Intelligent control system for sand blasting uniformity of glove based on multi-sensor data fusion

Abstract

The application discloses a glove sand blasting uniformity intelligent control system based on multi-sensor data fusion, relates to the technical field of automatic control of flexible curved surface workpiece sand blasting processing, and sequentially comprises a perception acquisition unit, a data preprocessing unit, a model collaborative decision unit and an execution driving unit to form a closed loop control link.The perception acquisition unit synchronously collects nine kinds of physical parameters; the data preprocessing unit outputs a regular data set; the model collaborative decision unit sequentially performs three-stage operation through an uniformity quantization model, a curved surface curvature adaptive parameter model and a multi-working condition collaborative compensation model, solves final process control variables and partition differentiated control variables; and the execution driving unit controls a three-axis servo, a proportional pressure regulating valve, a sand flow controller and a tensioning mechanism, so that partition parameter synchronous switching and boundary transition are realized.The system can reduce roughness standard deviation through closed loop iteration and can improve the sand blasting uniformity of the glove surface through uniformity quantization coefficient prompting.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for sandblasting of flexible curved workpieces, and in particular to an intelligent control system for uniform sandblasting of gloves based on multi-sensor data fusion. Background Technology

[0002] Sandblasting is a core process for improving the anti-slip performance and surface texture of protective gloves. Currently, most industries use traditional sandblasting equipment with fixed process parameters, relying on manual experience to set sandblasting pressure, gun speed, and spray distance, and completing basic operations with simple sensors. In actual processing, gloves are flexible curved surface components, and the curvature and stress requirements of the fingertips, palms, and wrists are different. Fixed parameters are difficult to match the changes in the curved surface, which can easily lead to local over-blasting or under-blasting, directly resulting in uneven surface roughness.

[0003] Meanwhile, fluctuations in the temperature and humidity of the processing environment, differences in the moisture content and particle size of the sand, and wear caused by long-term use of the nozzles will continuously affect the sandblasting effect. The existing system lacks online detection and real-time compensation methods, and can only perform random inspections and repairs after processing, which is not only inefficient, but also causes batch quality instability.

[0004] More notably, traditional equipment can only collect a small amount of operating data, lacks effective data denoising, outlier processing and spatiotemporal synchronization capabilities, cannot achieve multi-source information fusion judgment, and has not established a uniformity quantitative assessment and closed-loop correction mechanism. Overall, the control accuracy is low and the anti-interference capability is poor, making it difficult to meet the requirements of large-scale, high-quality production of high-end gloves. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent control system for the uniformity of glove sandblasting based on multi-sensor data fusion, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent control system for the uniformity of glove sandblasting based on multi-sensor data fusion, comprising:

[0007] The system comprises a sensing and acquisition unit, a data preprocessing unit, a model collaborative decision-making unit, and an execution driving unit, which are sequentially connected to form a closed-loop control link.

[0008] The sensing and acquisition unit is used to synchronously collect multi-source physical parameter data of the entire sandblasting operation. The multi-source physical parameter data includes at least the distance between the spray gun and the glove surface, air supply pressure, sand flow rate, glove surface roughness, ambient temperature and humidity, sand moisture content, sandblasting impact sound pressure, spray gun movement speed, and nozzle wear thickness.

[0009] The data preprocessing unit is used to perform noise reduction, outlier removal and spatiotemporal alignment on the multi-source physical parameter data, and output a spatiotemporally regularized dataset.

[0010] The model collaborative decision-making unit includes a sandblasting uniformity quantification model, a surface curvature adaptive parameter model, and a multi-condition collaborative compensation model. The model collaborative decision-making unit is used to sequentially input the spatiotemporal regularized dataset into the sandblasting uniformity quantification model, the surface curvature adaptive parameter model, and the multi-condition collaborative compensation model for three-level correlation operations, and output the final process control quantity and zoning control quantity.

[0011] The execution drive unit is used to control the action of the sandblasting actuator according to the final process control quantity and the zone control quantity, so as to realize the differentiated sandblasting control of the whole area and the zone.

[0012] The technical effects and advantages of this invention are as follows:

[0013] 1. This invention integrates multiple sensors such as laser displacement, pressure, flow rate, industrial camera, acoustic, and nozzle wear sensors to comprehensively collect key information such as surface morphology, roughness, environment, sand material, and loss, thus overcoming the shortcomings of traditional systems that only perceive one type of material.

[0014] 2. This invention improves data reliability and avoids the impact of noise and timing misalignment on control performance by using wavelet denoising, dynamic outlier removal, and spatiotemporal synchronization alignment.

[0015] 3. This invention establishes a quantitative model for global and regional uniformity, using numerical indicators to guide parameter adjustment, replacing manual experience, and making control more precise and stable;

[0016] 4. This invention calculates the glove curvature by fitting a three-dimensional surface and sets different process parameters for the fingertips, palms, and wrists, thus solving the problem of uneven sandblasting of irregular curved surfaces from the root.

[0017] 5. This invention quantifies and automatically corrects interferences such as nozzle wear, environmental changes, and sand condition online, improving the stability of the system under complex working conditions. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall method of the present invention;

[0019] Figure 2 This is a flowchart of the sensing and acquisition synchronization process of the present invention;

[0020] Figure 3 This is a diagram of the data preprocessing logic pipeline of the present invention;

[0021] Figure 4 This is a logic diagram of the model collaborative decision-making and hierarchical operation of the present invention;

[0022] Figure 5 This is a diagram illustrating the overall execution and closed-loop iterative control logic of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention provides, for example Figure 1-5 The intelligent control system for uniformity of glove sandblasting based on multi-sensor data fusion shown includes:

[0025] Sensing and Acquisition Unit: The unit synchronously collects physical parameters of the entire sandblasting operation by deploying multiple types of sensors. The sensing and acquisition unit includes a laser displacement sensor, a gas pressure sensor, a sand flow sensor, an industrial area array camera, a temperature and humidity sensor, a sand moisture content sensor, an acoustic sensor, a servo encoder, and an online nozzle wear detection sensor.

[0026] The working steps of the sensing and acquisition unit are as follows:

[0027] S1.1 Laser Displacement Sensor Acquisition: The laser displacement sensor is coaxially positioned above the spray gun nozzle to acquire the vertical distance from the spray gun outlet to the measured surface of the glove, denoted as [missing information]. The unit is mm, and the sampling period is 10ms. During the acquisition, the spatial coordinates (X, Y, Z three-dimensional coordinates) of the measurement point are recorded simultaneously, and the sampling interval is 0.5mm. The three-dimensional coordinate mapping relationship of the glove surface is established to provide a spatial reference for subsequent curvature calculation. At the same time, the initial coordinates after each clamping are recorded for subsequent area registration.

[0028] S1.2 Gas Pressure Sensor Acquisition: Connect the gas pressure sensor in series to the sandblasting air inlet main pipeline to acquire the real-time gas supply pressure of the pipeline, denoted as . Unit: MPa; sampling period: 10ms; pressure fluctuation values ​​are collected synchronously and recorded as... This provides supplementary data for quantifying nozzle wear.

[0029] S1.3 Sand Flow Sensor Data Acquisition: A sand flow sensor is installed in the sand conveying pipeline to collect the mass of sand conveyed per unit time, recorded as follows: Unit: g / s; sampling period: 10ms; synchronous output of sand particle size distribution coefficient, denoted as: The particle size was pre-calibrated through sand sieving experiments. The correspondence.

[0030] S1.4 Industrial Area Scan Camera Acquisition: The industrial area scan camera is positioned outside the observation window of the sandblasting chamber, with the lens facing the center area of ​​the glove. The surface texture image of the glove is acquired, and the quantitative value of the surface roughness is calculated and denoted as... The unit is μm, and the sampling period is 20ms. A partitioned acquisition mode is adopted, combining the three-dimensional coordinates acquired by the laser displacement sensor. A coordinate registration method is used (specifically: establishing a local coordinate system based on three feature points on the wrist edge of the glove, and normalizing and registering the coordinates of the fingertip, palm, and wrist regions) to ensure a one-to-one correspondence between the spatial positions of the three regions after each clamping. The roughness of each region is calculated separately and denoted as μm. , , (Corresponding to fingertips, palms, and wrists respectively).

[0031] S1.5 Temperature and Humidity Sensor Data Acquisition: The temperature and humidity sensor is placed inside the sandblasting chamber, 100mm from the glove surface. The ambient temperature and relative humidity inside the chamber are collected and recorded sequentially. , The units are ℃ and %RH, respectively, with a sampling period of 50ms; the rate of change of temperature and humidity is collected simultaneously and denoted as... , It is used for the quantification of environmental disturbances.

[0032] S1.6 Sand Moisture Content Sensor Data Acquisition: The sand moisture content sensor is buried inside the sand storage silo at a depth of 1 / 2 of the sand height. The mass moisture content of the sand is collected and recorded as follows. The unit is %, and the sampling period is 50ms; the trend of sand moisture content change is collected simultaneously and recorded as . .

[0033] S1.7 Acoustic Sensor Data Acquisition: The acoustic sensor is fixed to the side wall of the sandblasting chamber, 500mm from the nozzle outlet, to collect the sandblasting impact sound pressure intensity and the dominant frequency of the sound wave, which are recorded sequentially as follows: , The units are dB and Hz, respectively, with a sampling period of 10ms; the changes in the sound wave spectrum are collected synchronously and denoted as... .

[0034] S1.8 Servo Encoder Acquisition: A servo encoder is installed at the end of the spray gun's motion servo motor to acquire the three-dimensional linear velocity of the spray gun, denoted as... Unit: mm / s; synchronously collect the acceleration of the spray gun movement, denoted as... It is used for speed closed-loop correction.

[0035] S1.9 Nozzle Wear Online Detection Sensor Data Acquisition: The nozzle wear online detection sensor is placed on the inner wall of the nozzle, 10mm from the outlet, to acquire the real-time value of the wear thickness on the inner wall of the nozzle, recorded as follows: The unit is mm, and the sampling period is 50ms; it works in conjunction with the data collected by S1.2 and S1.7 to achieve multi-dimensional quantification of nozzle wear.

[0036] Data preprocessing unit: The raw data sequence output by the sensing and acquisition unit is processed sequentially for denoising, outlier removal, and spatiotemporal alignment to output a valid dataset, which is then sent to the model collaborative decision-making unit.

[0037] The data preprocessing unit operates as follows:

[0038] S2.1 Wavelet Threshold Denoising: The original data sequences output from each sensor are decomposed into two-dimensional wavelet bases using the db4 wavelet basis. The decomposition layer is 3. An adaptive threshold adjustment algorithm is introduced, and the threshold calculation formula is as follows: ( The threshold is dynamically adjusted based on the data fluctuation amplitude (where the data sequence length is used), retaining the low-frequency effective operating condition signal components and the high-frequency operating condition change signal components, filtering out high-frequency electromagnetic interference and vibration noise, and obtaining the denoised data sequence. The original data sequence is a set of time-series values ​​continuously collected by the sensor at a fixed sampling period; the denoised data sequence is a set of effective time-series values ​​after removing invalid interference components and retaining the operating condition change characteristics.

[0039] S2.23σ criterion outlier removal: Calculate the time-series mean of the denoised data sequence. With time series standard deviation Introducing dynamic deviation coefficient (Value range: 1.2~1.8) The real-time adjustment logic is as follows: ,in (Gas pressure fluctuation coefficient) (Comprehensive Environmental Disturbance Coefficient), meaning the greater the fluctuation in operating conditions, The larger the value, the more likely it is to satisfy the condition. The values ​​are calculated to remove abnormal values ​​that exceed the threshold, while retaining reasonable data fluctuations caused by changes in operating conditions to avoid erroneous removal.

[0040] The formula for calculating the mean is:

[0041] Formula for calculating standard deviation:

[0042] S2.3 Spatiotemporal Synchronization Alignment: Using the 10ms interrupt clock of the servo control system as a unified reference, all data after preprocessing is mapped to the same time beat, matching the spatial coordinate correspondence acquired in S1.1, and using linear interpolation to fill the data gaps in different sampling periods to form a spatiotemporally regularized dataset. The spatiotemporally regularized dataset is a combination of multi-physical parameter data that is time-synchronized and has one-to-one spatial point correspondence.

[0043] Model collaborative decision-making unit: The spatiotemporal regularized dataset output by the data preprocessing unit is sequentially fed into the three-layer correlation operation model to output the final process control quantity and zonal control quantity.

[0044] The working steps of the model collaborative decision-making unit are as follows:

[0045] S3.1 Sandblasting Uniformity Quantization Model Calculation: Based on the partition roughness data in the spatiotemporal regularized dataset, the global uniformity quantization coefficient and partition uniformity correction coefficient are calculated and used as the basis for system parameter adjustment, the basis for partition parameter correction, and the constraint conditions for subsequent calculations.

[0046] S3.1.1 Calculate the global and regional average roughness: Select 10 evenly distributed detection points (30 detection points in total) on each of the three regions of the glove surface (finger tip, palm, and wrist), and extract the roughness value of each detection point. , , Calculate the average roughness of each partition. , , and global average roughness : , ;

[0047] The partition weight coefficients of 0.4, 0.4, and 0.2 (corresponding to fingertips, palms, and wrists) were determined through five sets of comparative experiments based on the usage requirements of each area of ​​the glove (palms and fingertips are the main force-bearing and anti-slip areas).

[0048] S3.1.2 Calculate the uniformity quantization coefficient and partition correction coefficient: Calculate the roughness dispersion deviation based on the global and partition average roughness, and normalize it to obtain the uniformity quantization coefficient $U$ and the partition uniformity correction coefficient. , , :

[0049] , ;

[0050] Set standard limit value ;when At that time, subsequent parameter adaptive calculation and compensation calculation are initiated, and simultaneously based on , , Differential adjustments are made to the process parameters for each zone; when At the same time, while maintaining the existing process parameters unchanged, fine-tuning the parameters of each zone is still performed to ensure overall uniformity and stability. The calculated results are then used to... , , , It is simultaneously transmitted to the next two levels of the computing model.

[0051] S3.2 Calculation of Adaptive Parametric Model for Surface Curvature: Receive Uniformity Quantization Coefficient , , , The distance data and spatial coordinate data in the spatiotemporal warping dataset are used to generate initial process parameters for surface adaptation and partition correction parameters through two-level operations, providing basic parameters for subsequent compensation calculations;

[0052] S3.2.1 Calculation of surface curvature and its variation trend: Based on the distance data and spatial coordinates (X, Y, Z) collected by the laser displacement sensor, a B-spline 3D surface fitting algorithm is used for fitting. The specific implementation details are as follows:

[0053] 1. Basis function selection: Cubic B-spline basis function is used, the node vectors are uniformly distributed, and the node spacing is consistent with the laser sampling spacing (0.5mm).

[0054] 2. Control point selection: 50 feature points (covering fingertips, palms, and wrists) evenly distributed on the glove surface are selected as B-spline surface control points. The coordinates of the control points are directly determined by the three-dimensional coordinates collected by the laser displacement sensor.

[0055] 3. Resampling and Boundary Conditions: The resampling interval is 0.3 mm to ensure the smoothness of the fitted surface; the boundary conditions adopt natural boundaries (second derivative is 0) to adapt to the surface shape of the glove edge;

[0056] 4. Calculation of higher-order derivatives: The surface profile of the glove is obtained by fitting. Subsequently, to avoid amplifying noise through numerical difference, a Gaussian filter (with a kernel size of 5×5 and a standard deviation of 0.8) was used to smooth the fitted surface, and then the first-order partial derivatives were calculated using the central difference method. , Second-order partial derivatives , , Finally, the curvature of the surface at the measurement point is calculated. and rate of change of curvature : , ( (Sampling period 10ms)

[0057] The The surface curvature of the glove measuring point characterizes the degree of local bending. The rate of change of curvature represents the trend of change in the degree of curvature of the surface.

[0058] S3.2.2 Solving for adaptive initial parameters and partition correction parameters: In Under these conditions, combined with preset material reference parameters , , Curvature of coupled surface $K$ and correction coefficient for partition uniformity , , The adaptive initial parameters are obtained by solving. , , and partition correction parameters , , ( ): , , , , ;

[0059] curvature in the formula (unit ), Comprehensive environmental disturbance coefficient ( (Units: ℃ / ms + %RH / ms + % / ms) and uniformity of sand particle size distribution (Dimensionless) Pressure fluctuation coefficient (Dimensionless) All values ​​were first dimensionless (divided by their respective reference values). The benchmark value is , The benchmark value is To ensure uniformity of all dimensions; the coefficients (0.05, 0.02, 0.03, 0.01, 0.10, 0.01) in the above formulas were all obtained through three sets of experiments under different working conditions (see the core data in the experiment report below for specific calibration data). , , These are the partition weighting coefficients; , , Preset reference air pressure, reference spray distance, and reference moving speed for gloves of corresponding materials; , , To adapt the initial air pressure, initial spray distance, and initial moving speed to the curved surface; , , This refers to the correction amount for the process parameters of each zone. , , and , , It is then transmitted to the next-level computational model.

[0060] S3.3 Multi-condition collaborative compensation model calculation: Simultaneously receiving uniformity quantization coefficients The initial parameters for surface adaptation and the parameters for partition correction, as well as the pressure, sound pressure, temperature, humidity, moisture content, and main frequency data of the sound waves in the spatiotemporal regularized dataset, are processed by two levels of computational variables, chemical condition disturbances, and collaborative corrections to output the final process control variables and partition control variables.

[0061] S3.3.1 Quantitative Analysis of Disturbance Equivalents under Three Working Conditions: Disturbance equivalents for nozzle wear, environmental conditions, and sand particle size were quantified separately. All coefficients were obtained through three sets of experiments under different working conditions (see the core data in the experimental report below for specific calibration data).

[0062] Nozzle wear equivalent (Based on data collected from S1.9) ):

[0063] Environmental interference coefficient (combined) ):

[0064] Sand particle size interference coefficient (combined) ):

[0065] In the formula The pressure deviation rate is dimensionless. The sound pressure deviation rate is dimensionless. Nozzle wear thickness, in mm. Temperature deviation, unit: °C Humidity deviation, unit: %RH The moisture content of the sand is expressed as a percentage. The dimensionless comprehensive coefficient of environmental disturbance, Main frequency deviation rate, dimensionless The particle size distribution deviation is dimensionless; The reference sound pressure for the new nozzle (calibrated through a no-load test of the new nozzle; the calibration values ​​are shown in the sensor calibration data table below). The reference acoustic frequency corresponding to standard particle size sand (calibrated through sandblasting experiments with standard particle size sand, the calibration value is shown in the sensor calibration data table below). This refers to the uniformity of sand particle size distribution.

[0066] S3.3.2 Solving for final process control quantities and zone control quantities: Using the initial parameters of surface adaptation as a basis, the uniformity quantization coefficient, the corrected disturbance equivalent, and the zone correction parameters are coupled to collaboratively solve for the final process control quantities. , , , and zoning control quantity , , , ( ): , , , , , ,

[0067] The coefficients (0.10, 0.02, 0.01, 0.03) in the above formula were all obtained through three sets of experiments under different working conditions (see the core data of the experiment report below for specific calibration data). Preset the baseline sand flow rate (calibrated through material compatibility testing); , , , The final air supply pressure, final injection distance, final moving speed, and final sand flow rate output by the system; , , , This provides the final process control parameters for each zone, enabling differentiated regulation between zones.

[0068] It should be further explained that the constraint relationships between the models at each layer are as follows:

[0069] Coefficients output by the uniformity quantization model , , , The start and stop conditions of the adaptive parameter model for surface curvature are limited, while the correction range of the multi-condition collaborative compensation model is constrained.

[0070] The initial parameters and partition correction parameters output by the surface curvature adaptive parameter model are used as the computational basis for the multi-condition collaborative compensation model.

[0071] The multi-condition collaborative compensation model integrates all the calculation results of the first two levels of the model, superimposes the quantified values ​​of the condition disturbance, and converges them into a set of final process control quantities and zone control quantities.

[0072] Execution-driven unit: The final process control quantity and zonal control quantity output by the model collaborative decision-making unit are sent to the execution-driven unit. Each mechanism executes according to the corresponding parameters, and the zonal differentiated sandblasting is achieved by using a single spray gun path planning method.

[0073] S5.1 Three-Axis Servo Motion Module Control: Based on , and control quantities for each zone , Combined with the acceleration of the spray gun movement The system employs a PID closed-loop control algorithm (PID parameters: proportional coefficient Kp=5.0, integral coefficient Ki=0.1, derivative coefficient Kd=0.5, calibrated experimentally) to control the movement of the spray gun in the X / Y / Z three-dimensional space. Zone parameter switching is triggered by path nodes. Specifically, the system pre-sets the coordinate ranges of the fingertips, palm, and wrist in the control system. When the spray gun moves to the boundary node of the corresponding zone, the system automatically triggers the corresponding zone's response. , Parameter switching with a response time of ≤1ms enables precise full-range control and zone-specific adjustment of spray distance and moving speed, with a positioning accuracy of ±0.08mm and a speed tracking error of ≤0.8%.

[0074] S5.2 Electrical proportional pressure regulating valve control: based on and control quantities for each zone An adaptive pressure regulation algorithm is adopted to stabilize the air supply pressure of the sandblasting pipeline in a closed loop, while also taking into account the air pressure fluctuation coefficient. The pressure regulation response speed is adjusted in real time, with an adjustment accuracy of ±0.01MPa and a pressure fluctuation range of ≤0.005MPa.

[0075] S5.3 Sand Mass Flow Controller Control: Based on and control quantities for each zone Combined with the uniformity of sand particle size distribution The closed-loop control of the sand supply in the sand conveying pipeline stabilizes the sand flow rate at the final sand flow rate value, with a control accuracy of ±1g / s and a flow rate fluctuation of ≤0.1g / s.

[0076] S5.4 Glove Tensioning Mechanism Control: Maintains a constant workpiece surface shape to avoid deformation-induced deviations in detection and calculation; a new real-time surface shape monitoring module is added, which, combined with distance data collected by a laser displacement sensor, adjusts the tension force in real time to ensure tension fluctuation ≤3%, ensuring consistent surface shape in all areas of the glove and guaranteeing uniformity of zoned sandblasting.

[0077] S5.5 Zone Control Module Linkage: Coordinates the three-axis servo motion module, electrical proportional pressure regulating valve, and sand mass flow controller, employing a synchronous triggering mechanism to ensure that when the spray gun enters the target area, the corresponding zone... , , , Parameters are switched synchronously to avoid poor transition at partition boundaries; a 5mm transition section is used at partition boundaries, and parameters change linearly within the transition section to ensure uniform roughness transition.

[0078] It should be noted that the regulation process is as follows:

[0079] System initialization: After the system is powered on, sensor calibration and reference parameter loading are completed, experimental calibration coefficients and PID parameters are loaded, and zeroing calibration of each mechanism is completed;

[0080] Synchronous acquisition of parameters: The physical parameters of the entire domain are synchronously acquired with a fixed control cycle of 10ms. Steps S1.1 to S1.9 are executed to synchronously complete the coordinate registration of the glove area.

[0081] Data preprocessing: The collected parameters are sent to the data preprocessing unit, and steps S2.1 to S2.3 are executed to complete the data normalization and output the spatiotemporally normalized dataset;

[0082] Model collaborative computation: The normalized data is sent to the model collaborative decision-making unit, and the computation is completed by executing steps S3.1 to S3.3, outputting the final process control quantity and zoning control quantity;

[0083] Actuator action: The final process control quantity and zone control quantity are sent to the execution drive unit to execute steps S5.1 to S5.5 to complete the sandblasting operation, and the roughness data during the operation is collected simultaneously;

[0084] Closed-loop iteration: The roughness data collected in real time is fed back to the data preprocessing unit, and the process returns to step S6.2 for iterative iteration until the uniformity quantization coefficient meets the requirements. And each partition The single-workpiece processing flow is terminated; the system convergence is verified by experiments, and stability is achieved with ≤20 iterations.

[0085] Taking the sandblasting process of nitrile protective gloves as an example, the preset benchmark parameters (calibrated experimentally) are as follows: , , , , , (New nozzle no-load test calibration) (Standard particle size sand blasting test calibration), zoning weight coefficient , , .

[0086] Experimental conditions: ambient temperature 25±5℃, relative humidity 60±5%RH, sand particle size 0.1~0.3mm, nozzle model φ8mm, hand clamping using a special clamp, coordinate registration completed through 3 feature points on the wrist after each clamping; the experimental sample size is 50 pieces, divided into 3 groups of different working conditions (working condition 1: standard working condition, working condition 2: high humidity working condition, working condition 3: nozzle wear working condition), 15 pieces in each group, and the remaining 5 pieces are verification samples.

[0087] Actual measured data collected under operating condition 1 (standard operating condition): , , Curvature of the measured surface rate of change of curvature Pressure fluctuation coefficient uniformity of sand particle size distribution Comprehensive environmental disturbance coefficient Nozzle wear thickness Pressure, sound pressure, and frequency offset are all within the normal fluctuation range.

[0088] Perform steps S2.1 to S2.3 to complete data preprocessing and obtain a spatiotemporally warped dataset; perform steps S3.1 to S3.3 to calculate the global average roughness. Average roughness of the partition , , Uniformity quantization coefficient , partition uniformity correction coefficient , , The corrected curvature was obtained by calculation. Adaptive initial parameters , , Partition correction parameters , , Etc; Quantification yields the corrected perturbation equivalent , , The final output includes global control variables and zone control variables.

[0089] Execute steps S5.1 to S5.5, the drive unit operates according to the final control value and the zone control value, and after 12 control cycle iterations, the uniformity quantization coefficient is determined. The surface roughness is stabilized to above 0.96, and the roughness deviation of each partition $U_i$ is stabilized to above 0.93. The overall surface roughness deviation of the workpiece does not exceed ±0.15μm, and the surface roughness deviation of each partition does not exceed ±0.1μm.

[0090] Comparative experimental results: Compared with the existing fixed-parameter sandblasting system, the standard deviation of the surface roughness of the gloves processed by this system decreased from 0.32μm to 0.15μm, and the uniformity was improved by 53.1% (see the measured data comparison table below for specific comparison data); the disturbance compensation response time decreased from 50ms to 35ms, and the response speed was improved by 30%. These data are based on the statistical results of 3 sets of working conditions and 50 samples. After significance testing (P<0.05), the difference is statistically significant.

[0091] Core coefficient calibration data table

[0092]

[0093]

[0094]

[0095] Sensor calibration data sheet

[0096]

[0097] Comparison table of measured data (Working condition 1: Standard working condition)

[0098]

[0099] IV. Core Data of Iterative Curves During Convergence (Case 1)

[0100]

[0101] Note: 1. All calibration experiments were repeated 3 times, and the average value was taken as the final calibration value. The experimental repeatability error was ≤2%; 2. The sample size of the comparative experiment was 50 pieces. After significance test (P<0.05), the difference between the detection results of this system and the existing system was statistically significant.

[0102] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-sensor data fusion based glove sandblasting uniformity intelligent control system, characterized in that, include: The system comprises a sensing and acquisition unit, a data preprocessing unit, a model collaborative decision-making unit, and an execution driving unit, which are sequentially connected to form a closed-loop control link. The sensing and acquisition unit is used to synchronously collect multi-source physical parameter data of the entire sandblasting operation. The multi-source physical parameter data includes at least the distance between the spray gun and the glove surface, air supply pressure, sand flow rate, glove surface roughness, ambient temperature and humidity, sand moisture content, sandblasting impact sound pressure, spray gun movement speed, and nozzle wear thickness. The data preprocessing unit is used to perform noise reduction, outlier removal and spatiotemporal alignment on the multi-source physical parameter data, and output a spatiotemporally regularized dataset. The model collaborative decision-making unit includes a sandblasting uniformity quantification model, a surface curvature adaptive parameter model, and a multi-condition collaborative compensation model. The model collaborative decision-making unit is used to sequentially input the spatiotemporal regularized dataset into the sandblasting uniformity quantification model, the surface curvature adaptive parameter model, and the multi-condition collaborative compensation model for three-level correlation operations, and output the final process control quantity and zoning control quantity. The execution drive unit is used to control the action of the sandblasting actuator according to the final process control quantity and the zone control quantity, so as to realize the differentiated sandblasting control of the whole area and the zone.

2. The multi-sensor data fusion based glove sandblasting uniformity intelligent control system according to claim 1, wherein, The sensing and acquisition unit includes: a laser displacement sensor, a gas pressure sensor, a sand flow sensor, an industrial area array camera, a temperature and humidity sensor, a sand moisture content sensor, an acoustic sensor, a servo encoder, and an online nozzle wear detection sensor.

3. The multi-sensor data fusion based glove sandblasting uniformity intelligent control system according to claim 1, wherein, The data preprocessing unit includes a wavelet threshold denoising module, which is used to perform wavelet decomposition on the original data sequences output by each sensor and denoise them using an adaptive threshold. The 3σ criterion outlier removal module is used to introduce a dynamic deviation coefficient to remove outliers from the denoised data sequence. The spatiotemporal synchronization and alignment module is used to map all processed data to the same time beat and match the spatial coordinate correspondence, forming a spatiotemporally regularized dataset, using the interrupt clock of the servo control system as a unified reference.

4. The multi-sensor data fusion based glove sandblasting uniformity intelligent control system according to claim 1, wherein, The sand blasting uniformity quantification model is used for calculating a global uniformity quantification coefficient U and a respective partition uniformity correction coefficient according to the partition roughness data in the spatiotemporal regularized dataset and starting a subsequent model operation when U is less than a standard defined value.

5. The multi-sensor data fusion based glove sandblasting uniformity intelligent control system according to claim 4, wherein, The curved surface curvature adaptive parameter model is used for receiving the uniformity quantization coefficient U, the partition uniformity correction coefficient and distance data and spatial coordinate data in the spatiotemporal regular data set, obtaining the curved surface curvature K of the glove surface through curved surface fitting, and combining a preset material reference parameter to solve the curved surface adaptive initial process parameter and the partition correction parameter.

6. The multi-sensor data fusion based glove sandblasting uniformity intelligent control system according to claim 5, wherein, The multi-condition collaborative compensation model is used to simultaneously receive the uniformity quantization coefficient U, the initial process parameters and zoning correction parameters for surface adaptation, and the pressure data, sound pressure data, temperature and humidity data, moisture content data and sound wave frequency data in the spatiotemporal regularized dataset. It quantifies the nozzle wear equivalent, environmental interference coefficient and sand particle size interference coefficient respectively, and outputs the final process control quantity and zoning control quantity based on the initial process parameters for surface adaptation after collaborative correction.

7. The multi-sensor data fusion based glove sandblasting uniformity intelligent control system according to claim 1, wherein, The execution drive unit includes: a three-axis servo motion module for controlling the movement of the spray gun according to the final moving speed and the zone moving speed; an electric proportional pressure regulating valve for controlling the pipeline pressure according to the final air supply pressure and the zone air supply pressure; and a sand mass flow controller for controlling the sand supply amount according to the final sand flow rate and the zone sand flow rate. The glove tensioning mechanism is used to maintain a constant curvature of the glove. The partition control module is used to coordinate the synchronous switching and linear transition of parameters among various actuators at the partition boundaries.

8. The multi-sensor data fusion based glove sandblasting uniformity intelligent control system according to claim 1, wherein, In the closed-loop control link, the surface roughness data of the glove after sandblasting is collected in real time by an industrial area array camera and fed back to the data preprocessing unit, and the process is iterated until the uniformity quantization coefficient meets the set threshold.

Images