Vehicle frame surface electrostatic dust collection system and method

Abstract

The invention relates to the technical field of electrostatic dust collection, and discloses a vehicle frame surface electrostatic dust collection system and method.The system comprises a data collecting and processing module, a feature modeling analysis module, a prediction model operation module, an evaluation optimization feedback module, a case library management module and a self-adaptive learning module; by integrating multi-source data acquisition and an intelligent preprocessing mechanism, a comprehensive environment-equipment-dust state monitoring system is constructed, the integrity and accuracy of data sensed by the dust removal system are ensured, dust accumulation characteristics are dynamically analyzed based on real-time data flow, dust removal requirements under different working conditions can be adaptively recognized, and the dust removal efficiency is improved. The accuracy and response timeliness of a dust removal decision are ensured, and the environment adaptability of a dust removal system is improved; by introducing a deep learning prediction model and a multi-dimensional feature fusion mechanism, intelligent prediction of a dust accumulation trend and a dust removal effect is realized, so that the system can judge an optimal dust removal opportunity in advance and dynamically adjust equipment parameters.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrostatic dust removal, in particular to a frame surface electrostatic dust removal system and method. BACKGROUND

[0002] Electrostatic dust removal is a method of dust removal for gas, in which dust-containing gas is electrically separated when passing through a high-voltage electrostatic field, and dust particles, after being combined with negative ions to carry negative electricity, tend to discharge on the surface of the anode and deposit; electrostatic dust removal is a dust collection method that uses an electrostatic field to ionize gas and thus make dust particles carry electricity and be adsorbed onto an electrode; in a strong electric field, air molecules are ionized into positive ions and electrons, and the electrons, in the process of rushing to the positive electrode, encounter dust particles, making the dust particles carry negative electricity and be adsorbed onto the positive electrode and collected.

[0003] At present, during the operation of a frame in a complex and changeable outdoor environment, dust and pollutants are easily accumulated on the surface of the frame; the traditional electrostatic dust removal system adopts a fixed threshold triggering mechanism or a fixed time starting mechanism, and thus cannot sense the actual dust accumulation state of the frame surface and the change of environmental conditions in real time, resulting in a lack of precision and adaptability in dust removal operation.

[0004] Therefore, the present application provides a frame surface electrostatic dust removal system and method to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a frame surface electrostatic dust removal system and method to solve the problem of a lack of precision and adaptability in dust removal operation as mentioned in the background.

[0006] To achieve the above object, the present application provides the following technical solutions: a frame surface electrostatic dust removal system and method, the method comprising the following steps: S1, collecting dust distribution data, environmental parameter data and dust removal equipment operation data of the frame surface; S2, preprocessing the collected data, including data cleaning, abnormal value processing and normalization processing, to generate a standardized data set; S3, based on the standardized data set, using a time series analysis method to extract periodic characteristics and trend characteristics of dust accumulation; S4, constructing an environment-dust removal effect correlation model to analyze the change rule of dust removal efficiency under different environmental conditions; S5, establishing a dust removal equipment state evaluation model to evaluate the influence of high-voltage electrostatic generator and electrode plate aging on dust removal performance; S6, inputting the periodic characteristics, trend characteristics, environment-dust removal effect correlation model output and equipment state evaluation results into a deep learning prediction model; S7, using an adaptive optimization algorithm to dynamically adjust the parameters of the deep learning prediction model; S8, output the dust removal control instruction of the vehicle frame surface, including the starting time, the intensity parameter and the duration; S9, compare the actual dust removal effect data with the predicted result, and continuously optimize the prediction model.

[0007] Preferably, the dust distribution data, the environmental parameter data and the dust removal equipment operation data of the vehicle frame surface in S1 include the following steps: S11, collect the dust concentration data of each area of the vehicle frame surface through the dust sensor array, the sampling interval is 1-5 minutes, and the front, middle and rear parts of the vehicle frame are covered; S12, obtain humidity, temperature, wind speed and pollution index data from the environmental monitoring unit, wherein the humidity data is suitable for correcting the dust adsorption characteristics; S13, collect the dust removal equipment operation data, including the voltage output of the high-voltage electrostatic generator, the electrode plate current efficiency and the energy consumption parameter; S14, align the collected multi-source data according to the time stamp and the spatial position, and establish a vehicle frame dust removal multi-dimensional database.

[0008] Preferably, the data preprocessing in S2 includes the following steps: S21, detect and eliminate abnormal dust concentration data points by using the sliding window method, and the abnormality determination standard is that the deviation from the window mean value is more than three times the standard deviation; S22, apply linear interpolation method to complete the missing environmental humidity data; S23, perform minute-level, hour-level and day-level normalization processing on the dust concentration and temperature data; S24, map the equipment operation parameters to the [0, 1] interval to unify the parameter scale.

[0009] Preferably, the extraction of periodic characteristics and trend characteristics of dust accumulation in S3 includes the following steps: S31, extract the daily periodic characteristics of dust accumulation by using Fourier transform, and identify the morning and evening peak accumulation rules; S32, apply exponential smoothing method to decompose the long-term trend component of the dust data, and predict the seasonal accumulation change; S33, construct an autoregressive model to analyze the random fluctuation characteristics of the dust, and capture the influence of sudden pollution events; S34, vectorize the extracted periodic characteristics, trend characteristics and random characteristics to generate a feature matrix for model input.

[0010] Preferably, the establishment of the environment-dust removal effect correlation model in S4 includes the following steps: S41, establish a humidity-dust removal efficiency nonlinear regression model to fit the electrostatic adsorption decay curve under high humidity environment, wherein the dust removal efficiency With humidity The relationship is expressed by the formula: ; in, This indicates the dust removal efficiency under the current humidity level. This indicates the maximum dust removal efficiency under standard humidity conditions. Indicates the humidity attenuation coefficient. Indicates the ambient humidity value; S42. Analyze the influence coefficient of wind speed on dust diffusion and adjust the dust removal intensity using a wind speed correction strategy. S43. Use decision tree algorithm to evaluate the combined effect of multiple pollutants and quantify the contribution of particulate matter type to dust removal effect; S44. Construct an environmental sensitivity matrix, identify the priority of key environmental factors, and optimize the model response speed.

[0011] Preferably, the establishment of the dust removal equipment status assessment model in step S5 includes the following steps: S51. Establish a performance degradation model for a high-voltage electrostatic generator based on historical current data to predict the voltage output decline trend. S52. The degree of surface oxidation and contamination is assessed using electrode plate image recognition technology, and its impact on dust removal efficiency is quantified. S53. Analyze the variation of equipment energy consumption with operating time and establish an energy efficiency evaluation curve; S54. Construct an equipment health status scoring system and output a status assessment report that includes maintenance recommendations.

[0012] Preferably, step S6, which involves inputting periodic features, trend features, the output of the environment-dust removal effect correlation model, and the equipment status assessment results into the deep learning prediction model, includes the following steps: S61. Construct a long short-term memory network prediction model based on an attention mechanism, dynamically weighting each feature input, where the input gate calculation is expressed by the formula: ; in, Indicates the input gate in time The output, This represents the sigmoid activation function. This represents the weight matrix of the input gate. This represents the hidden state of the previous time step. This represents the input features at the current time step. This represents the bias vector of the input gate; S62. Set periodic feature input gate, environmental feature input gate and equipment status input gate in the prediction model; S63, dynamically adjust feature weights using a gating mechanism, prioritizing high sensitivity environmental factors; S64, design residual connection structure to avoid deep network degradation, stabilize prediction results.

[0013] Preferably, the dynamic adjustment in S7 includes the following steps: S71, initialize model parameters, set learning rate and regularization coefficient to prevent overfitting; S72, use particle swarm optimization algorithm to search for the optimal hyperparameter combination to minimize prediction error; S73, design dynamic learning rate adjustment strategy, automatically scale optimization step size according to real-time error feedback; S74, establish parameter adaptive update mechanism to respond to sudden environmental changes.

[0014] Preferably, the output of the dust removal control instruction of the vehicle frame surface in S8 includes the following steps: S81, output the ultra-short-term dust removal control instruction for the next 5-30 minutes, including voltage adjustment and start-up delay; S82, generate short-term dust removal probability prediction interval for the next 1-6 hours, provide confidence level reference; S83, provide medium and long-term dust removal trend prediction for the next 1-7 days, guide preventive maintenance; S84, visualize and display the control instruction and historical effect comparison chart to assist user decision-making.

[0015] Preferably, the system includes: Data acquisition and processing module, collect vehicle frame surface data through dust monitoring unit, collect external factors through environmental parameter acquisition unit, and generate original data set using equipment state detection unit; Feature modeling and analysis module, receive the original data set, extract periodic features through time series decomposition unit, build environment-dust removal effect mapping relationship using environmental correlation modeling unit, and output feature vector set through equipment state evaluation unit; Prediction model operation module, receive the feature vector set, build control model architecture through deep learning network construction unit, dynamically adjust model parameters through parameter optimization unit, and generate control signal using dust removal instruction output unit; Evaluation and optimization feedback module, receive the control signal execution effect data, calculate the actual deviation through error analysis unit, generate optimization instructions using model parameter update unit, and display evaluation report through effect visualization unit; Case library management module, store historical dust removal case data, call reference cases through case feature extraction unit and similarity matching unit, and feed back to the prediction model operation module through the decision support unit to optimize the control strategy; An adaptive learning module integrates the evaluation data output by the evaluation optimization feedback module, analyzes the prediction characteristics through a system capability portrait unit, determines the model improvement direction through a short board identification unit, and generates an optimization scheme through a data acquisition strategy control unit.

[0016] Compared with the prior art, the vehicle frame surface electrostatic dust removal system and method provided by the application have the following beneficial effects: 1. In the application, when performing electrostatic dust removal on the surface of the vehicle frame, a comprehensive environment-equipment-dust state monitoring system is constructed through integrated multi-source data acquisition and intelligent preprocessing mechanism, ensuring the integrity and accuracy of the dust removal system sensing data. Based on real-time data flow, the dust accumulation characteristics are dynamically analyzed, the dust removal demand under different working conditions is adaptively identified, the accuracy and timeliness of the dust removal decision are ensured, and the environmental adaptability of the dust removal system is improved.

[0017] 2. In the application, when performing electrostatic dust removal on the surface of the vehicle frame, a deep learning prediction model and a multi-dimensional feature fusion mechanism are introduced to realize intelligent prediction of dust accumulation trend and dust removal effect, so that the system can judge the best dust removal opportunity in advance and dynamically adjust the equipment parameters. In complex environments, stable dust removal performance can still be maintained, avoiding the problems of excessive dust removal and insufficient dust removal, improving the energy efficiency ratio and the intelligent level of the system.

[0018] 3. In the application, when performing electrostatic dust removal on the surface of the vehicle frame, a closed-loop optimization and adaptive learning mechanism is constructed to compare the dust removal effect with the prediction result in real time and feedback the adjustment of the model parameters, so that the system has the ability of continuous optimization and fault self-diagnosis. It can continuously adapt to changes in the actual use environment of the vehicle frame in long-term operation, reduce maintenance requirements and prolong the service life of the equipment, improve the reliability and economy of the system. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The flowchart of the vehicle frame surface electrostatic dust removal method of the application is shown in the figure. Figure 2 The framework diagram of the vehicle frame surface electrostatic dust removal system of the application is shown in the figure. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the application will be described below in conjunction with the drawings in the embodiments of the application. Obviously, the described embodiments are only a part of the embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0021] Specific embodiment: vehicle frame surface electrostatic dust removal system and method, the method comprising the following steps: S1, collect dust distribution data, environmental parameter data and dust removal equipment operation data on the surface of the vehicle frame; S2, pre-process the collected data, including data cleaning, outlier processing and normalization processing, to generate a standardized data set; S3, based on the standardized data set, use time series analysis method to extract periodic and trend features of dust accumulation; S4, build an environment-dust removal effect correlation model to analyze the change rule of dust removal efficiency under different environmental conditions; S5, establish a dust removal equipment state evaluation model to evaluate the influence of high-voltage electrostatic generator and electrode plate aging on dust removal performance; S6, input the periodic and trend features, environment-dust removal effect correlation model output and equipment state evaluation results into the deep learning prediction model; S7, use adaptive optimization algorithm to dynamically adjust the parameters of the deep learning prediction model; S8, output dust removal control instructions for the surface of the vehicle frame, including start time, intensity parameters and duration; S9, compare the actual dust removal effect data with the prediction results to continuously optimize the prediction model.

[0022] In S1, the dust distribution data, environmental parameter data and dust removal equipment operation data on the surface of the vehicle frame are collected, including the following steps: S11, collect dust concentration data of each area on the surface of the vehicle frame through a dust sensor array, with a sampling interval of 1-5 minutes, covering the front, middle and rear parts of the vehicle frame; The dust sensor array is composed of multiple optical dust sensors arranged in a grid on the surface of the vehicle frame. Each sensor measures the dust particle concentration in real time through laser scattering principle and sends the data to the central processing unit through the communication module; S12, obtain humidity, temperature, wind speed and pollutant index data from the environmental monitoring unit, and correct the dust adsorption characteristics according to the humidity data; S13, collect dust removal equipment operation data, including voltage output of high-voltage electrostatic generator, electrode plate current efficiency and energy consumption parameters; S14, align the collected multi-source data by timestamp and spatial position to establish a multi-dimensional database for vehicle frame dust removal.

[0023] In S2, the data pre-processing includes the following steps: S21, use sliding window method to detect and eliminate abnormal dust concentration data points, with the abnormality determination standard being more than three times the standard deviation from the window mean; Specifically, for time series data points , set the window size , calculate the mean and standard deviation of data within the window: ; ; wherein represents the dust concentration value at time t, represents the window length, represents the window mean, represents the window standard deviation; , it is determined that is an outlier and is removed; S22, apply linear interpolation to complete the missing environmental humidity data, ensure data set continuity: ; wherein represents the interpolated humidity value at time , , represents the nearest time point before and after the missing point, , represents the corresponding humidity value; S23, normalize the dust concentration and temperature data at the minute, hour and day levels; S24, map the device operating parameters to the [0, 1] interval to unify the parameter scale.

[0024] The periodic and trend features of dust accumulation in S3 include the following steps: S31, use Fourier transform to extract the daily periodicity of dust accumulation and identify the morning and evening peak accumulation rules; Apply discrete Fourier transform to convert time series data to frequency domain and calculate frequency components: ; ; wherein represents the dust concentration time series, represents the number of data points, represents the frequency domain coefficient, is the imaginary part in the complex plane, represents the position index in the frequency domain, is the position of the data point in the sequence; S32, apply exponential smoothing method to decompose the long-term trend component of dust data and predict seasonal accumulation changes; Exponential smoothing model: ; wherein represents the smoothed trend value, representing time of observation, representing smoothing parameter, S33, construct an autoregressive model to analyze the random fluctuation characteristics of dust and capture the impact of sudden pollution events; The autoregressive model is represented as: wherein representing dust concentration at time , represents a constant, representing autoregressive coefficient, representing model order, representing white noise error term; S34, vectorize the extracted periodic, trend and random feature vectors and store them to generate a feature matrix for model input.

[0025] S4, constructing an environment-dedusting effect correlation model, including the following steps: S41, establishing a humidity-dedusting efficiency nonlinear regression model to fit the electrostatic adsorption decay curve under high humidity environment, wherein the dedusting efficiency and humidity are related by the formula: wherein, representing dedusting efficiency under current humidity, representing maximum dedusting efficiency under standard humidity, representing humidity decay coefficient, representing environmental humidity value; S42, analyzing the influence coefficient of wind speed on dust diffusion, and adopting a wind speed correction strategy to adjust the dedusting intensity; Based on the real-time data of the wind speed sensor, the wind speed influence factor is calculated, and the output voltage and pulse frequency of the electrostatic dust collector are dynamically adjusted according to the wind speed; The specific strategy is: when the wind speed is lower than the threshold value, maintain the baseline dedusting intensity; when the wind speed increases, increase the dedusting voltage in a linear relationship to compensate for the wind-induced dust diffusion; when the wind speed is too high, start the strong wind mode and increase the dedusting frequency to ensure the coverage area; S43, using a decision tree algorithm to evaluate the comprehensive effect of multiple pollutants and quantifying the contribution of particulate matter types to the dedusting effect; A decision tree model is constructed, with environmental parameters and pollutant data as input features and dedusting efficiency as output target.

[0026] The split point is selected by information gain and Gini impurity to generate a binary tree structure. ​​​​

[0027] After the training is completed, the model is used to predict the dust removal effect under different conditions, and the feature importance is analyzed to quantify the contribution of each pollutant; S44, construct an environmental sensitivity matrix to identify the priority of key environmental factors and optimize the response speed of the model.

[0028] The environmental sensitivity matrix is defined as: ; Wherein represents the influence coefficient of the first environmental factor on the first dust removal performance index, the row of the matrix represents the environmental factor, and the column represents the performance index, which is suitable for priority ranking. The dust removal equipment state evaluation model in S5 includes the following steps: S51, establish a high-voltage electrostatic generator performance attenuation model based on historical current data to predict the voltage output decline trend; S52, evaluate the surface oxidation and contamination degree through electrode plate image recognition technology, and quantify the loss rate of dust removal efficiency; S53, analyze the change rule of equipment energy consumption with running time, and establish an energy efficiency evaluation curve; S54, construct a device health state scoring system, and output a state evaluation report containing maintenance suggestions.

[0029] In S6, the periodic feature, the trend feature, the environment-dust removal effect correlation model output and the equipment state evaluation result are input into the deep learning prediction model, including the following steps: S61, construct a long short-term memory network prediction model based on attention mechanism, dynamically weight each feature input, wherein the input gate calculation is represented by the formula: ; Wherein, represents the output of the input gate at time , represents the sigmoid activation function, represents the weight matrix of the input gate, represents the hidden state of the previous time step, represents the input feature of the current time step, represents the bias vector of the input gate; S62, set the periodic feature input gate, the environmental feature input gate and the equipment state input gate in the prediction model; S63, dynamically adjust the feature weight by using the gating mechanism, and give priority to the high sensitivity environmental factors; S64, design a residual connection structure to avoid deep network degradation and stabilize the prediction result.

[0030] The dynamic adjustment in S7 includes the following steps: S71, initialize model parameters, set learning rate and regularization coefficient to prevent overfitting; S72, search for the optimal hyperparameter combination using the particle swarm optimization algorithm to minimize the prediction error; In the particle swarm optimization algorithm, each particle position represents a set of hyperparameters, and the velocity update formula is: ; The position update formula is: ; Where represents the velocity of the particle in dimension , represents the position, represents the individual optimum, represents the global optimum, represents the inertia weight, , represents the learning factor, , represents a random number; S73, design a dynamic learning rate adjustment strategy to automatically scale the optimization step size based on real-time error feedback; Adaptive learning rate adjustment is used based on error change rate: ; Where represents the learning rate at time , represents the initial learning rate, represents the decay coefficient, represents the error change amount, represents the current error value; When the error decreases, the learning rate decreases to fine-tune; when the error increases, the learning rate increases to accelerate convergence; S74, establish a parameter adaptive update mechanism to respond to sudden environmental changes: By monitoring environmental sensor data in real time, when a sudden change is detected, the parameter update process is triggered: First, suspend the current model prediction and enable the emergency dust removal mode; Second, quickly re-adjust the model parameters based on the latest data and use reinforcement learning algorithms to optimize the response strategy; Finally, update the case library and adjust the future parameter update frequency to ensure that the system quickly adapts to abnormal conditions.

[0031] S8 outputs the dust removal control instructions for the vehicle frame surface, including the following steps: S81, output the ultra-short-term dust removal control instruction for the next 5-30 minutes, including voltage adjustment and start-up delay; S82, generate a short-term dust removal probability prediction interval for the next 1-6 hours, providing a confidence level reference; S83, provide a medium and long-term dust removal trend prediction for the next 1-7 days to guide preventive maintenance; S84, visualize the control instruction and historical effect comparison chart to assist user decision-making.

[0032] The system comprises: A data acquisition and processing module acquires vehicle frame surface data through a dust monitoring unit, collects external factors through an environmental parameter acquisition unit, and generates an original data set using a device state detection unit; A feature modeling and analysis module receives the original data set, extracts periodic features through a time series decomposition unit, builds an environment-dust removal effect mapping relationship using an environmental correlation modeling unit, and outputs a feature vector set through a device state evaluation unit; A prediction model operation module receives the feature vector set, builds a control model architecture through a deep learning network construction unit, dynamically adjusts model parameters through a parameter optimization unit, and generates control signals using a dust removal instruction output unit; An evaluation and optimization feedback module receives control signal execution effect data, calculates actual deviation through an error analysis unit, generates optimization instructions using a model parameter update unit, and displays evaluation reports through an effect visualization unit; A case library management module stores historical dust removal case data, calls reference cases through a case feature extraction unit and a similarity matching unit, and feeds back to the prediction model operation module through a decision support unit to optimize control strategies; An adaptive learning module integrates evaluation data output by the evaluation and optimization feedback module, analyzes prediction features through a system capability profiling unit, determines model improvement direction through a short board identification unit, and generates optimization schemes through a data acquisition strategy control unit.

[0033] The operation steps of the system and method are as follows: Step 1, multi-source collaborative perception stage First, a distributed dust sensor array is used to capture the particulate matter concentration of each area on the vehicle frame surface in real time. The sensors are arranged in a grid layout to cover the key areas of the front, middle and rear of the vehicle frame, and laser scattering technology is used to accurately quantify the dust accumulation thickness. At the same time, the environmental monitoring unit synchronously collects humidity, temperature, wind speed and pollutant composition data, among which the humidity sensor dynamically corrects the dust adsorption characteristic parameters; the device state detection module continuously acquires high-voltage electrostatic generator output voltage curve, electrode plate current efficiency and other operating parameters. All heterogeneous data streams are fused into a unified time series data set through a space-time alignment engine, establishing a full-dimensional perception basis for subsequent analysis.

[0034] Step two, dynamic feature modeling phase Based on the standardized data set, the system adopts a three-level feature analysis architecture: Periodic feature extraction: through the Fourier transformer to decompose the diurnal fluctuation law of dust accumulation, identify the pollution peak mode of morning and evening commuting period; Environmental correlation modeling: Constructing humidity-dust removal efficiency decay curve, quantifying the decline trend of electrostatic adsorption efficiency under high humidity environment, while establishing wind speed influence matrix, dynamically adjusting dust removal intensity to offset the effect of wind-induced dust diffusion; Device health assessment: Based on historical current data to establish electrostatic generator performance decay model, combined with electrode plate surface image recognition technology, quantifying the loss rate of dust removal efficiency caused by oxidation layer, generating device health score report.

[0035] Step three, intelligent prediction decision-making phase After receiving the feature vector set, the deep learning prediction engine processes multiple source inputs through a spatio-temporal fusion mechanism: Core architecture: Embedding attention gate mechanism in long short-term memory network, periodic feature input gate captures daily and weekly accumulation rules, environmental feature input gate dynamically weights the impact of sudden sandstorm events, and device state input gate adjusts weight distribution according to electrode plate aging degree; Adaptive parameter optimization: Adopting particle swarm algorithm to search for optimal hyperparameter combination in real time, when detecting environmental mutations, triggering dynamic learning rate adjustment mechanism to accelerate model convergence to match real-time working conditions; Multi-scale instruction generation: Outputting 5-30 minute ultra-short-term dust removal voltage adjustment instructions, synchronously generating 6-hour dust removal probability prediction with confidence interval, guiding the system to achieve balance between energy consumption and effect.

[0036] Step four, closed-loop optimization evolution phase After dust removal, the system starts a three-layer feedback mechanism: Error traceability analysis: Compare the actual dust removal effect with the predicted value, locate the deviation source from environmental perception error and model parameter lag; Case library linkage optimization: Call historical similar cases, inject feature weight correction strategies through decision support unit; Self-evolution ability construction: Based on the continuous running prediction ability portrait, the short board identification unit automatically strengthens weak links, and the data acquisition strategy control unit dynamically increases the sampling frequency of sensors in high error period, forming an intelligent closed loop of "perception-decision-optimization".

[0037] Step five, human-machine collaborative execution phase The visualization engine converts key data into decision dashboards: real-time display of the superimposed analysis of the chassis surface dust thermal map and the predicted dust removal trajectory, marking the high pollution risk area; at the same time, generating equipment maintenance warning prompts, guiding manual intervention cleaning, component replacement. All control commands are transmitted to the electrostatic generator execution unit through the industrial bus to realize seamless connection from intelligent prediction to physical dust removal.

[0038] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by an "including a" statement does not exclude the existence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0039] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, alternatives, and variations can be made in the embodiments without departing from the spirit and scope of the present application as defined by the appended claims and their equivalents.

Claims

1. A method of electrostatically dusting a surface of a vehicle frame, characterized by: The method comprises the following steps: S1, collecting dust distribution data, environmental parameter data and dust removal equipment operation data on the surface of the vehicle frame; S2, preprocessing the collected data, including data cleaning, outlier processing and normalization processing, to generate a standardized data set; S3, based on the standardized data set, using time series analysis method to extract periodic and trend characteristics of dust accumulation; S4, constructing an environment-dust removal effect correlation model to analyze the change rule of dust removal efficiency under different environmental conditions; S5, establishing a dust removal equipment state evaluation model to evaluate the influence of high-voltage electrostatic generator and electrode plate aging on dust removal performance; S6, inputting the periodic and trend characteristics, environment-dust removal effect correlation model output and equipment state evaluation results into a deep learning prediction model; S7, using an adaptive optimization algorithm to dynamically adjust the parameters of the deep learning prediction model; S8, outputting dust removal control instructions on the surface of the vehicle frame, including start time, intensity parameters and duration; S9, comparing the actual dust removal effect data with the prediction results to continuously optimize the prediction model.

2. The method of claim 1, wherein: In S1, collecting dust distribution data, environmental parameter data and dust removal equipment operation data on the surface of the vehicle frame, comprising the following steps: S11, collecting dust concentration data of each area on the surface of the vehicle frame through a dust sensor array, with a sampling interval of 1-5 minutes, covering the front, middle and rear parts of the vehicle frame; S12, obtaining humidity, temperature, wind speed and pollutant index data from the environmental monitoring unit, wherein the humidity data is suitable for correcting dust adsorption characteristics; S13, collecting dust removal equipment operation data, including voltage output of high-voltage electrostatic generator, electrode plate current efficiency and energy consumption parameters; S14, aligning the collected multi-source data by timestamp and spatial position to establish a multi-dimensional database for vehicle frame dust removal.

3. The method of claim 1, wherein: In S2, the data preprocessing comprises the following steps: S21, using a sliding window method to detect and eliminate abnormal dust concentration data points, with an abnormality determination standard of deviating from the window mean value by more than three times the standard deviation; S22, applying linear interpolation method to complete the missing environmental humidity data; S23, performing minute, hour and day level normalization processing on dust concentration and temperature data; S24, mapping equipment operation parameters to the [0, 1] interval to unify the parameter scale.

4. The method of claim 1, wherein: In S3, the periodic and trend characteristics of dust accumulation are extracted, comprising the following steps: S31, using Fourier transform to extract the daily periodic characteristics of dust accumulation to identify the morning and evening peak accumulation rules; S32, applying exponential smoothing method to decompose the long-term trend component of dust data to predict seasonal accumulation changes; S33, constructing an autoregressive model to analyze the random fluctuation characteristics of dust to capture the influence of sudden pollution events; S34, vectorizing the extracted periodic, trend and random characteristics to generate a feature matrix for model input.

5. The method of claim 1, wherein: In S4, the environment-dust removal effect correlation model is constructed, comprising the following steps: S41, a humidity-dedusting efficiency nonlinear regression model is established, and the electrostatic adsorption decay curve under high humidity environment is fitted, wherein the relationship between the dedusting efficiency and humidity is represented by the formula: ； wherein, represents the dust removal efficiency at the current humidity, represents the maximum dust removal efficiency at the standard humidity, represents the humidity attenuation coefficient, represents the ambient humidity value; S42, analyzing the influence coefficient of wind speed on dust diffusion, and using wind speed correction strategy to adjust dust removal intensity; S43, using decision tree algorithm to evaluate the comprehensive effect of multiple pollutants and quantifying the contribution of particulate matter types to dust removal effect; S44, construct an environmental sensitivity matrix to identify the priority of key environmental factors and optimize the response speed of the model.

6. The method of claim 1, wherein: The S5 of establishing a dust removal equipment state evaluation model comprises the following steps: S51, establish a high-voltage electrostatic generator performance attenuation model based on historical current data to predict the voltage output decline trend; S52, evaluate the surface oxidation and contamination degree through electrode plate image recognition technology, and quantify the loss rate of dust removal efficiency; S53, analyze the change rule of equipment energy consumption with running time, and establish an energy efficiency evaluation curve; S54, construct a device health state scoring system, and output a state evaluation report containing maintenance suggestions.

7. The method of claim 1, wherein: The S6 of inputting the periodic feature, trend feature, environment-dust removal effect correlation model output and equipment state evaluation result into the deep learning prediction model comprises the following steps: S61, construct a long short-term memory network prediction model based on an attention mechanism, dynamically weight each feature input, and the input gate calculation is represented by the formula: ； wherein, represents the output of the input gate at time , represents a sigmoid activation function, represents a weight matrix of the input gate, represents the hidden state of the previous time step, represents the input feature of the current time step, represents a bias vector of the input gate; S62, set the periodic feature input gate, environmental feature input gate and equipment state input gate in the prediction model; S63, dynamically adjust the feature weight by using the gating mechanism, and give priority to the high sensitivity environmental factors; S64, design a residual connection structure to avoid deep network degradation and stabilize the prediction result.

8. The method of claim 1, wherein: The S7 of dynamically adjusting comprises the following steps: S71, initialize the model parameters, set the learning rate and regularization coefficient to prevent overfitting; S72, search for the optimal hyperparameter combination by using the particle swarm optimization algorithm to minimize the prediction error; S73, design a dynamic learning rate adjustment strategy to automatically scale the optimization step according to real-time error feedback; S74, establish a parameter adaptive updating mechanism to respond to sudden environmental changes.

9. The method of claim 1, wherein: The S8 of outputting the dust removal control instruction of the vehicle frame surface comprises the following steps: S81, output the ultra-short-term dust removal control instruction for the next 5-30 minutes, including voltage adjustment and start-up delay; S82, generate a short-term dust removal probability prediction interval for the next 1-6 hours to provide a confidence level reference; S83, provide a medium and long-term dust removal trend prediction for the next 1-7 days to guide preventive maintenance; S84, visually display the control instruction and historical effect comparison chart to assist user decision-making.

10. A vehicle body surface electrostatic dust removal system for implementing the vehicle body surface electrostatic dust removal method according to any one of claims 1 to 9, characterized by: The system comprises: A data acquisition and processing module acquires vehicle frame surface data through a dust monitoring unit, collects external factors through an environmental parameter acquisition unit, and generates an original data set using a device state detection unit; A feature modeling and analysis module receives the original data set, extracts periodic features through a time series decomposition unit, builds an environment-dust removal effect mapping relationship using an environmental correlation modeling unit, and outputs a feature vector set through a device state evaluation unit; A prediction model operation module receives the feature vector set, builds a control model architecture through a deep learning network construction unit, dynamically adjusts model parameters through a parameter optimization unit, and generates a control signal using a dust removal instruction output unit; An evaluation and optimization feedback module receives the control signal execution effect data, calculates the actual deviation through an error analysis unit, generates optimization instructions using a model parameter updating unit, and displays an evaluation report through an effect visualization unit. A case library management module stores historical dust removal case data, calls reference cases through a case feature extraction unit and a similarity matching unit, and feeds back to a prediction model operation module through a decision support unit to optimize a control strategy; An adaptive learning module integrates evaluation data output by the evaluation optimization feedback module, analyzes prediction features through a system capability portrait unit, determines a model improvement direction through a short board identification unit, and generates an optimization scheme through a data collection strategy control unit.

Images