vibration, root mean square value of vibration velocity, and working condition type code.

[0073] The first hidden layer has more nodes than the second hidden layer, and both hidden layers use the ReLU activation function. The first hidden layer learns and constructs composite features from several input features, while the second hidden layer learns the weights and logical relationships between these composite features. Specifically, the first hidden layer has more nodes and is used for high-order feature extraction and preliminary mapping of complex nonlinear relationships; the second hidden layer has slightly fewer nodes and is used to further combine these high-order features to form more abstract "perceptual concepts".

[0074] The first hidden layer is responsible for learning and constructing complex composite features from the five input features. For example, it might form a "low-frequency high-energy impact" node, which strongly corresponds to the subjective evaluation of "very shaky, unacceptable".

[0075] The second hidden layer learns the weights and logical relationships between these composite features based on the output of the first layer. For example, it might learn that: "Under acceleration conditions, the 'high frequency, medium energy' feature has a lower weight in terms of its impact on the score, while under idling conditions, the 'low frequency, medium energy' feature has a very high negative impact weight." The activation function of the output layer is either the Sigmoid function or a linear function, which maps the output of the hidden layer to the predicted score.

[0076] The subjective-objective consistency model is trained based on historical datasets, which includes the following steps.

[0077] S4.1, Extract from historical database A set of valid training samples constitutes the training set. ;in, For the first The input feature vector of each sample, This represents the root mean square value of vibration acceleration. Indicates the vibration kurtosis value. Indicates the peak frequency of vibration. This represents the root mean square value of the vibration velocity. This is the operating condition encoding vector, which uses one-hot encoding to represent the operating condition to which this set of data belongs; The target output value is the value obtained by comparing the input feature vector with the target output value. Subjective vibration perception score of the accessory given by a professional evaluation team under the same working condition. .

[0078] S4.2, input feature vector The input is fed into the BP neural network model, and the input layer receives... The 3D feature vector, the hidden layer performs a nonlinear transformation on the input data, the 4D feature vector, the 5D feature vector, the 6D feature vector, the 7D feature vector, the 8D feature vector, the 9D feature vector, the 10 ... Output of hidden layer Calculated using the following formula:

[0079] in, This is the output of the previous layer. For the first The weight matrix of the layer, For the first The layer's bias vector, For the first The activation function of the layer.

[0080] S4.3, the output layer receives the output of the last hidden layer. The final predicted score is generated using a linear activation function. :

[0081] in, The weight vector of the output layer. This is the bias constant for the output layer.

[0082] S4.4, through the loss function Quantitative model predictions Compared with true subjective ratings The differences between them; The loss function is calculated using the backpropagation algorithm. Relative to the parameters of each layer of the model , , , The gradient is calculated, and the gradient descent optimization algorithm is used to update all parameters of the model based on the calculated gradient. For any parameter... Its update formula is: in, This is the learning rate.

[0083] S4.5 Repeat steps 2 to 4 to traverse multiple batches or periods in the training dataset until the preset termination condition is met.

[0084] The termination condition is any of the following: the loss function value drops below a preset threshold, the model's performance on the independent validation dataset no longer improves significantly, or the preset maximum number of training iterations is reached.

[0085] Through the above training process, a model with a fixed network structure and optimal parameter set that maintains consistency between objective and subjective vibration is finally obtained. This model can accurately capture the complex nonlinear mapping relationship between objective vibration indicators and human subjective comfort perception.

[0086] Preferably, the loss function is constructed to incorporate prior knowledge or constraints during model training, including a weighted term based on the importance of the working condition, a focus reinforcement term based on the key interval of the score, or an uncertainty adaptive term based on the clustering of objective data.

[0087] Option 1 incorporates a weighted factor based on the importance of operating conditions.

[0088] The importance difference of different working conditions is introduced during the model training stage to ensure that the model does not learn equally on all working conditions, but focuses more on learning those working conditions that occur more frequently and have a greater impact on driver comfort.

[0089] The weighted mean squared error loss function is defined as follows:

[0090] In the formula, This represents the batch sample size. This is the working condition weighting function, whose value is based on the first... The working condition to which each sample belongs For example: idling condition Uniform speed condition Accelerated operating conditions Braking conditions .

[0091] This loss function embeds prior knowledge of engineering load conditions into the optimization objective. During the model training phase, it guides the model parameters to be updated towards "more accurately fitting high-weight load conditions," thus the trained model naturally has higher prediction accuracy and reliability for important load conditions.

[0092] Option 2 incorporates focus reinforcement items based on key scoring intervals.

[0093] This approach can enhance the model's predictive ability near key scoring intervals (especially the pass / fail boundary) and reduce serious errors such as misclassifying "some jitter" as "slight jitter".

[0094] The boundary reinforcement loss function is defined as follows:

[0095] In the formula, The boundary reinforcement coefficient controls the intensity of attention given to key regions. The center point of the critical interval, To control the width of the region of interest.

[0096] This loss function is obtained through a... A Gaussian kernel centered on the sample is used to dynamically adjust the loss. This is based on the true score of the sample. Near the critical boundary When the Gaussian term increases, it amplifies the loss contribution of that sample. This forces the model to devote more "attention" to learning complex patterns near the boundary, significantly improving the model's discriminative ability and robustness in the most sensitive areas of engineering decision-making.

[0097] Option 3 incorporates an adaptive uncertainty term based on the clustering degree of objective data.

[0098] Subjective ratings inherently possess ambiguity, and different evaluators may give reasonable scores for the same vibration state. To address this issue and make the model training process more intelligent and robust, this solution introduces an uncertainty-adaptive loss function based on the degree of objective data clustering.

[0099] The uncertainty in model learning primarily stems from the ambiguity of the mapping relationships within the objective feature space. If a sample's objective features exhibit high consistency within its local neighborhood, and the corresponding subjective scores are also similar, this indicates a clear "subjective-objective mapping" relationship in that region, and the model should give this sample higher attention and continue learning it. Conversely, if a sample's objective features correspond to highly dispersed subjective scores within its local neighborhood, this indicates a complex or contradictory mapping relationship in that region, and the model should adopt a more conservative approach during learning, reducing its learning weight.

[0100] The adaptive loss function is defined as follows:

[0101] In the formula, For the first The objective characteristic uncertainty measure of a sample is calculated as follows: in the feature space of the objective evaluation indicators of the historical dataset, find the value that is consistent with... Find the nearest neighbor samples and calculate the standard deviation of the subjective feeling scores corresponding to these nearest neighbor samples; The weight function is monotonically decreasing, for example Used to measure uncertainty Mapped to loss weights.

[0102] This loss function introduces "data uncertainty." Instead of being static, it dynamically adjusts its learning weights based on the degree of subjective rating divergence caused by the objective characteristics of each training sample within its local neighborhood. For samples with consistent objective characteristics and subjective ratings (low uncertainty), the model focuses on learning them; for samples with dispersed objective characteristics and even identical subjective ratings (high uncertainty), the model's learning is relatively conservative. This makes the training process more intelligent and robust, better handling the inherent ambiguity in subjective evaluations.

[0103] S5. For the attachment to be evaluated, the objective evaluation indicators calculated under several preset working conditions are input into the trained objective consistency model to obtain the predicted score for each working condition. Based on the preset weight coefficients of each working condition, the predicted scores for each working condition are weighted and calculated to obtain the final vibration evaluation result of the attachment.

[0104] When it is necessary to evaluate the vibration comfort of a new heavy truck or its accessory design variant, firstly, following the same specifications as in the model training phase, sensors are placed on target accessories such as the left and right fenders, left and right bumpers, left and right foot pedals, left and right rearview mirrors, and left and right taillight brackets. Under preset conditions such as start / stop, idling, constant speed, full throttle acceleration, and braking, vibration acceleration signals are collected, and a set of objective evaluation indicators are calculated. Subsequently, the objective evaluation index calculated for each working condition, together with the one-hot encoded identifier of that working condition, forms a feature vector, which is then input into the trained and validated subjective-objective consistency model. The model performs forward propagation calculations using its internally stored weights and bias parameters, instantly outputting the predicted score of the attachment for the corresponding single working condition, denoted as... For example, the predicted scores for idling conditions were obtained respectively. Uniform speed working condition prediction score Accelerated operating condition prediction score and braking condition prediction score .

[0105] To obtain a single index that reflects the overall vibration level of the accessories throughout a complete driving cycle, the predicted scores for each of the above-mentioned operating conditions need to be weighted and fused. This embodiment pre-sets weighting coefficients to reflect the differences in the contribution of different operating conditions to the driver's overall comfort perception. The weighting calculation follows the formula below:

[0106] In a preferred embodiment, the weighting coefficient is set as: idling condition weighting. =0.3, weight for uniform speed condition =0.4, weight of full-throttle acceleration condition =0.2, braking condition weight =0.1. This weighting reflects the actual driving characteristics, with constant speed driving time accounting for the largest proportion, followed by idling, and acceleration and braking transient conditions accounting for a smaller proportion.

[0107] Final calculation This is the final vibration evaluation result for this attachment. This result is a continuous value between 1 and 10, which can be directly mapped to the preset semantic evaluation level.

[0108] The above text provides a detailed description of an embodiment of a vibration evaluation method for heavy-duty truck accessories. Based on the vibration evaluation method for heavy-duty truck accessories described in the above embodiment, this invention also provides a vibration evaluation system for heavy-duty truck accessories corresponding to the method.

[0109] Figure 2This is a schematic block diagram of a heavy-duty truck accessory vibration evaluation system provided in an embodiment of the present invention. In this embodiment, the heavy-duty truck accessory vibration evaluation system 200 can be divided into multiple functional modules according to its functions, such as... Figure 2 As shown. The module referred to in this invention is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and which are stored in memory.

[0110] The signal acquisition and preprocessing module 210 is used to acquire vibration acceleration signals of the target attachment under several preset working conditions and to preprocess the acquired vibration acceleration signals.

[0111] The objective evaluation index calculation module 220 is used to calculate several objective evaluation indices based on the preprocessed vibration acceleration signal.

[0112] The subjective scoring calculation module 230 is used to obtain the subjective perception score of the vibration level of the corresponding target accessory by the evaluator under the corresponding single preset working condition.

[0113] The subjective-objective consistency model training module 240 is used to train a subjective-objective consistency model based on a historical dataset. Each sample in the historical dataset includes an objective evaluation index obtained under a preset working condition and a subjective feeling score corresponding to the preset working condition. The model outputs a predicted vibration score of the surrounding area under the working condition based on the input objective evaluation index.

[0114] The vibration evaluation module 250 is used to input the objective evaluation indicators calculated under several preset working conditions of the attachment to be evaluated into the trained objective consistency model to obtain the predicted score corresponding to each working condition; based on the preset weight coefficient of each working condition, the predicted score corresponding to each working condition is weighted and calculated to obtain the final vibration evaluation result of the attachment.

[0115] The heavy-duty truck accessory vibration evaluation system of this embodiment is used to implement the aforementioned heavy-duty truck accessory vibration evaluation method. Therefore, the specific implementation of this system can be found in the embodiment section of the heavy-duty truck accessory vibration evaluation method above. Thus, the specific implementation can be referred to the description of the corresponding embodiments, and will not be elaborated here.

[0116] Furthermore, since the heavy truck accessory vibration evaluation system in this embodiment is used to implement the aforementioned heavy truck accessory vibration evaluation method, its function corresponds to the function of the above method, and will not be repeated here.

[0117] Figure 3This is a schematic diagram of a terminal 300 provided in an embodiment of the present invention, including: a processor 310, a memory 320, and a communication unit 330. The processor 310 is used to implement the flow steps of the above-described embodiment of the heavy truck accessory vibration evaluation method when implementing the heavy truck accessory vibration evaluation program stored in the memory 320.

[0118] This invention also provides a computer storage medium, which may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. The computer storage medium stores a heavy-duty truck accessory vibration evaluation program. When the heavy-duty truck accessory vibration evaluation program is executed by a processor, it implements the process steps of the above-described embodiment of the heavy-duty truck accessory vibration evaluation method.

[0119] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A vibration evaluation method for heavy truck accessories, characterized in that, Includes the following steps: The vibration acceleration signals of the target attachments are collected under several preset working conditions, and the collected vibration acceleration signals are preprocessed. Based on the preprocessed vibration acceleration signal, several objective evaluation indicators are calculated. Obtain subjective assessment scores of the vibration levels of the corresponding target accessories by evaluators under the corresponding single preset working conditions; A model that maintains consistency between subjective and objective evaluation is trained based on a historical dataset. Each sample in the historical dataset includes an objective evaluation index obtained under a preset working condition and a subjective feeling score corresponding to that preset working condition. The model outputs a predicted vibration score of the surrounding structures under that working condition based on the input objective evaluation index. For the attachment to be evaluated, the objective evaluation indicators calculated under several preset working conditions are input into the trained objective consistency model to obtain the predicted score for each working condition. Based on the preset weight coefficients of each working condition, the predicted scores for each working condition are weighted and calculated to obtain the final vibration evaluation result of the attachment.

2. The vibration evaluation method for heavy truck accessories according to claim 1, characterized in that, Objective evaluation indicators include root mean square value of vibration acceleration, vibration kurtosis, peak frequency of vibration, and root mean square value of vibration velocity.

3. The vibration evaluation method for heavy truck accessories according to claim 2, characterized in that, The subjective-objective consistency model adopts a BP neural network model architecture, including an input layer, two hidden layers, and an output layer; The input layer receives a standardized feature vector composed of input features and distributes the feature vector to the hidden layer. The first hidden layer has more nodes than the second hidden layer, and the activation function of both hidden layers is the ReLU function; the first hidden layer is used to learn and construct composite features from several input features, and the second hidden layer is used to learn the weights and logical relationships between composite features; The activation function of the output layer is either the Sigmoid function or a linear function, which maps the output of the hidden layer to the predicted score.

4. The vibration evaluation method for heavy truck accessories according to claim 3, characterized in that, The subjective-objective consistency model is trained based on historical datasets and includes: Step 1: Extract from historical database A set of valid training samples constitutes the training set. ;in, For the first The input feature vector of each sample, This represents the root mean square value of vibration acceleration. Indicates the vibration kurtosis value. Indicates the peak frequency of vibration. This represents the root mean square value of the vibration velocity. This is the operating condition encoding vector, which uses one-hot encoding to represent the operating condition to which this set of data belongs; The target output value is the value obtained by comparing the input feature vector with the target output value. Subjective vibration perception score of the accessory given by a professional evaluation team under the same working condition. ; Step 2, input feature vector The input is fed into the BP neural network model, and the input layer receives... The 3D feature vector, the hidden layer performs a nonlinear transformation on the input data, the 4D feature vector, the 5D feature vector, the 6D feature vector, the 7D feature vector, the 8D feature vector, the 9D feature vector, the 10 ... Output of hidden layer Calculated using the following formula: in, This is the output of the previous layer. For the first The weight matrix of the layer, For the first The layer's bias vector, For the first Activation function of the layer; Step 3: The output layer receives the output of the last hidden layer. The final predicted score is generated using a linear activation function. : in, The weight vector of the output layer. The bias constant of the output layer; Step 4, using the loss function Quantitative model predictions Compared with true subjective ratings The differences between them; The loss function is calculated using the backpropagation algorithm. Relative to the parameters of each layer of the model , , , The gradient is calculated, and the gradient descent optimization algorithm is used to update all parameters of the model based on the calculated gradient. For any parameter... Its update formula is: in, The learning rate; Step 5: Repeat steps 2 to 4 to iterate through multiple batches or periods in the training dataset until the preset termination condition is met.

5. The vibration evaluation method for heavy truck accessories according to claim 4, characterized in that, The loss function is constructed to incorporate prior knowledge or constraints during model training, including a weighted term based on the importance of the working condition, a focus reinforcement term based on the key interval of the scoring, or an uncertainty adaptive term based on the clustering degree of objective data. When it incorporates an adaptive term for uncertainty based on the clustering degree of objective data, the loss function The expression is: in, For the first The objective characteristic uncertainty measure of a sample is calculated as follows: in the feature space of the objective evaluation indicators of the historical dataset, find the value that is consistent with... Find the nearest neighbor samples and calculate the standard deviation of the subjective feeling scores corresponding to these nearest neighbor samples; The weight function is monotonically decreasing.

6. The vibration evaluation method for heavy truck accessories according to any one of claims 1 to 5, characterized in that, The target accessories include left and right fenders, left and right bumpers, left and right foot pedals, left and right rearview mirrors, and left and right rear taillight brackets.

7. The vibration evaluation method for heavy truck accessories according to any one of claims 1 to 5, characterized in that, The preset operating conditions include at least two of the following: start / stop condition, idle condition, idle condition with air conditioning on, full throttle acceleration condition, constant speed condition, and braking condition.

8. A vibration evaluation system for heavy truck accessories, characterized in that, include: The signal acquisition and preprocessing module is used to acquire vibration acceleration signals of target accessories under several preset working conditions and to preprocess the acquired vibration acceleration signals. The objective evaluation index calculation module is used to calculate several objective evaluation indices based on the preprocessed vibration acceleration signal. The subjective scoring calculation module is used to obtain the subjective perception score of the vibration level of the corresponding target accessory by the evaluator under the corresponding single preset working condition; The subjective-objective consistency model training module is used to train a subjective-objective consistency model based on a historical dataset. Each sample in the historical dataset includes an objective evaluation index obtained under a preset working condition and a subjective feeling score corresponding to that preset working condition. The model outputs a predicted vibration score of the surrounding area under that working condition based on the input objective evaluation index. The vibration evaluation module is used to input the objective evaluation indicators calculated under several preset working conditions of the attachment to be evaluated into the trained objective consistency model to obtain the predicted score corresponding to each working condition; based on the preset weight coefficient of each working condition, the predicted score corresponding to each working condition is weighted and calculated to obtain the final vibration evaluation result of the attachment.

9. A terminal, characterized in that, include: The memory is used to store vibration evaluation programs for heavy truck accessories; A processor is configured to implement the steps of the heavy truck accessory vibration evaluation method as described in any one of claims 1 to 7 when executing the heavy truck accessory vibration evaluation program.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores a heavy truck accessory vibration evaluation program, which, when executed by a processor, implements the steps of the heavy truck accessory vibration evaluation method as described in any one of claims 1 to 7.

Images