A method for remote monitoring of the working state of a paint spray gun

Abstract

The present application belongs to the technical field of industrial automation and intelligent manufacturing, and specifically relates to a coating spray gun working state remote monitoring method, comprising: deploying multi-module sensor to collect workshop environment data and pre-processing; modeling the initial environment data using environmental influence and superimposing correction with standard process parameters; collecting spray gun working state data and pre-processing, synchronously fusing initial environment data and spray gun initial state data; adopting machine learning analysis method to infer and calculate the fused state data, and generating monitoring report and early warning information; adopting closed-loop control and rule mapping measures on the state evaluation result to generate disposal instructions and issue to the execution terminal; the present application can remotely and real-time monitor and evaluate the working state of the coating spray gun, and automatically output executable disposal instructions, thereby reducing the dependence on manual inspection, reducing quality defects and downtime risks, and improving spraying consistency and production stability.

Description

Technical Field

[0001] This invention belongs to the field of industrial automation and intelligent manufacturing technology, and specifically relates to a method for remote monitoring of the working status of a paint spray gun. Background Technology

[0002] Chinese patent application number CN202011289317.3 discloses an intelligent sensing adhesive application control system. This invention includes a host computer controller, a display, an adhesive applicator, and an industrial robot. The control system incorporates a vision module, a force control module, and an intelligent decision-making module for adhesive application assembly process parameters. The vision module acquires part specification information, and the intelligent decision-making module retrieves adhesive application assembly process parameters. Furthermore, the vision module obtains part position information to determine the part's assembly location. The force control module communicates with the host computer controller to acquire force and torque information of the part during the adhesive application assembly process, and the force control algorithm adjusts the industrial robot's posture to complete the assembly.

[0003] While existing spray gun monitoring technologies in the automotive painting industry have achieved basic parameter display and threshold alarms for spraying pressure and flow rate, and can maintain basic production stability through manual inspections, the working status of spray guns in high-speed continuous production workshops with multiple robotic arms is affected by a combination of factors, including changes in ambient temperature and humidity, fluctuations in paint viscosity and temperature, gradual nozzle clogging, fluctuations in feed pressure, and spray angle drift. This can easily lead to quality problems such as uneven coating thickness, orange peel, and sagging. Existing technologies generally use single threshold alarms and manual inspections, which cannot distinguish between normal fluctuations and gradual abnormal trends in real time, and are difficult to identify problems such as slow pressure drops, unstable flow rates, and early signs of clogging. At the same time, existing technologies lack a mechanism to correlate spray gun operating data with environmental conditions, paint quality, and coating results. This makes it difficult to quickly locate the root cause after quality anomalies occur, and it cannot output clear fault diagnoses and actionable maintenance and adjustment paths, resulting in delayed response, low troubleshooting efficiency, and a high risk of line downtime.

[0004] To address the aforementioned issues, this invention proposes a remote monitoring method for the working status of paint spray guns. This method involves multi-source real-time acquisition and time-synchronized processing of environmental data from the spraying station, working status data of the spray gun itself, and paint quality data to obtain a unified data stream. The data stream is then analyzed using feature fusion and machine learning, and features such as pressure, flow rate, angle, viscosity, temperature, and humidity are modeled and inferred to obtain spray gun health assessment results, abnormal trend prediction results, and risk warning information. The analysis results are then remotely evaluated and controlled via cloud computing to obtain automatic adjustment commands for spraying parameters. This enables early warning before defects form and allows for root cause identification to guide parameter adjustment and maintenance. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] This invention provides a method for remote monitoring of the working status of paint spray guns, aiming to solve the following problems:

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] A method for remotely monitoring the working status of a paint spray gun includes:

[0009] Step S1: Deploy multi-module sensors to collect environmental data in the workshop, obtain raw environmental data, preprocess the raw environmental data, and generate initial environmental data.

[0010] Step S2: Based on the initial environmental data, the initial environmental data is modeled using environmental impact and superimposed with standard process parameters to obtain the target spraying parameters after environmental compensation.

[0011] Step S3: Based on the target spraying parameters, collect the working status data of the spray gun, obtain the original status data of the spray gun and preprocess it to generate the initial status data of the spray gun. Synchronously fuse the initial environmental data and the initial status data of the spray gun to generate fused status data.

[0012] Step S4: Based on the fused status data, machine learning analysis methods are used to perform inference calculations on the fused status data to obtain status assessment results and generate monitoring reports and early warning information.

[0013] Step S5: Based on the monitoring report and early warning information, take closed-loop control and rule mapping measures on the status assessment results, generate disposal instructions and send them to the execution terminal.

[0014] As a preferred embodiment, the specific steps for deploying multiple module sensors to collect environmental data in the workshop and obtain raw environmental data are as follows:

[0015] By deploying multi-module sensors at each spraying station in the spraying workshop, temperature sensors, humidity sensors, and air pressure sensors are installed in the spray gun operation area, material supply area, and return air area, and environmental data of the workshop is collected according to a unified sampling period to obtain raw environmental data.

[0016] As a preferred embodiment, the specific steps for preprocessing the raw environmental data to generate initial environmental data are as follows:

[0017] By timestamping the raw environmental data, timestamps are added to the data from each sensor according to the unified clock of the acquisition gateway, and sampling is aligned. Noise filtering is performed on the raw environmental data, and median filtering is used to eliminate noise from high-frequency jitter data. Outlier removal is performed on the raw environmental data, and drift points and abrupt jumps are identified and removed based on physically reasonable intervals and abrupt change thresholds. Missing data is filled in by using linear interpolation for short-term missing data and valid values ​​for continuous missing data, generating initial environmental data including temperature, humidity, and air pressure.

[0018] As a preferred embodiment, the specific steps of modeling the initial environmental data using environmental impact based on the initial environmental data, and superimposing and correcting it with standard process parameters to obtain the environmentally compensated target spraying parameters are as follows:

[0019] The initial environmental data is used to model environmental impact, establish the correspondence between environmental parameters and spraying process parameters, and output environmental compensation. The specific steps of environmental impact modeling are as follows: determine the standard process parameter set based on the established process benchmark, establish the standard environmental condition range as the benchmark, process the initial environmental data in segments, and statistically analyze the temperature, humidity and air pressure in a fixed time window to generate the current environmental state quantity.

[0020] As a preferred embodiment, the specific steps of modeling the initial environmental data based on environmental impact and superimposing and correcting it with standard process parameters to obtain the environmentally compensated target spraying parameters further include:

[0021] By using an online environmental impact coefficient identification algorithm to calculate the initial environmental data, the correspondence between environmental parameters and spraying process parameters is established and the environmental compensation amount is output, so that the standard process parameters are corrected to the target spraying parameters that are suitable for the current environment.

[0022] The definition of the online identification algorithm for environmental impact coefficients is as follows:

[0023] T1 is constructed using temperature (TEP), humidity (WET), and air pressure (PAR) under standard environmental conditions as a benchmark to construct an environmental deviation vector. and compensation vector , where X is the environmental deviation vector. The median temperature. The median humidity. Here, T represents the median air pressure, t is the transpose sign, t is the time index, and Con is the compensation vector. As a form of stress compensation, For traffic compensation, For atomization compensation;

[0024] T2, based on the environmental deviation vector and compensation vector, obtains the environmental impact coefficient matrix through recursive identification, calculates the real-time environmental compensation amount, and then superimposes the environmental compensation amount with the standard process parameter set to obtain the target spraying parameters. The update formulas for the recursive identification are as follows:

[0025] ,

[0026] Where AD is the gain vector, t is the time index, P is the covariance matrix, and X is the environmental bias vector. The forgetting factor is T, and T is the transpose sign.

[0027] ,

[0028] Where EN is the environmental impact coefficient matrix, and t is the time index. The compensation vector is calculated in real time, where Con is the compensation vector, X is the environmental deviation vector, and AD is the gain vector.

[0029] ,

[0030] Where P is the covariance matrix and t is the time index. Let AD be the forgetting factor, X be the environmental bias vector, and T be the transpose sign.

[0031] T3, by superimposing and correcting the standard process parameter set, adds the real-time environmental compensation amount to the standard spraying pressure, standard spraying flow rate and standard atomization parameters respectively to obtain the target spraying parameters; by superimposing and correcting the environmental compensation coefficient with the standard process parameters, by applying the corresponding compensation amount to the standard spraying pressure, standard spraying flow rate and standard atomization parameters respectively and performing boundary limiting, the environmentally compensated target spraying parameters are generated, and the target spraying parameters are output to the subsequent spray gun status monitoring and synchronization fusion steps.

[0032] As a preferred implementation, based on the target spraying parameters, the original state data of the spray gun is obtained by collecting the working state data of the spray gun. Specifically, pressure sensors, flow sensors, and angle or attitude sensors are respectively installed in the spray gun feeding channel, atomization channel, and attitude execution part. Spraying pressure, spraying flow, and spraying angle data are collected at a uniform sampling period, and the collected data are aggregated through a data acquisition gateway to form the original state data of the spray gun. The original state data of the spray gun is preprocessed to generate the initial state data of the spray gun. Specifically, the original state data of the spray gun is timestamped and sampled and aligned to ensure that the pressure, flow, and angle data are on the same time reference. The original state data of the spray gun is denoised and median filtering is used to eliminate random jitter. The original state data of the spray gun is outlier processed by removing abrupt jumps based on the sensor range boundaries and mutation thresholds, and short-term missing data is filled by interpolation to obtain continuous and stable initial state data of the spray gun.

[0033] As a preferred embodiment, the specific steps for synchronously fusing the initial environmental data and the initial state data of the spray gun to generate fused state data are as follows:

[0034] By synchronously fusing the initial environmental data and the initial state data of the spray gun, fused state data is generated. Specifically, the synchronous fusing measures are as follows: the initial environmental data and the initial state data of the spray gun are aligned and matched in units of time windows. The ambient temperature, ambient humidity, ambient air pressure, spraying pressure, spraying flow rate, and spraying angle within the same time window are combined into the same fused record. The target spraying parameters are written into the fused record as a reference benchmark to characterize the deviation relationship between the target and the actual situation. Fusion state data containing environmental quantities, spray gun state quantities, and target spraying parameters is generated and used for state assessment and early warning generation.

[0035] As a preferred implementation, the specific steps of using machine learning analysis methods to perform inference calculations on the fused state data based on the fused state data, obtaining state assessment results, and generating monitoring reports and early warning information are as follows:

[0036] Based on fused status data, a status assessment result is obtained by applying machine learning analysis and rule-based output measures to the fused status data, and a monitoring report and early warning information are generated. The implementation of the machine learning analysis is as follows: the fused status data is segmented according to a fixed time window, and the average pressure, average flow rate, average angle, pressure change rate, flow fluctuation amplitude, angle drift, and deviation between the target spraying parameters and the actual spray gun parameters within the window are extracted to generate a status feature vector; the status feature vector is normalized and feature concatenation is performed to form the model input; the model input is sent to the cloud inference module, which calls the pre-trained spray gun status assessment model to perform inference calculations, and outputs a status assessment result including a health score, anomaly type identifier, anomaly level identifier, and risk score, and binds the status assessment result with the corresponding timestamp, workstation number, and spray gun number for storage.

[0037] As a preferred embodiment, the specific steps of using machine learning analysis methods to perform inference calculations on the fused state data based on the fused state data, obtaining state assessment results, and generating monitoring reports and early warning information further include:

[0038] The method for generating early warning information is as follows: by applying a residual trend early warning algorithm to the fusion state data and target spraying parameters, a trend early warning score is obtained and an early warning information is generated. When the triggering conditions are met, the early warning level, early warning reason, triggering time, associated spray gun number and key evidence indicators are output.

[0039] The residual trend early warning algorithm is defined as follows:

[0040] R1, based on the target spraying parameters Compared with actual spray gun parameters By performing a difference construction, the residual vector is obtained as follows: Furthermore, by normalizing the residual vector using standard deviation, a normalized residual vector is obtained, where M is the residual vector. This refers to the actual spraying pressure. For the target spraying pressure, This represents the actual spraying flow rate. For the target spraying flow rate, These are the actual atomization parameters. Here are the target atomization parameters, t is the time index, and T is the transpose sign;

[0041] R2, based on the normalized residual vector, obtains the smoothed residual statistic by performing an exponentially weighted shift statistic on the normalized residual vector. Furthermore, cumulative statistics are performed on the smoothed residuals to obtain the cumulative asymptotic anomalies. Where S is the smoothing residual statistic, Here, L is the smoothing coefficient, and L is the normalized residual vector. This represents the gradual accumulation of abnormalities. The function is a non-negative cutoff function, and e is the drift tolerance term;

[0042] R3, calculated by combining the smoothed residual statistic and the cumulative asymptotic anomaly, yields the trend warning score. Risk is the trend warning score. Let the L2 norm represent the overall strength of the vector. The first norm represents the sum of the absolute values ​​of vectors.

[0043] As a preferred implementation, the specific steps of taking closed-loop control and rule mapping measures based on the monitoring report and early warning information, generating disposal instructions, and issuing them to the execution terminal are as follows:

[0044] Based on monitoring reports and early warning information, closed-loop control and rule mapping measures are implemented on the status assessment results to generate disposal instructions and send them to the execution terminal. The specific steps of the closed-loop control and rule mapping measures are as follows: key deviation indicators, anomaly type identifiers, and anomaly level identifiers in the monitoring report are read to form disposal trigger conditions, and the trigger conditions are mapped to a set of disposal actions according to a preset disposal rule library; the disposal action set is constrained and verified, upper and lower limits are applied to the adjustment amount, and workstation interlock conditions are verified for shutdown actions, and disposal instructions containing target values, adjustment steps, execution times, and execution priorities are generated. The disposal instructions are then sent to the execution terminal via the industrial wireless network; after receiving the disposal instructions, the execution terminal completes parameter setting or maintenance action triggering, and sends the execution receipt, post-execution spray gun status data, and disposal result markers back to the cloud, realizing a closed-loop disposal process of assessment—disposal—receipt—reassessment.

[0045] Beneficial effects

[0046] 1. By collecting workshop environmental data in real time and modeling and compensating for environmental impact, the target spraying parameters after environmental compensation are obtained. These parameters are used to dynamically correct standard process parameters under conditions of temperature, humidity, and air pressure fluctuations. This solves the problem of coating consistency decline caused by spraying parameter drift due to environmental changes, and forms a definite technical effect of maintaining stable spraying even under environmental changes.

[0047] 2. By synchronizing and fusing the spray gun's working status data with the initial environmental data in time, fused status data is generated. This data is used to establish a deviation benchmark between the target spraying parameters and the actual spray gun parameters and to provide traceable parameter evidence. This solves the problem of fragmented data in existing technologies, which makes it difficult to quickly locate the root cause of quality anomalies, thus forming a "traceable and locatable" deterministic technical effect.

[0048] 3. By performing cloud-based inference and evaluation on the fused status data and outputting monitoring reports and early warning information, closed-loop control and rule mapping measures are adopted on the status evaluation results to generate disposal instructions and issue them to the execution terminal. This solves the problems of delayed response and low investigation efficiency caused by the reliance on manual inspection and experience-based handling in existing technologies, and forms a deterministic technical effect of automatic early warning, automatic handling and closed-loop review. Attached Figure Description

[0049] Figure 1 This is a flowchart of the present invention.

[0050] Figure 2 This is a comparison chart of the technical effects of the present invention, in which the black bars represent the present invention and the gray bars represent the prior art. Detailed Implementation

[0051] To make the technical means, creative features, and achieved objectives and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention. Unless otherwise specified, the experimental methods in the following embodiments are conventional methods, and the materials and reagents used in the following embodiments are commercially available unless otherwise specified.

[0052] Example 1 combined Figure 1 The flowchart shown below illustrates a method for remotely monitoring the working status of a paint spray gun. The specific implementation steps are as follows:

[0053] Step S1: Deploy multi-module sensors to collect environmental data in the workshop, obtain raw environmental data, preprocess the raw environmental data, and generate initial environmental data.

[0054] Step S1 addresses the issues of missing workshop environmental data, high data noise, and inconsistent timing leading to unreliable subsequent analyses. It obtains initial environmental data through standardized data collection and preprocessing.

[0055] Specifically, by deploying multi-module sensors at each spraying station in the spraying workshop, temperature, humidity, and air pressure sensors are installed in the spray gun operating area, material supply area, and return air area, respectively, and environmental data of the workshop is collected according to a unified sampling period to obtain raw environmental data. The raw environmental data is timestamped by adding timestamps to the data of each sensor according to the unified clock of the acquisition gateway and aligning the samples. The raw environmental data is denoised by applying median filtering to eliminate noise from high-frequency jitter data. Outliers are removed by identifying and removing drift points and jump points based on physically reasonable intervals and abrupt change thresholds. Missing data is filled by using linear interpolation for short-term missing data and valid values ​​for continuous missing data to generate initial environmental data containing temperature, humidity, and air pressure, and outputting a standardized environmental time series for subsequent modeling.

[0056] The physical reasonable range and mutation threshold are determined based on the sensor range, the allowable operating conditions of the spraying workshop, and historical stable production data to determine the upper and lower limits of each environmental parameter as the physical reasonable range, and the mutation threshold is set according to the range of parameter changes at adjacent sampling times. When the sampled value exceeds the physical reasonable range, it is determined to be invalid data and discarded. When the difference between adjacent sampled values ​​exceeds the mutation threshold, it is determined to be a sudden jump and discarded or replaced with an adjacent valid value.

[0057] Step S2: Based on the initial environmental data, the initial environmental data is modeled using environmental impact and superimposed with standard process parameters to obtain the target spraying parameters after environmental compensation.

[0058] Step S2 is used to solve the problem that environmental fluctuations directly cause spraying parameter drift and the same process has inconsistent effects under different environments. By modeling environmental impact and superimposing parameters for correction, the target spraying parameters after environmental compensation are obtained.

[0059] Specifically, environmental impact modeling is performed on the initial environmental data to establish the correspondence between environmental parameters and spraying process parameters, and to output environmental compensation amounts. The specific steps of environmental impact modeling are as follows: Based on a predetermined process baseline, a set of standard process parameters is determined to establish a standard environmental condition range as a baseline. The initial environmental data is segmented, with a preset sampling period and preset time window used to divide the initial environmental data into multiple adjacent time periods. Within each time period, environmental temperature, humidity, and air pressure are statistically calculated, including at least the mean, maximum, minimum, and fluctuation range. The system generates environmental state quantities for the corresponding time period based on the initial environmental data, which are then used for subsequent environmental impact modeling and calculation of target spraying parameters. An online environmental impact coefficient identification algorithm is used to calculate the initial environmental data, establishing a correspondence between environmental parameters and spraying process parameters and outputting environmental compensation quantities. These compensation quantities are used to convert the current environmental conditions into correction quantities for the spraying process parameters, thereby modifying the standard process parameters to adapt to the target spraying parameters for the current environment. For example, by superimposing the environmental compensation quantities onto the standard spraying pressure, standard spraying flow rate, and standard atomization parameters, the environmentally compensated target spraying parameters are obtained, which are used to maintain atomization and film formation stability under the current temperature, humidity, and air pressure conditions.

[0060] The environmental impacts refer to the effects of changes in environmental parameters in the spraying workshop on spraying process parameters and spraying effects. These environmental parameters include ambient temperature, ambient humidity, and ambient air pressure. The effects are manifested in the following ways: changes in ambient temperature cause changes in paint flowability and atomization state, thus affecting spraying flow rate and atomization parameters; changes in ambient humidity cause changes in the film-forming process, thus affecting spraying pressure and atomization parameters; and changes in ambient air pressure cause changes in atomized gas density, thus affecting atomization parameters and spraying pressure. For example, when the ambient temperature rises, paint flowability increases, leading to a larger output per unit time. The system maintains stable spray width and particle size by reducing the target spraying flow rate or atomization parameters. Conversely, when the ambient temperature decreases, paint flowability decreases, leading to increased output resistance. The system improves output resistance by increasing the spraying flow rate or atomization parameters. The system aims to achieve the target spray flow rate or increase atomization parameters to ensure the film thickness meets process requirements. When ambient humidity increases, the film formation process is more susceptible to moisture, increasing the risk of surface defects. The system improves atomization uniformity and adhesion stability by increasing atomization intensity and compensating for this. When ambient humidity decreases, the film formation process is more prone to surface drying. The system reduces the risk of particle coarsening by decreasing atomization intensity or adjusting spray pressure. When ambient air pressure decreases, the atomized gas density decreases, weakening the atomization effect. The system maintains a fine atomized particle effect by increasing atomization parameters and compensating for this by increasing spray pressure. When ambient air pressure increases, the atomized gas density rises, leading to excessive atomization. The system prevents scattering and film thickness fluctuations by decreasing atomization parameters and providing reverse compensation for this.

[0061] The definition of the online identification algorithm for environmental impact coefficients is as follows:

[0062] T1 is constructed using temperature (TEP), humidity (WET), and air pressure (PAR) under standard environmental conditions as a benchmark to construct an environmental deviation vector. and compensation vector , where X is the environmental deviation vector. The median temperature. The median humidity. Here, T represents the median air pressure, t is the transpose sign, t is the time index, and Con is the compensation vector. As a form of stress compensation, For traffic compensation, For atomization compensation;

[0063] T2, based on the environmental deviation vector and compensation vector, obtains the environmental impact coefficient matrix through recursive identification, calculates the real-time environmental compensation amount, and then superimposes the environmental compensation amount with the standard process parameter set to obtain the target spraying parameters. The update formulas for the recursive identification are as follows:

[0064] ,

[0065] Where AD is the gain vector, t is the time index, P is the covariance matrix, and X is the environmental bias vector. The forgetting factor is T, and T is the transpose sign.

[0066] ,

[0067] Where EN is the environmental impact coefficient matrix, and t is the time index. The compensation vector is calculated in real time, where Con is the compensation vector, X is the environmental deviation vector, and AD is the gain vector.

[0068] ,

[0069] Where P is the covariance matrix and t is the time index. Let AD be the forgetting factor, X be the environmental bias vector, and T be the transpose sign.

[0070] T3 corrects the standard process parameter set by superimposing the real-time environmental compensation amount onto the standard spraying pressure, standard spraying flow rate, and standard atomization parameters to obtain the target spraying parameters; it then corrects the environmental compensation coefficient by superimposing it onto the standard process parameters, applying corresponding compensation amounts to the standard spraying pressure, standard spraying flow rate, and standard atomization parameters, and applying boundary limits to generate the environmentally compensated target spraying parameters, which are then output to the subsequent spray gun status monitoring and synchronization fusion steps.

[0071] In one embodiment, the fresh air system in the automotive painting workshop is turned on after the morning shift change, causing environmental changes at the painting station in a short period of time: the ambient temperature rises from the reference temperature, the ambient humidity decreases from the reference humidity, and the ambient air pressure fluctuates. If the painting is still carried out according to the standard painting pressure, standard painting flow rate, and standard atomization parameters in the established process reference, quality risks such as dry painting edges, coarser atomized particles, and increased fluctuations in coating thickness may occur. The common practice on site is for operators to adjust the pressure and flow rate according to their experience. However, since the environmental changes are continuous and different spray guns respond differently, manual parameter adjustment has the problems of response lag and inconsistent compensation, which can easily cause coating consistency differences for the same model at different time periods.

[0072] The online environmental impact coefficient identification algorithm of this invention calculates the current environmental deviation vector in real time and automatically obtains the environmental compensation amount by recursively updating the environmental impact coefficient matrix. This compensation amount is then superimposed on the standard spraying pressure, standard spraying flow rate, and standard atomization parameters to generate the environmentally compensated target spraying parameters, ensuring that the spray gun maintains stable atomization and film formation conditions during environmental changes. Therefore, the online environmental impact coefficient identification algorithm solves the problems of spraying parameter drift caused by environmental fluctuations and slow response and inconsistency across spray guns due to reliance on manual experience compensation in practical scenarios. Furthermore, it reduces the probability of defects such as uneven coating thickness, orange peel, and sagging caused by sudden environmental changes.

[0073] The established process benchmarks are process parameter benchmarks formed and solidified during the selection and process finalization stages of spraying equipment, based on specified vehicle models, parts, coating systems and spraying stations. They are determined by process specifications or equipment calibration results, including standard environmental condition ranges and standard spraying pressure, standard spraying flow rate, standard atomization parameters and standard spraying angle within the standard environmental condition ranges. They are also established to correspond with vehicle model identification, station identification, spray gun model and coating formula identification, and are used as reference benchmarks for subsequent environmental compensation and condition assessment.

[0074] The set of standard process parameters includes at least standard spraying pressure, standard spraying flow rate, standard atomization parameters, and standard spraying angle.

[0075] Step S3: Based on the target spraying parameters, collect the working status data of the spray gun, obtain the original status data of the spray gun and preprocess it to generate the initial status data of the spray gun. Synchronously fuse the initial environmental data and the initial status data of the spray gun to generate fused status data.

[0076] Step S3 is used to solve the problem that the spray gun status data, environmental data, and process objectives cannot be linked and the data is not synchronized, making it difficult to locate the root cause. By collecting, pre-processing and merging the spray gun status data, fused status data is generated to provide a unified input for diagnosis and prediction.

[0077] Specifically, based on the target spraying parameters obtained in step S2, the original state data of the spray gun is obtained by collecting the working state data of the spray gun. Specifically, pressure sensors, flow sensors, and angle or attitude sensors are respectively installed in the spray gun's feeding channel, atomization channel, and attitude execution part. Spraying pressure, spraying flow, and spraying angle data are collected at a unified sampling period, and the collected data is aggregated through a data acquisition gateway to form the original state data of the spray gun. Preprocessing is performed on the original state data of the spray gun to generate initial state data of the spray gun. Specifically, the original state data of the spray gun is timestamped and sampled to ensure that the pressure, flow, and angle data are on the same time reference. Noise reduction processing is performed on the original state data of the spray gun, and median filtering is used to eliminate random jitter. Outlier processing is performed on the original state data of the spray gun. Based on the sensor range boundary and mutation threshold, abrupt jumps are eliminated, and short-term missing data is filled by interpolation to obtain continuous and stable initial state data of the spray gun. Synchronous fusion of the initial environmental data and the initial state data of the spray gun is implemented to generate fused state data. Specifically, the synchronous fusion measures involve aligning and matching the initial environmental data and the initial state data of the spray gun in time windows, combining the ambient temperature, ambient humidity, ambient air pressure, spraying pressure, spraying flow rate, and spraying angle within the same time window into a single fused record, and writing the target spraying parameters as a reference into the fused record to characterize the deviation relationship between the target and the actual situation. This generates fused state data containing environmental quantities, spray gun state quantities, and target spraying parameters, which is then output to step S4 for state assessment and early warning generation.

[0078] Step S4: Based on the fused status data, machine learning analysis methods are used to perform inference calculations on the fused status data to obtain status assessment results and generate monitoring reports and early warning information.

[0079] Step S4 addresses the problems of existing technologies being unable to distinguish between normal fluctuations and gradual anomalies, unable to identify early signs of blockage and quality risks, and having delayed alarms. It uses machine learning reasoning to calculate and obtain status assessment results and output monitoring reports and early warning information.

[0080] Specifically, based on the fused status data generated in step S3, machine learning analysis and rule-based output measures are applied to the fused status data to obtain status assessment results and generate monitoring reports and early warning information. The implementation of the machine learning analysis is as follows: by segmenting the fused status data according to a fixed time window, the average pressure, average flow rate, average angle, pressure change rate, flow fluctuation amplitude, angle drift, and deviation between the target spraying parameters and the actual spray gun parameters within the window are extracted to generate a status feature vector; by performing normalization and feature concatenation processing on the status feature vector, a model input is formed; by sending the model input to the cloud inference module, the pre-trained spray gun status assessment model is called to perform inference calculations, and a status assessment result including health score, abnormality type identifier, abnormality level identifier, and risk score is output, and the status assessment result is bound and stored with the corresponding timestamp, workstation number, and spray gun number.

[0081] The monitoring report is generated by taking structured arrangement measures on the status assessment results and fused status data to generate a monitoring report containing the current environmental status, target spraying parameters, actual spray gun parameters, deviation index, health score, anomaly type and anomaly level, and outputting it to the remote monitoring interface and data interface in a preset format.

[0082] The method for generating the early warning information is as follows: by applying a residual trend early warning algorithm to the fusion state data and the target spraying parameters, a trend early warning score is obtained and an early warning information is generated. This is used to identify progressive abnormal trends such as progressive nozzle blockage, slow decrease in spraying pressure, slight fluctuation in spraying flow rate, and gradual degradation of atomization performance before coating defects are formed. When the triggering conditions are met, the early warning level, early warning reason, triggering time, associated spray gun number, and key evidence indicators are output, thereby providing an executable basis for the subsequent step S5 to generate disposal instructions.

[0083] The triggering condition is that any component in the progressive abnormal accumulation reaches or exceeds the corresponding component threshold, and the component shows a unidirectional cumulative increase within a consecutive preset number of time windows; when any of the above conditions are met, an early warning is determined to be triggered, and the early warning level, early warning reason, trigger time, associated spray gun number and key evidence indicators are generated.

[0084] The residual trend early warning algorithm is defined as follows:

[0085] R1, based on the target spraying parameters Compared with actual spray gun parameters By performing a difference construction, the residual vector is obtained as follows: Furthermore, by normalizing the residual vector using standard deviation, a normalized residual vector is obtained, where M is the residual vector. This refers to the actual spraying pressure. For the target spraying pressure, This represents the actual spraying flow rate. For the target spraying flow rate, These are the actual atomization parameters. Here are the target atomization parameters, t is the time index, and T is the transpose sign;

[0086] R2, based on the normalized residual vector, obtains the smoothed residual statistic by performing an exponentially weighted shift statistic on the normalized residual vector. Furthermore, cumulative statistics are performed on the smoothed residuals to obtain the cumulative asymptotic anomalies. Where S is the smoothing residual statistic, Here, L is the smoothing coefficient, and L is the normalized residual vector. This represents the gradual accumulation of abnormalities. The function is a non-negative cutoff function, and e is the drift tolerance term;

[0087] R3, calculated by combining the smoothed residual statistic and the cumulative asymptotic anomaly, yields the trend warning score. Risk is the trend warning score. Let the L2 norm represent the overall strength of the vector. The first norm represents the sum of the absolute values ​​of vectors;

[0088] In one embodiment, during continuous production, a spray gun in an automotive painting workshop experiences a gradual accumulation of paint particles on its nozzle, causing the nozzle channel to narrow. Initially, this anomaly does not cause instantaneous changes in spray pressure or flow rate. Instead, it manifests as a slow decrease in spray pressure, slight fluctuations in flow rate, and a gradual deterioration in atomization. Due to the small magnitude and slow process, when using a traditional fixed threshold alarm method, the pressure and flow rate remain within the threshold range for extended periods, preventing the system from triggering an alarm. The anomaly is often only discovered on-site after quality issues such as uneven coating thickness, orange peel texture, or localized drying occur, increasing the risk of rework and line shutdown.

[0089] The residual trend early warning algorithm of this invention constructs residuals by analyzing the deviation between actual spray gun parameters and target spraying parameters, and performs smoothing statistics and continuous accumulation on the residuals to obtain a trend early warning score. When the nozzle gradually becomes clogged, causing the deviation to continue to increase, the trend early warning score increases continuously and triggers an early warning before the defect forms. At the same time, it outputs the cause of the early warning based on the dominant direction of the deviation accumulation, such as outputting a flow-related early warning or a pressure-related early warning. Thus, the residual trend early warning algorithm solves the problems in real-world scenarios where gradual anomalies are difficult to identify by threshold alarms, blockage precursors cannot be detected in advance, and quality defects occur first due to alarm lag. It also enables maintenance personnel to perform cleaning of nozzles or switching to a backup spray gun before defects appear.

[0090] The cloud-based inference module is a computing processing unit deployed on a cloud platform. It is connected to the workshop data acquisition gateway via network communication. It is used to receive model input formed by fused status data, and call the spray gun status assessment model to perform inference calculations, outputting health scores, anomaly type identifiers, anomaly level identifiers, and risk scores. The cloud-based inference module includes a data access unit, a feature processing unit, a model inference unit, and a result output unit. The data access unit is used to receive and cache data, the feature processing unit is used to normalize and concatenate the data, the model inference unit is used to perform forward calculations of the model, and the result output unit is used to write the inference results into the database and push them to the monitoring interface and early warning module.

[0091] The pre-trained spray gun condition assessment model is established during the offline training phase based on historical spray gun operation data, corresponding quality judgment results, and fault maintenance records. Its input consists of a state feature vector composed of pressure, flow rate, angle and its rate of change, fluctuation amplitude, drift amount, and the deviation between the target spraying parameters and the actual spray gun parameters. Its output includes a spray gun health score, anomaly type identifier, anomaly level identifier, and risk score. During the training phase, the model iteratively learns from labeled samples and solidifies the parameters on validation samples. After training, the model parameters are deployed to the cloud inference module for real-time condition assessment of online fused condition data.

[0092] The pre-trained spray gun status assessment model obtains a training sample set by constructing and labeling historical production data from the spraying workshop. This training sample set includes at least: environmental data, spray gun operating status data, target spraying parameters, coating quality judgment results, and maintenance records. The historical fusion status data is segmented into multiple sample segments within fixed time windows. Each sample segment corresponds to the same spray gun number, the same workstation number, and the same time window. By associating and labeling the quality judgment results with the maintenance records, a label is obtained for each sample segment: when the coating quality judgment for the corresponding sample segment is qualified and there is no maintenance record, it is labeled as normal; when nozzle cleaning, nozzle replacement, abnormal material supply handling, or shutdown maintenance records appear shortly after the sample segment, it is labeled as abnormal based on the maintenance reason. The system categorizes anomalies, such as insufficient pressure, abnormal flow, atomization degradation, angle deviation, and clogging trends. When a sample segment is deemed unqualified, the unqualified type and the cause of spray gun maintenance are used together to correct the anomaly category, ensuring that the label matches the actual root cause. Training input features are generated by extracting features from each sample segment. These features include at least: average pressure, average flow, average angle, pressure change rate, flow fluctuation amplitude, angle drift, and the deviation between the target spraying parameters and the actual spray gun parameters. Subsequently, training and validation are implemented for the samples, and supervised learning is used to iteratively update the model parameters, ensuring the model output matches the sample labels. When the validation set indicators meet preset requirements, the model parameters are solidified, resulting in a pre-trained spray gun status evaluation model, which is then deployed to the cloud inference module.

[0093] During operation, the model applies the same time window segmentation and feature extraction measures to the fused state data obtained in step S3 to generate a state feature vector as model input. The model input includes: current environmental state quantities: statistical values ​​of temperature, humidity, and air pressure; target spraying parameters: target pressure, target flow rate, and target atomization parameters; actual spray gun parameters: actual pressure, actual flow rate, actual angle / attitude, and atomization characterization; and the deviation between the target and the actual parameters and its variation characteristics: mean deviation, deviation slope, fluctuation amplitude, and drift. The above inputs enable the model to perform unified evaluation in scenarios where environmental changes, target parameter changes, and actual spray gun changes coexist, thereby avoiding misjudgment based solely on pressure or flow rate.

[0094] The output of the spray gun status assessment model is a status assessment result, including the following four items, which are bound to the spray gun number, workstation number, and timestamp: Health score: used to characterize the degree of deviation of the current working state of the spray gun from the stable benchmark; Anomaly type identifier: used to indicate whether the anomaly belongs to the pressure, flow, atomization, angle, or clogging trend category; Anomaly level identifier: used to distinguish between general anomalies and severe anomalies, and as the basis for whether to trigger action; Risk score: used to quantify quality risk and failure risk, providing a basis for early warning and action strategy selection; The above output results are used to generate a monitoring report, and together with the early warning information of the residual trend early warning algorithm, drive the generation of the action instruction in step S5;

[0095] The specific steps for iteratively updating model parameters using supervised learning are as follows: The model is trained using batch input of labeled training samples. The state feature vector of each sample is input into the spray gun state assessment model to obtain the model's predicted output. The predicted output is compared with the true label of the sample to calculate the prediction error. The true label includes a normal category label, an abnormal level label, and a corresponding risk interval label. Then, the influence of each model parameter on the error is obtained by backpropagation and differentiation of the prediction error. The model parameters are then updated using gradient descent, and the parameters are corrected according to a preset learning rate to reduce the prediction error of the next sample of the same type. The above process of inputting samples—calculating prediction—calculating error—backpropagation—parameter updating is repeated for all training batches, and the training set is traversed multiple times. After each training round, the model is evaluated using validation samples. When the validation effect reaches a preset index and no longer shows significant improvement over multiple rounds, the iteration is stopped and the model parameters are fixed, thus obtaining the pre-trained spray gun state assessment model.

[0096] Step S5: Based on the monitoring report and early warning information, take closed-loop control and rule mapping measures on the status assessment results, generate disposal instructions and send them to the execution terminal;

[0097] Step S5 addresses the issues of relying on repeated manual investigations after an anomaly is discovered, difficulty in providing executable maintenance paths, and slow response. Through closed-loop control and rule mapping, the evaluation results are transformed into handling instructions and sent to the execution terminal to achieve rapid handling and stable quality.

[0098] Specifically, based on the monitoring report and early warning information generated in step S4, closed-loop control and rule mapping measures are implemented on the status assessment results to generate disposal instructions and send them to the execution terminal. The specific steps of the closed-loop control and rule mapping measures are as follows: by reading the key deviation indicators, abnormality type identifiers, and abnormality level identifiers in the monitoring report, disposal trigger conditions are formed, and the trigger conditions are mapped to a set of disposal actions according to the preset disposal rule library. The set of disposal actions includes at least spraying pressure adjustment, spraying flow adjustment, atomization parameter adjustment, spraying angle correction, spray gun cleaning, nozzle replacement prompt, switching to a backup spray gun, and shutdown maintenance. By constraining and verifying the set of disposal actions, upper and lower limits are applied to the adjustment amount, and workstation interlock condition verification is performed for shutdown actions. Disposal instructions containing target values, adjustment steps, execution times, and execution priorities are generated and sent to the execution terminal through the industrial wireless network. After receiving the disposal instructions, the execution terminal completes parameter setting or maintenance action triggering and sends the execution receipt, post-execution spray gun status data, and disposal result markers back to the cloud, realizing a closed-loop disposal process of assessment—disposal—receipt—reassessment.

[0099] The preset handling rule library is a set of rules pre-established and stored in the cloud platform or local controller. It is used to map the status assessment results and early warning information into executable handling actions. The handling rule library consists of process rules, equipment rules, and safety interlock rules, and is defined in the structure of trigger condition—handling action—action parameter—priority—constraint condition. The trigger condition includes at least an anomaly type identifier, an anomaly level identifier, a trend early warning score, a key deviation index, and its duration window. Handling actions include: spraying pressure adjustment, spraying flow adjustment, atomization parameter adjustment, spraying angle correction, spray gun cleaning, nozzle replacement prompt, switching to a backup spray gun, and shutdown for maintenance. Action parameters include at least a target value, adjustment step size, number of executions, and execution interval. Constraint conditions include: upper and lower limits of parameters, adjustment rate limit, workstation interlock conditions, and safety shutdown conditions, thereby ensuring that the handling instructions can be effectively executed for anomalies while meeting the requirements of production safety and process consistency.

[0100] Combination Figure 2 The figure shows a comparison of the technical effects of a remote monitoring method for the working status of a paint spray gun. The black bars represent the technical effects of the present invention, and the gray bars represent the technical effects of the prior art. Figure 2 It can be seen that the technical effect of the present invention is superior to that of the prior art.

[0101] Example 2, a method for remotely monitoring the working status of a paint spray gun based on Example 1, is as follows:

[0102] Step 1: Deploy multi-module sensors to collect environmental data in the workshop, obtain raw environmental data, preprocess the raw environmental data, and generate initial environmental data;

[0103] Specifically, raw environmental data is obtained by deploying temperature, humidity, and air pressure sensors at the spraying station to collect data. The raw environmental data is then preprocessed, including timestamp calibration, noise reduction filtering, outlier removal, and missing value completion, to generate initial environmental data.

[0104] Step 2: Based on the initial environmental data, model the initial environmental data using environmental impact, and overlay and correct it with standard process parameters to obtain the target spraying parameters after environmental compensation.

[0105] Specifically, the current environmental state quantity is generated by statistically analyzing the initial environmental data according to a fixed time window; the environmental compensation quantity is obtained by taking environmental compensation measures on the current environmental state quantity; the target spraying parameters are obtained by superimposing the environmental compensation quantity onto the standard spraying pressure, standard spraying flow rate and standard atomization parameters and applying boundary limiting, and then output to step 3.

[0106] Step 3: Based on the target spraying parameters, collect the working status data of the spray gun, obtain the original status data of the spray gun and preprocess it to generate the initial status data of the spray gun. Synchronously fuse the initial environmental data and the initial status data of the spray gun to generate fused status data.

[0107] Specifically, the original state data of the spray gun is obtained by collecting pressure, flow rate and atomization characterization parameters; the initial state data of the spray gun is generated by taking preprocessing measures such as alignment, noise reduction, anomaly handling and missing data completion on the original state data of the spray gun; and the fused state data is generated by taking the initial environmental data, the initial state data of the spray gun and the target spraying parameters into the same time window alignment and fusion measures.

[0108] Step 4: Based on the fused state data, machine learning analysis methods are used to perform inference calculations on the fused state data to obtain state assessment results and generate monitoring reports and early warning information; among which, the early warning information is generated by residual trend early warning.

[0109] Specifically, by segmenting the fused status data into fixed time windows and extracting the average pressure, average flow rate, rate of change, fluctuation range, and "target-actual" deviation, the model input is formed; by calling the pre-trained spray gun status assessment model for inference calculation, the status assessment results of health score, abnormality type identifier, abnormality level identifier, and risk score are obtained, and a monitoring report is generated.

[0110] Simultaneously, by using a residual trend early warning algorithm to calculate the fusion state data and target spraying parameters, a trend early warning score is obtained and early warning information is generated. This is used to identify progressive nozzle blockage, slow pressure decrease, and slight flow fluctuations before coating defects form. The residual trend early warning algorithm includes at least the following steps: obtaining residuals by subtracting the actual spraying pressure, actual spraying flow rate, and actual atomization characterization from the corresponding target values; obtaining trend characterization by smoothing and continuously accumulating the residuals; obtaining a trend early warning score by calculating the trend characterization score and combining it with a duration window; and generating early warning information including early warning level, early warning reason, trigger time, associated spray gun number, and key evidence indicators when the trend early warning score meets the triggering conditions.

[0111] In one example scenario, particles gradually adhere to the nozzle, causing the flow deviation to accumulate unidirectionally over multiple time windows. Before the traditional threshold alarm is triggered, the system generates a flow trend warning and pushes the warning information to the handling module.

[0112] Step 5: Based on the monitoring report and early warning information, take closed-loop control and rule mapping measures on the status assessment results, generate disposal instructions and send them to the execution terminal;

[0113] By matching the status assessment results, trend warning scores, anomaly type identifiers, and key deviation indicators with rules, the action is determined from the preset action rule library; the action is limited and interlocked by taking measures to generate the action instructions; the action instructions are sent to the execution terminal to trigger the adjustment of spraying pressure / flow / atomization parameters, or to trigger spray gun cleaning, nozzle replacement prompts, switching to backup spray guns, and shutdown for maintenance; and the execution receipts and post-action data returned by the execution terminal are reviewed in a closed loop to complete the remote closed-loop control of early warning, action, and review.

[0114] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for remotely monitoring the working status of a paint spray gun, characterized in that: Step S1: Deploy multi-module sensors to collect environmental data in the workshop, obtain raw environmental data, preprocess the raw environmental data, and generate initial environmental data. Step S2: Based on the initial environmental data, environmental impact modeling is performed on the initial environmental data to establish the correspondence between environmental parameters and spraying process parameters, and to output the environmental compensation amount. The specific steps of environmental impact modeling are as follows: A standard process parameter set is determined according to the established process benchmark to establish a standard environmental condition range as a benchmark. The initial environmental data is segmented, and temperature, humidity, and air pressure are statistically analyzed using a fixed time window to generate the current environmental state quantity. This quantity is then superimposed and corrected with the standard process parameters. The initial environmental data is calculated using an online environmental impact coefficient identification algorithm to establish the correspondence between environmental parameters and spraying process parameters and to output the environmental compensation amount. The standard process parameters are then corrected to target spraying parameters adapted to the current environment. The definition of the online identification algorithm for environmental impact coefficients is as follows: T1 is constructed using temperature (TEP), humidity (WET), and air pressure (PAR) under standard environmental conditions as a benchmark to construct an environmental deviation vector. and compensation vector , where X is the environmental deviation vector. The median temperature. The median humidity. Here, T represents the median air pressure, t is the transpose sign, t is the time index, and Con is the compensation vector. As a form of stress compensation, For traffic compensation, For atomization compensation; T2, based on the environmental deviation vector and compensation vector, obtains the environmental impact coefficient matrix through recursive identification, calculates the real-time environmental compensation amount, and then superimposes the environmental compensation amount with the standard process parameter set to obtain the target spraying parameters. The update formulas for the recursive identification are as follows: ， Where AD is the gain vector, t is the time index, P is the covariance matrix, and X is the environmental bias vector. The forgetting factor is T, and T is the transpose sign. ， Where EN is the environmental impact coefficient matrix, and t is the time index. The compensation vector is calculated in real time, where Con is the compensation vector, X is the environmental deviation vector, and AD is the gain vector. ， Where P is the covariance matrix and t is the time index. Let AD be the forgetting factor, X be the environmental bias vector, and T be the transpose sign. T3, by superimposing and correcting the standard process parameter set, adds the real-time environmental compensation amount to the standard spraying pressure, standard spraying flow rate and standard atomization parameters respectively to obtain the target spraying parameters; by superimposing and correcting the environmental compensation coefficient with the standard process parameters, by applying the corresponding compensation amount to the standard spraying pressure, standard spraying flow rate and standard atomization parameters respectively and performing boundary limiting, the environmentally compensated target spraying parameters are generated, and the target spraying parameters are output to the subsequent spray gun status monitoring and synchronization fusion steps; Step S3: Based on the target spraying parameters, collect the working status data of the spray gun, obtain the original status data of the spray gun and preprocess it to generate the initial status data of the spray gun. Synchronously fuse the initial environmental data and the initial status data of the spray gun to generate fused status data. Step S4: Based on the fused status data, machine learning analysis methods are used to perform inference calculations on the fused status data to obtain status assessment results and generate monitoring reports and early warning information. The implementation of the machine learning analysis is as follows: by segmenting the fused status data into fixed time windows, the average pressure, average flow rate, average angle, pressure change rate, flow fluctuation amplitude, angle drift, and deviation between the target spraying parameters and the actual spray gun parameters within the window are extracted to generate a status feature vector; by performing normalization and feature concatenation processing on the status feature vector, a model input is formed; by sending the model input to the cloud inference module, the pre-trained spray gun status assessment model is called to perform inference calculations, and the status assessment results containing health scores, abnormality type identifiers, abnormality level identifiers, and risk scores are output, and the status assessment results are bound and stored with the corresponding timestamp, workstation number, and spray gun number. The method for generating early warning information is as follows: by applying a residual trend early warning algorithm to the fusion state data and target spraying parameters, a trend early warning score is obtained and an early warning information is generated. When the triggering conditions are met, the early warning level, early warning reason, triggering time, associated spray gun number and key evidence indicators are output. The residual trend early warning algorithm is defined as follows: R1, based on the target spraying parameters Compared with actual spray gun parameters By performing a difference finite element method, we obtain the residual vector , and then normalize the residual vector by applying standard deviation normalization, we obtain the normalized residual vector , where M is the residual vector. This refers to the actual spraying pressure. For the target spraying pressure, This represents the actual spraying flow rate. For the target spraying flow rate, These are the actual atomization parameters. Here are the target atomization parameters, t is the time index, and T is the transpose sign; R2, based on the normalized residual vector, obtains the smoothed residual statistic by performing an exponentially weighted shift statistic on the normalized residual vector. Furthermore, cumulative statistics are performed on the smoothed residuals to obtain the cumulative asymptotic anomalies. Where S is the smoothing residual statistic, Here, L is the smoothing coefficient, and L is the normalized residual vector. This represents the gradual accumulation of abnormalities. The function is a non-negative cutoff function, and e is the drift tolerance term; R3, calculated by combining the smoothed residual statistic and the cumulative asymptotic anomaly, yields the trend warning score. Risk is the trend warning score. The 2-norm represents the overall strength of a vector. The first norm represents the sum of the absolute values ​​of vectors; Step S5: Based on the monitoring report and early warning information, take closed-loop control and rule mapping measures on the status assessment results, generate disposal instructions and send them to the execution terminal.

2. The method for remote monitoring of the working status of a paint spray gun according to claim 1, characterized in that: The specific steps for deploying multi-module sensors to collect environmental data in the workshop and obtain raw environmental data are as follows: By deploying multi-module sensors at each spraying station in the spraying workshop, temperature sensors, humidity sensors, and air pressure sensors are installed in the spray gun operation area, material supply area, and return air area, and environmental data of the workshop is collected according to a unified sampling period to obtain raw environmental data.

3. The method for remote monitoring of the working status of a paint spray gun according to claim 1, characterized in that: The specific steps for preprocessing raw environmental data to generate initial environmental data are as follows: By timestamping the raw environmental data, timestamps are added to the data from each sensor according to the unified clock of the acquisition gateway, and sampling is aligned. Noise filtering is performed on the raw environmental data, and median filtering is used to eliminate noise from high-frequency jitter data. Outlier removal is performed on the raw environmental data, and drift points and abrupt jumps are identified and removed based on physically reasonable intervals and abrupt change thresholds. Missing data is filled in by using linear interpolation for short-term missing data and valid values ​​for continuous missing data, generating initial environmental data including temperature, humidity, and air pressure.

4. The method for remote monitoring of the working status of a paint spray gun according to claim 1, characterized in that: Based on the target spraying parameters, the specific steps for collecting spray gun working status data, obtaining raw spray gun status data, and preprocessing it to generate initial spray gun status data are as follows: Based on the target spraying parameters, the original state data of the spray gun is obtained by collecting its working status data. Specifically, pressure sensors, flow sensors, and angle or attitude sensors are installed in the spray gun's feeding channel, atomization channel, and attitude execution part, respectively. Spraying pressure, spraying flow, and spraying angle data are collected at a uniform sampling period, and the collected data is aggregated through a data acquisition gateway to form the original state data of the spray gun. The original state data of the spray gun is preprocessed to generate the initial state data of the spray gun. Specifically, the original state data of the spray gun is timestamped and sampled to align, so that the pressure, flow, and angle data are on the same time reference. The original state data of the spray gun is denoised and median filtering is used to eliminate random jitter. The original state data of the spray gun is outlier processed by removing abrupt jumps based on the sensor range boundaries and mutation thresholds, and short-term missing data is filled by interpolation to obtain continuous and stable initial state data of the spray gun.

5. The method for remote monitoring of the working status of a paint spray gun according to claim 1, characterized in that: The specific steps for synchronously fusing the initial environmental data and the initial state data of the spray gun to generate fused state data are as follows: By synchronously fusing the initial environmental data and the initial state data of the spray gun, fused state data is generated. Specifically, the synchronous fusing measures are as follows: the initial environmental data and the initial state data of the spray gun are aligned and matched in units of time windows. The ambient temperature, ambient humidity, ambient air pressure, spraying pressure, spraying flow rate, and spraying angle within the same time window are combined into the same fused record. The target spraying parameters are written into the fused record as a reference benchmark to characterize the deviation relationship between the target and the actual situation. Fusion state data containing environmental quantities, spray gun state quantities, and target spraying parameters is generated and used for state assessment and early warning generation.

6. The method for remote monitoring of the working status of a paint spray gun according to claim 1, characterized in that: Based on monitoring reports and early warning information, the specific steps for generating and issuing disposal instructions to the execution terminal by implementing closed-loop control and rule mapping measures based on the status assessment results are as follows: Based on monitoring reports and early warning information, closed-loop control and rule mapping measures are implemented on the status assessment results to generate disposal instructions and send them to the execution terminal. The specific steps of the closed-loop control and rule mapping measures are as follows: key deviation indicators, anomaly type identifiers, and anomaly level identifiers in the monitoring report are read to form disposal trigger conditions, and the trigger conditions are mapped to a set of disposal actions according to a preset disposal rule library; the disposal action set is constrained and verified, upper and lower limits are applied to the adjustment amount, and workstation interlock conditions are verified for shutdown actions, and disposal instructions containing target values, adjustment steps, execution times, and execution priorities are generated. The disposal instructions are then sent to the execution terminal via the industrial wireless network; after receiving the disposal instructions, the execution terminal completes parameter setting or maintenance action triggering, and sends the execution receipt, post-execution spray gun status data, and disposal result markers back to the cloud, realizing a closed-loop disposal process of assessment—disposal—receipt—reassessment.

Images