Bridge cable tower automatic anticorrosion spraying method based on machine vision

Abstract

This invention relates to an automated anti-corrosion spraying method for bridge pylons based on machine vision, specifically in the field of spraying. This solution achieves online real-time sensing and adaptive compensation of the working viscosity of the coating. It uses flow resistance characteristics to deduce the effective dynamic viscosity of the coating and combines this with a process knowledge base to intelligently determine the optimal atomization parameters, thus forming precise feedforward control. Simultaneously, a feedback closed loop based on the actual atomization state is introduced to dynamically correct parameter deviations. Finally, the system can perform confidence-driven incremental optimization of the knowledge base based on control experience during stable operation. This closed loop enables the spraying process to autonomously adapt to changes in coating properties and external disturbances, ensuring uniform coating thickness and firm adhesion under complex working conditions, significantly improving the consistency of anti-corrosion work quality, material utilization, and long-term reliability.

Description

Technical Field

[0001] This invention relates to the field of spraying, and more specifically, to an automated anti-corrosion spraying method for bridge towers based on machine vision. Background Technology

[0002] In the long-term operation and maintenance of major infrastructure such as large bridge towers, anti-corrosion coating is a key link to ensure the structural durability and safety. With the development of automation technology, the use of intelligent equipment equipped with spraying robots to replace high-risk and inefficient manual high-altitude operations has become a clear trend. However, due to the height of the tower structure, anti-corrosion spraying operations often need to be carried out continuously for several hours or even longer. During this period, the ambient temperature and humidity may fluctuate drastically due to the alternation of day and night and conditions such as proximity to water and wind. This complex and dynamic construction environment not only poses a challenge to the stable operation of the equipment, but also directly affects the physicochemical state of the sprayed coating. When the coating is circulated or left to stand in the supply pipeline and container for a long time, its key physical properties such as viscosity and temperature will drift significantly due to factors such as solvent evaporation and heat exchange. The change in physical properties will directly change the rheological characteristics of the coating at the atomizing nozzle, causing the droplet size distribution, spray fan shape and flight trajectory after atomization to deviate from the ideal setting, thereby affecting the adhesion efficiency and film uniformity of the coating on complex three-dimensional curved surfaces.

[0003] While existing automated spraying systems can accurately reproduce the spray gun trajectory and programmatically control basic process parameters, they have significant shortcomings in addressing the deeper process interference caused by the time-varying properties of coatings. Current technologies often focus on trajectory planning and paint dispensing switch control, or simplify the feeding system to a constant pressure and constant flow source. The atomization parameters of the spray gun, such as atomizing air pressure, fan-shaped air pressure, and coating outlet pressure, are usually set to fixed values ​​based on initial tests. This results in the system lacking online sensing and closed-loop control capabilities for the actual coating spraying status. When the coating viscosity increases due to environmental or time factors, the atomization quality deteriorates under fixed parameters. Coarse particles are produced; when the viscosity decreases, sagging is easily caused. Although advanced systems may integrate coating temperature control units, there is still no effective online monitoring and compensation method for viscosity, a more core physical property parameter. During construction, manual intermittent sampling and measurement are still required, which is inefficient and cannot be implemented in high-altitude automated scenarios. Therefore, the existing technical solutions are essentially open-loop or semi-open-loop control, which cannot ensure that the coating is always within the process window during the entire continuous spraying operation. This leads to potential non-uniformity and instability risks in the final coating quality, becoming a key bottleneck restricting the development of automatic spraying technology towards high quality and high reliability. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing an automated anti-corrosion spraying method for bridge towers based on machine vision, thereby solving the problems mentioned in the background art.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: specifically, it includes the following steps: Step S1: Pressure sensors are installed at the outlet of the paint supply pump and the inlet of the atomizing nozzle in the spraying system, respectively. Flow meters and temperature sensors are installed in the paint supply pipeline. The paint supply pump outlet pressure and atomizing nozzle inlet pressure measured by the pressure sensors, the paint volume flow rate measured by the flow meters, and the paint temperature measured by the temperature sensors are collected synchronously at a fixed sampling period. Based on the paint supply pump outlet pressure, atomizing nozzle inlet pressure, and paint volume flow rate, the real-time total flow resistance characteristic is calculated. The real-time total flow resistance characteristic represents the flow resistance from the paint supply pump to the atomizing nozzle. Step S2: Input the real-time total flow resistance characteristic, the real-time collected paint volume flow rate and paint temperature into the preset parameterized flow resistance model, and use the recursive parameter estimation algorithm to perform real-time inversion and update of the effective dynamic viscosity parameters of the paint in the parameterized flow resistance model to obtain the estimated value of the current effective dynamic viscosity of the paint. Step S3: Take the estimated effective dynamic viscosity of the coating and the real-time coating temperature as input, query the preset multi-dimensional process knowledge base, and output a set of optimal atomization parameter feedforward settings through interpolation calculation. The optimal atomization parameter feedforward settings include at least the atomization air pressure setting and the coating pressure setting. Step S4: The optimal atomization parameter feedforward setpoint is sent as the basic control quantity to the corresponding pressure regulating actuator. At the same time, the indirect feedback signal reflecting the actual atomization state is collected, and a feedback compensation quantity is generated based on the deviation between the indirect feedback signal and the preset atomization state expectation value. The optimal atomization parameter feedforward setpoint and the feedback compensation quantity are superimposed to obtain the final control command. During spraying, the effective final control command and the effective data pair consisting of the corresponding effective dynamic viscosity estimate of the coating and the coating temperature are used to fine-tune the multi-dimensional process knowledge base. In a preferred embodiment, in step S1, pressure sensors are respectively arranged at the outlet of the paint supply pump and the inlet of the atomizing nozzle of the spraying system, and flow meters and temperature sensors are arranged in the paint supply pipeline as follows: A first pressure sensor is installed on a straight pipe section downstream of the fluid outlet flange of the paint supply pump to measure the outlet pressure of the paint supply pump; a second pressure sensor is installed on the fluid inlet connection of the atomizing nozzle to measure the pressure at the inlet of the atomizing nozzle; a flow meter is installed on a straight pipe section with stable flow velocity in the main pipeline between the paint supply pump and the atomizing nozzle to measure the volumetric flow rate of the paint flowing through it; a temperature sensor is installed on the outer wall of the pipeline near the flow meter or inside the paint storage container to measure the paint temperature. The same central control unit synchronously collects the measured values ​​of the first pressure sensor, the second pressure sensor, the flow meter, and the temperature sensor at a fixed sampling period, and packages them into data packets with the same timestamp.

[0006] In a preferred embodiment, the specific process for calculating the real-time total flow resistance characteristic is as follows: First, the measured value of the first pressure sensor is subtracted from the measured value of the second pressure sensor to obtain the real-time pressure difference between the pump and the nozzle. Secondly, the measured values ​​of paint volumetric flow rate collected synchronously are subjected to first-order low-pass filtering to obtain smoothed paint volumetric flow rate values. Finally, the real-time total flow resistance characteristic is calculated as follows: multiply the real-time pressure difference by the nominal density constant of the coating at standard temperature and pressure to obtain a first intermediate product; multiply the smoothed volumetric flow rate of the coating by the nominal dynamic viscosity constant of the coating at standard temperature and pressure to obtain a second intermediate product; divide the first intermediate product by the second intermediate product, and the quotient is the real-time total flow resistance characteristic.

[0007] In a preferred embodiment, the preset parameterized flow resistance model is constructed in step S2 as follows: Based on the equivalent pipeline geometric constants from the paint supply pump to the atomizing nozzle, the nominal physical property constants of the paint, and the empirical correction coefficients determined through previous calibration experiments, a parametric flow resistance model is established that expresses the real-time total flow resistance characteristic as a function of the estimated effective dynamic viscosity of the paint, the paint temperature, and the smoothed paint volumetric flow rate. The parametric flow resistance model includes a linear term based on the classical laminar flow law and a comprehensive empirical correction term. The comprehensive empirical correction term is composed of the product of the empirical constant, the exponential part which is the product of the negative temperature influence correction coefficient and the difference between the paint temperature and the standard temperature, and the power part which is the base of the ratio of the smoothed paint volumetric flow rate to the reference flow rate and the exponential part which is the flow nonlinearity correction exponent. The linear term based on the classical laminar flow law is specifically the first coefficient obtained by dividing the equivalent pipe length by the fourth power of the equivalent hydraulic radius, and then multiplying it by the ratio of the estimated effective dynamic viscosity of the coating to the nominal dynamic viscosity of the coating.

[0008] In a preferred embodiment, the specific process of using a recursive parameter estimation algorithm to perform real-time inversion and updating of the effective dynamic viscosity parameters of the coating in the parameterized flow resistance model is as follows: Within each control cycle synchronized with the data acquisition cycle, firstly, the estimated effective dynamic viscosity of the coating obtained from the previous control cycle, along with the coating temperature and smoothed volumetric flow rate values ​​acquired in real time during the current cycle, are input into the parameterized flow resistance model to calculate the real-time total flow resistance characteristic predicted by the model for the current cycle. Secondly, the real-time total flow resistance characteristic predicted by the model is subtracted from the real-time total flow resistance characteristic measured and calculated in step S1 for the current cycle to obtain the model prediction error. Finally, using the recursive least squares method, the gain is dynamically calculated based on the model prediction error and the sensitivity of the parameterized flow resistance model to the effective dynamic viscosity parameter of the coating. This gain is then used to calculate the correction increment, which is added to the estimated effective dynamic viscosity of the coating from the previous cycle to obtain and output the estimated effective dynamic viscosity of the coating for the current cycle. This process is continuously repeated to achieve real-time tracking of the changes in the actual physical properties of the coating by the estimated effective dynamic viscosity of the coating.

[0009] In a preferred embodiment, the construction process of the multidimensional process knowledge base in step S3 is as follows: In the laboratory, spraying tests are conducted on one or more coatings intended for use at multiple discrete effective dynamic viscosity values ​​covering the working range, and at multiple discrete coating temperatures covering the working range. By adjusting the atomizing air pressure and coating pressure, and using atomization quality assessment methods, the optimal values ​​of atomizing air pressure and coating pressure are determined for each set of discrete combinations of effective dynamic viscosity values ​​and coating temperatures to produce the best atomization effect. Each set of discrete combinations of effective dynamic viscosity values ​​and coating temperatures is treated as a coordinate point, and the corresponding optimal values ​​of atomizing air pressure and coating pressure are stored together as a set of data in a structured database, thus forming a multidimensional process knowledge base with effective dynamic viscosity and coating temperature as a joint index.

[0010] In a preferred embodiment, the specific process of calculating and outputting a set of optimal atomization parameter feedforward settings through interpolation is as follows: In each control cycle, firstly, the received effective dynamic viscosity estimate of the coating and the real-time coating temperature are used as query points to locate in the two-dimensional grid of the multi-dimensional process knowledge base. The four nearest grid vertices surrounding the query point are determined, and the optimal values ​​of atomized air pressure and coating pressure stored in these four vertices are read. Secondly, the distance from the query point to these four grid vertices is calculated. Then, based on the calculated four normalized process distances, a Gaussian weighting function is used to calculate the weight coefficient corresponding to each grid vertex. The calculation process is as follows: the product of a weight decay coefficient of negative two times and the square of the normalized process distance is used as the exponent to calculate the value of the natural exponential function, thus obtaining the initial weight of the grid vertex. The initial weights of the four grid vertices are summed to obtain the total weight. The initial weight of each grid vertex is divided by the total weight, and the result is the final weight coefficient of the grid vertex. Finally, the optimal values ​​of atomized air pressure stored by each of the four grid vertices are weighted and summed, and the weighted sum is output as the atomized air pressure setpoint for this period. The optimal values ​​of coating pressure stored by each of the four grid vertices are weighted and summed, and the weighted sum is output as the coating pressure setpoint for this period. The atomized air pressure setpoint and the coating pressure setpoint together constitute a set of optimal atomization parameter feedforward setpoints.

[0011] In a preferred embodiment, the calculation process for the normalized process distance is as follows: The difference between the estimated effective dynamic viscosity of the coating at the query point and the effective dynamic viscosity of the coating at any grid vertex is calculated. This difference is then divided by the average grid spacing of the effective dynamic viscosity dimension of the coating in the multidimensional process knowledge base to obtain the first normalized difference. The difference between the coating temperature at the query point and the coating temperature at the same grid vertex is calculated. This difference is then divided by the average grid spacing of the coating temperature dimension in the knowledge base to obtain the second normalized difference. The squares of the first and second normalized differences are added together, and the square root of the sum is taken. The result is the normalized process distance from the query point to the corresponding grid vertex.

[0012] In a preferred embodiment, the specific process of acquiring an indirect feedback signal reflecting the actual atomization state in step S4, and generating a feedback compensation amount based on the deviation between the indirect feedback signal and the preset expected atomization state value, is as follows: In each control cycle, the operating current signal of the motor driving the atomizing turbine or the motor driving the paint supply pump in the pressure regulating actuator is synchronously acquired as the indirect feedback signal. This operating current signal is preprocessed to remove the DC component and perform bandpass filtering, retaining the frequency bands reflecting the dynamic characteristics of the atomization process. A short-time Fourier transform is performed on the preprocessed operating current signal to convert it from the time domain to the frequency domain, obtaining the current signal spectrum within that control cycle. In the current signal spectrum, the energy of two characteristic frequency bands is defined and calculated: the first is the main fluctuation frequency band energy, whose frequency range is a narrow band interval around the fundamental frequency of the atomizing turbine rotation; the second is the random fluctuation frequency band energy, whose frequency range is a relatively wide high-frequency interval significantly higher than the fundamental frequency of rotation. Based on the calculated main fluctuation frequency band energy and random fluctuation frequency band energy, an atomization stability index is calculated. The specific calculation process for this atomization stability index is as follows: First, the energy of the random fluctuation frequency band is added to a normal coefficient, and the energy of the main fluctuation frequency band is also added to the same normal coefficient. Second, the sum of the former is divided by the sum of the latter to obtain a basic quotient. Next, the logarithm of this basic quotient is calculated to the base 10. Finally, the logarithm result is multiplied by 10, and the product is the atomization stability index. The real-time calculated value of this atomization stability index is compared with a pre-determined expected value of the atomization state, representing the optimal atomization state, to obtain the atomization state deviation. This atomization state deviation is input into a proportional-integral controller. The output of this proportional-integral controller is the feedback compensation amount for the atomizing air pressure and the feedback compensation amount for the coating pressure.

[0013] In a preferred embodiment, the specific process of fine-tuning the multidimensional process knowledge base by comparing the actual effective final control command with the corresponding effective dynamic viscosity estimate of the coating and the coating temperature is as follows: During operation, a control calmness index reflecting the system's control calmness is calculated in real time. The specific calculation process for this index is as follows: Within a fixed number of control cycles over the past period, the absolute values ​​of the feedback compensation for atomized air pressure and the absolute values ​​of the feedback compensation for coating pressure are statistically analyzed. These two absolute values ​​within the same cycle are added together, and the arithmetic mean of all these sums is calculated. The result is the control calmness index. This fixed number is defined as the moving average window length, and a normal coefficient is set as the control calmness threshold. When the control calmness index continuously exceeds a preset number of cycles while remaining below the control calmness threshold, it is determined that the current system is operating smoothly, and the optimal atomization parameter feedforward setpoint has a high degree of matching with actual requirements. At this point, the final control command actually in effect at the current moment... The corresponding combination of the estimated effective dynamic viscosity of the coating and the coating temperature is marked as a high-confidence effective data pair. When fine-tuning is triggered, the effective dynamic viscosity of the coating and the coating temperature in the effective data pair are first located in the two-dimensional grid of the multidimensional process knowledge base to determine the four nearest neighbor grid vertices affected by it, and the spatial interpolation weights from the data point to these four vertices are calculated. Then, for each affected grid vertex, the optimal value of the atomized air pressure or the optimal value of the coating pressure currently stored at the vertex is updated by recursive weighted average using the spatial interpolation weights, a global small learning rate, the optimal value of the atomized air pressure or the optimal value of the coating pressure currently stored at the vertex, and the atomized air pressure value and the coating pressure value contained in the final control command in the effective data pair.

[0014] The beneficial effects of this invention are as follows: This solution realizes online real-time sensing and adaptive compensation of the working viscosity of the coating. The effective dynamic viscosity of the coating is derived by analyzing the flow resistance characteristics, and the optimal atomization parameters are intelligently decided by combining the process knowledge base to form precise feedforward control. At the same time, a feedback closed loop based on the actual atomization state is introduced to dynamically correct parameter deviations. Finally, the system can perform confidence-driven incremental optimization of the knowledge base based on the control experience during stable operation. This closed loop enables the spraying process to autonomously adapt to changes in the physical properties of the coating and external disturbances, ensuring uniform coating thickness and firm adhesion under complex working conditions, and significantly improving the quality consistency, material utilization rate and long-term reliability of anti-corrosion operations. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0018] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application. Example

[0019] This embodiment provides, for example Figure 1 The method for automatic anti-corrosion spraying of bridge pylons based on machine vision is shown, and specifically includes the following steps: Step S1: Pressure sensors are installed at the outlet of the paint supply pump and the inlet of the atomizing nozzle in the spraying system, respectively. Flow meters and temperature sensors are installed in the paint supply pipeline. The paint supply pump outlet pressure and atomizing nozzle inlet pressure measured by the pressure sensors, the paint volume flow rate measured by the flow meters, and the paint temperature measured by the temperature sensors are collected synchronously at a fixed sampling period. Based on the paint supply pump outlet pressure, atomizing nozzle inlet pressure, and paint volume flow rate, the real-time total flow resistance characteristic is calculated. The real-time total flow resistance characteristic represents the flow resistance from the paint supply pump to the atomizing nozzle. Step S2: Input the real-time total flow resistance characteristic, the real-time collected paint volume flow rate and paint temperature into the preset parameterized flow resistance model, and use the recursive parameter estimation algorithm to perform real-time inversion and update of the effective dynamic viscosity parameters of the paint in the parameterized flow resistance model to obtain the estimated value of the current effective dynamic viscosity of the paint. Step S3: Take the estimated effective dynamic viscosity of the coating and the real-time coating temperature as input, query the preset multi-dimensional process knowledge base, and output a set of optimal atomization parameter feedforward settings through interpolation calculation. The optimal atomization parameter feedforward settings include at least the atomization air pressure setting and the coating pressure setting. Step S4: The optimal atomization parameter feedforward setpoint is sent as the basic control quantity to the corresponding pressure regulating actuator. At the same time, the indirect feedback signal reflecting the actual atomization state is collected, and a feedback compensation quantity is generated based on the deviation between the indirect feedback signal and the preset atomization state expectation value. The optimal atomization parameter feedforward setpoint and the feedback compensation quantity are superimposed to obtain the final control command. During spraying, the effective final control command and the effective data pair consisting of the corresponding effective dynamic viscosity estimate of the coating and the coating temperature are used to fine-tune the multi-dimensional process knowledge base.

[0020] In this embodiment, it is specifically necessary to explain the following steps in step S1: Pressure sensors are installed at the outlet of the paint supply pump and the inlet of the atomizing nozzle in the spraying system, and flow meters and temperature sensors are installed in the paint supply pipeline. A first pressure sensor is installed on a straight pipe section downstream of the fluid outlet flange of the paint supply pump to measure the outlet pressure of the paint supply pump. The length of the straight pipe section is not less than 10 times its inner diameter to ensure that the fluid has fully developed before flowing through the measuring point of the first pressure sensor, eliminating the interference of eddies on the pressure measurement. A second pressure sensor is installed on the fluid inlet connection of the atomizing nozzle to measure the pressure at the inlet of the atomizing nozzle. A flow meter is installed on a straight pipe section with stable flow velocity on the main pipeline between the paint supply pump and the atomizing nozzle to measure the volumetric flow rate of the paint. A straight pipe section with stable flow velocity refers to a straight pipe section with an upstream diameter of at least 10 times the pipe diameter and a downstream diameter of at least 5 times the pipe diameter to ensure the measurement accuracy of the flow meter. A temperature sensor is installed on the outer wall of the pipeline adjacent to the flow meter or inside the paint storage container to measure the paint temperature. A straight pipe section with stable flow velocity refers to a straight pipe section with an upstream diameter of at least 10 times the pipe diameter and a downstream diameter of at least 5 times the pipe diameter to ensure the measurement accuracy of the flow meter. Synchronous acquisition with a fixed sampling period refers to the simultaneous reading of the measured values ​​of the first pressure sensor, the second pressure sensor, the flow meter, and the temperature sensor by the same central control unit under the same sampling clock trigger, through a synchronization signal line or an isochronous network. The group of synchronized measured values ​​is then packaged into a data packet with the same timestamp for storage and transmission. The fixed sampling period is set to a range of 1 millisecond to 20 milliseconds, preferably 10 milliseconds. This synchronization mechanism ensures strict time alignment of pressure, flow, and temperature data, providing a consistent timing basis for subsequent calculations and avoiding data fusion errors caused by sampling delays. Using the isochronous Ethernet protocol, microsecond-level synchronization accuracy can be achieved. The specific process for calculating the real-time total flow resistance characteristic is as follows: First, the measured value of the first pressure sensor is subtracted from the measured value of the second pressure sensor to obtain the real-time pressure difference between the pump and the nozzle; this real-time pressure difference directly reflects the pressure loss of the coating during the flow process from the pump to the nozzle. Secondly, the measured value of paint volume flow rate acquired synchronously is subjected to first-order low-pass filtering to eliminate the high-frequency pulsation noise generated by the reciprocating motion of the plunger of the paint supply pump, and a smoothed paint volume flow rate value is obtained. The cutoff frequency of the first-order low-pass filter is set to 1 / 5 to 1 / 10 of the main frequency of the plunger motion of the paint supply pump. For example, for a plunger pump with 600 revolutions per minute, its main frequency is 10Hz, so the cutoff frequency is preferably set to 2Hz. This filtering process can effectively filter out periodic pulsation noise while preserving the true flow trend, and prevent noise from interfering with the stability of subsequent flow resistance calculation. Finally, the real-time total flow resistance characteristic is calculated. This total flow resistance characteristic is a dimensionless normalized transient flow resistance coefficient. The calculation process is as follows: multiply the real-time pressure difference by the nominal density constant of the coating at standard temperature and pressure to obtain a first intermediate product; multiply the smoothed volumetric flow rate of the coating by the nominal dynamic viscosity constant of the coating at standard temperature and pressure to obtain a second intermediate product; divide the first intermediate product by the second intermediate product, and the quotient is the real-time total flow resistance characteristic. The nominal density constant and nominal dynamic viscosity constant are the laboratory standards of the coating at 25 degrees Celsius and one standard atmosphere. Under standard conditions, constant physical property parameters obtained from material data sheets or actual measurements are introduced into the normalization calculation with these two constants to eliminate the influence of differences in the basic physical properties of different batches or formulations of coatings on the absolute flow resistance value. The calculated normalized transient flow resistance coefficient has an ideal reference value that is close to a theoretical constant determined by the geometry of the pipeline and nozzle. When the actual viscosity of the coating increases, the value of this coefficient will increase proportionally, thereby sensitively characterizing the viscosity change. If the value of this coefficient continues to be higher than the reference value to reach a preset threshold (e.g., exceeding the reference value by 20%) and remains for more than 3 seconds, it can trigger an early warning of possible pipeline blockage. The normalized transient flow resistance coefficient characterizes the ratio of the actual flow resistance to the ideal flow resistance when the coating is at its nominal physical properties under the current flow conditions. This ratio is a key input for the online inversion of the effective dynamic viscosity of the coating in subsequent steps. A stable and sensitive normalized transient flow resistance coefficient signal can significantly improve the convergence speed and accuracy of viscosity estimation.

[0021] In this embodiment, it is specifically necessary to explain that the construction method of the preset parameterized flow resistance model in step S2 is as follows: Based on the equivalent pipeline geometric constants from the paint supply pump to the atomizing nozzle, the nominal physical property constants of the paint, and the empirical correction coefficients determined through previous calibration experiments, a parametric flow resistance model is established that expresses the real-time total flow resistance characteristic as a function of the estimated effective dynamic viscosity of the paint, the paint temperature, and the smoothed paint volumetric flow rate. The parametric flow resistance model includes a linear term based on the classical laminar flow law and a comprehensive empirical correction term. The linear term is proportional to the estimated effective dynamic viscosity of the paint and is used to characterize the main flow resistance caused by fluid viscosity. The comprehensive empirical correction term is used to compensate for the influence of paint temperature changes on the viscosity itself and the nonlinear flow effects that may be caused by flow rate changes. The comprehensive empirical correction term is composed of the product of the empirical constant, the exponential part which is the product of the negative temperature influence correction coefficient and the difference between the paint temperature and the standard temperature, and the power part which is the base of the ratio of the smoothed paint volumetric flow rate to the reference flow rate and the exponent of the flow nonlinearity correction index. The equivalent pipeline geometric constants, including the equivalent pipeline length and equivalent hydraulic radius, are calculated by performing fluid dynamic simplification on the actual physical pipeline (including straight pipes, elbows, valves, etc.) from the paint supply pump outlet to the atomizing nozzle inlet. The purpose is to characterize the complex pipeline system as an equivalent straight circular pipe with the same flow resistance. The nominal physical property constants of the paint, including the nominal dynamic viscosity and nominal density, refer to constant physical parameters measured on paint samples or provided by the material supplier in a stable laboratory environment at 25 degrees Celsius and one standard atmosphere. Empirical correction coefficients, namely empirical constants, negative temperature effect correction coefficients, flow nonlinearity correction exponents, and reference flow rates, are determined during the model "training" or "calibration" phase. After the spraying system is assembled, the paint is used in the laboratory under various temperature and stable flow rate combinations covering the expected working range, and the real-time total flow resistance characteristic is recorded after stabilization for each set of conditions. The model employs a multivariate nonlinear regression algorithm to minimize the overall error between the model's predicted values ​​and the measured values ​​under all operating conditions. A set of empirical correction coefficients is fitted and optimized, and the optimal combination of coefficients is the final determined value. The standard temperature is a predefined reference point used to calculate temperature differences, usually consistent with the temperature at which nominal physical constants are measured, such as 25 degrees Celsius. The advantage of this model structure is that the linear term provides a prediction basis based on physical laws, ensuring the model's accuracy under ideal conditions. The comprehensive empirical correction term acts as a "compensator," learning and compensating for deviations between actual complex fluid behavior and the ideal model through data-driven methods. This is particularly important for intrinsic viscosity changes caused by temperature variations (viscosity typically decreases with increasing temperature) and shear thinning or thickening effects that may occur in non-Newtonian fluids at high shear rates (high flow rates), thus significantly improving the model's prediction accuracy and robustness across the entire operating range. The linear term based on the classical laminar flow law is specifically the first coefficient obtained by dividing the equivalent pipe length by the fourth power of the equivalent hydraulic radius, and then multiplying it by the ratio of the estimated effective dynamic viscosity of the coating to the nominal dynamic viscosity of the coating; the empirical constant in the comprehensive empirical correction term is a fixed value determined by fitting the calibration experiment during the system commissioning phase; the negative temperature effect correction coefficient is used to quantify the attenuation or enhancement effect of temperature changes on flow resistance; the flow nonlinearity correction index is used to characterize the degree to which flow changes deviate from linear flow characteristics; the reference flow rate is a benchmark value used to normalize the actual flow rate; The specific process of using a recursive parameter estimation algorithm to perform real-time inversion and updating of the effective dynamic viscosity parameters of the coating in the parameterized flow resistance model is as follows: Within each control cycle synchronized with the data acquisition cycle, firstly, the estimated effective dynamic viscosity of the coating obtained from the previous control cycle, along with the coating temperature and smoothed volumetric flow rate values ​​acquired in real time during this cycle, are input into the parameterized flow resistance model to calculate the real-time total flow resistance characteristic predicted by the model for this cycle. Secondly, the real-time total flow resistance characteristic predicted by the model is subtracted from the real-time total flow resistance characteristic measured and calculated in step S1 for this cycle to obtain the model prediction error. Finally, using the recursive least squares method, the gain is dynamically calculated based on the model prediction error and the sensitivity of the parameterized flow resistance model to the effective dynamic viscosity parameter of the coating. This gain is then used to calculate the correction increment, which is added to the estimated effective dynamic viscosity of the coating from the previous cycle to obtain and output the estimated effective dynamic viscosity of the coating for this cycle. This ensures that the estimated value is constrained within a reasonable physical range based on the nominal dynamic viscosity of the coating. This process is continuously repeated to achieve real-time tracking of the changes in the actual physical properties of the coating by the estimated effective dynamic viscosity of the coating. Recursive least squares is an online estimation algorithm with a forgetting factor, typically ranging from 0.95 to 0.999 (e.g., 0.99). The forgetting factor gradually diminishes the influence of historical data, focusing instead on tracking recent parameter trends. This is crucial for tracking slowly time-varying coating viscosity. The model's sensitivity to the effective dynamic viscosity parameter of the coating is a specific value within each control cycle. It is calculated by taking the partial derivative of the parameterized flow resistance model's mathematical expression with respect to the estimated effective dynamic viscosity of the coating, and substituting the viscosity estimate from the previous cycle, the current temperature, and the flow rate. The gain is a time-varying weighting coefficient, determined by the algorithm's internal covariance matrix, the model sensitivity in the current cycle, and the forgetting factor. It determines the extent to which the current model's prediction error is utilized. To correct the viscosity estimate, the higher the sensitivity or the greater the prediction error, the greater the gain may be, and thus the greater the correction magnitude. The setting of a reasonable physical range is based on prior knowledge of the coating's physical properties. For example, the upper limit can be set to 5.0 times the nominal dynamic viscosity of the coating, and the lower limit can be set to 0.2 times the nominal dynamic viscosity of the coating. If the calculated effective dynamic viscosity estimate of the coating for this period exceeds this range, it will be forcibly limited to the corresponding boundary value. This boundary processing mechanism, as a protection strategy, can effectively prevent the estimate from diverging to physically unreasonable areas when the sensor is subjected to brief strong interference or the initial value of the model is set improperly. This ensures the reliability of the state information received by the subsequent control module. The entire inversion process forms a closed loop of "prediction-comparison-correction-constraint", which enables the effective dynamic viscosity estimate of the coating to adaptively approach its true value. The empirical constant, the negative temperature effect correction coefficient, the flow nonlinearity correction index, and the reference flow rate were determined by fitting a set of calibration experimental data during the system debugging phase, and were subsequently kept constant during operation; the model's sensitivity to viscosity parameters refers to the rate of change of the model output results with the estimated effective dynamic viscosity of the coating; the upper and lower limits of the reasonable physical range are set as several times the nominal dynamic viscosity of the coating, respectively.

[0022] In this embodiment, the construction process of the multi-dimensional process knowledge base in step S3 is specifically described as follows: In the laboratory, spraying tests are conducted on one or more coatings intended for use at multiple discrete effective dynamic viscosity values ​​covering the working range, and at multiple discrete coating temperatures covering the working range. The range of effective dynamic viscosity values ​​should be determined based on the coating's technical data sheet and the expected application environment, for example, from 0.01 Pa·s to 0.5 Pa·s, with discrete points selected at 0.02 Pa·s intervals. The range of coating temperatures should cover the lowest and highest possible application temperatures, for example, from 5 degrees Celsius to 40 degrees Celsius, with discrete points selected at 5-degree Celsius intervals. By adjusting the atomizing air pressure and coating pressure, and using atomization quality assessment methods, the optimal values ​​of atomizing air pressure and coating pressure are determined to produce the best atomization effect under each set of discrete combinations of effective dynamic viscosity values ​​and coating temperatures. Atomization quality assessment methods may include using a high-speed camera to photograph the atomized cone area, calculating the stability and symmetry of the atomization cone angle through image analysis, and simultaneously using laser particle size analysis. The instrument measures the Sottle mean diameter and particle size distribution span of the atomized droplets. The atomizing air pressure and coating pressure corresponding to the conditions where the atomizing cone angle is stabilized within ±3 degrees of the set value, and the Sottle mean diameter is minimized and the particle size distribution span is narrowest, are determined as the optimal values ​​for atomizing air pressure and coating pressure under that set of operating conditions. Each discrete combination of effective dynamic viscosity of the coating and coating temperature is treated as a coordinate point, and the optimal values ​​for atomizing air pressure and coating pressure corresponding to that coordinate point are stored together in a structured database, forming a multi-dimensional process knowledge base with effective dynamic viscosity and coating temperature as joint indexes. The structured database can be in the form of a two-dimensional lookup table, with the row index being an array of discrete effective dynamic viscosity values ​​of the coating arranged in ascending order, and the column index being an array of discrete coating temperature values ​​arranged in ascending order. Each cell stores a pair of floating-point numbers, representing the optimal values ​​for atomizing air pressure and coating pressure corresponding to that row and column index, respectively. This structure facilitates rapid interval positioning and data retrieval in real-time control. The specific process of calculating and outputting a set of optimal atomization parameter feedforward settings through interpolation is as follows: In each control cycle, firstly, the received estimated effective dynamic viscosity of the coating and the real-time coating temperature are used as query points to locate the four nearest grid vertices surrounding the query point in the two-dimensional grid of the multidimensional process knowledge base. The optimal values ​​of atomized air pressure and coating pressure stored at these four vertices are then read. The location process is as follows: In the array storing discrete effective dynamic viscosity values ​​of the coating, a binary search or sequential search is performed to find the largest array element that is less than or equal to the estimated effective dynamic viscosity value of the coating. This element is denoted as the viscosity corresponding to the low viscosity index value, and the next array element is denoted as the viscosity corresponding to the high viscosity index value. The same operation is performed in the array storing discrete coating temperature values ​​to obtain the temperature corresponding to the low temperature index value and the temperature corresponding to the high temperature index value. The coordinates of the four grid vertices surrounding the query point are then formed by combining these two viscosity values ​​and the two temperature values ​​in pairs. Secondly, the distance from the query point to these four grid vertices is calculated. Then, based on the calculated four normalized process distances, a Gaussian weighting function is used to calculate the weight coefficient corresponding to each grid vertex. The calculation process is as follows: the product of a negative two times the weight decay coefficient and the square of the normalized process distance is used as the exponent to calculate the value of the natural exponential function, thus obtaining the initial weight of the grid vertex. The weight decay coefficient is an adjustable parameter greater than zero, used to control the rate at which the weight decays with distance. The typical value range is between 1.0 and 10.0, for example, set to 2.0. The smaller this coefficient is, the more uniform the weight distribution and the smoother the interpolation result. The larger the initial weight, the more concentrated the weight is on the nearest vertex, and the closer the interpolation result is to the nearest neighbor interpolation. The initial weights of the four grid vertices are summed to obtain the total weight. The initial weight of each grid vertex is divided by the total weight, and the result is the final weight coefficient of that grid vertex. The sum of the final weight coefficients of the four grid vertices is strictly equal to one. Grid vertices closer to the query point are assigned larger weight coefficients. Finally, the optimal values ​​of atomized air pressure stored in each of the four grid vertices are weighted and summed, and the weighted sum is output as the atomized air pressure setpoint for this cycle. The weighted summation is calculated as follows: the initial weight of the first grid vertex... The optimal atomized air pressure is multiplied by the final weight coefficient of that vertex, then added to the optimal atomized air pressure of the second, third, and fourth grid vertices multiplied by their final weight coefficients. These four products are summed to obtain the atomized air pressure setpoint. The optimal paint pressure values ​​stored at each of the four grid vertices are then weighted and summed, and the resulting weighted sum is output as the paint pressure setpoint for this cycle. The calculation method for the paint pressure setpoint is similar to that of the atomized air pressure setpoint. With the same atomizing air pressure setting, the same four final weighting coefficients are used to sum the optimal coating pressure values ​​stored at the four vertices. The atomizing air pressure setting and the coating pressure setting constitute a set of optimal atomization parameter feedforward settings. This interpolation method ensures that when the query point is inside the grid, the output parameters are a smooth transition of the parameters of the four adjacent vertices, avoiding parameter jumps. When the query point coincides with a certain grid vertex, the weighting coefficient of that vertex will approach one, and the weights of other vertices will approach zero. The output parameters will naturally converge to the optimal value stored at that vertex, ensuring the accurate reproduction of the knowledge base benchmark. The calculation process for the normalized process distance is as follows: The difference between the estimated effective dynamic viscosity of the coating at the query point and the effective dynamic viscosity of the coating at any grid vertex is calculated. This difference is then divided by the average grid spacing of the effective dynamic viscosity dimension of the coating in the multidimensional process knowledge base to obtain the first normalized difference. The difference between the coating temperature at the query point and the coating temperature at the same grid vertex is calculated. This difference is then divided by the average grid spacing of the coating temperature dimension in the knowledge base to obtain the second normalized difference. The squares of the first and second normalized differences are added together, and the square root of the sum is taken. The result is the normalized process distance from the query point to the corresponding grid vertex.

[0023] In this embodiment, the specific process of collecting the indirect feedback signal reflecting the actual atomization state in step S4, and generating the feedback compensation amount based on the deviation between the indirect feedback signal and the preset expected atomization state value, is as follows: In each control cycle, the operating current signal of the motor driving the atomizing turbine or the motor driving the paint supply pump in the pressure regulating actuator is synchronously acquired as an indirect feedback signal. This operating current signal is preprocessed to remove the DC component and perform bandpass filtering, retaining the frequency bands reflecting the dynamic characteristics of the atomization process. The passband range of the bandpass filter is typically set from 10 Hz to 1 kHz to cover the main dynamic frequency components of the atomization process, while filtering out power frequency interference and higher-frequency switching noise. A short-time Fourier transform is performed on the preprocessed operating current signal to convert it from the time domain to the frequency domain, obtaining the current signal spectrum within that control cycle. The energy of two characteristic frequency bands is defined and calculated within the current signal spectrum. The first is the main fluctuation frequency band energy, which is a narrow band around the fundamental frequency of the atomizing turbine rotation, for example, within ±5 Hz of the fundamental frequency. This main fluctuation frequency band energy mainly characterizes the intensity of periodic disturbances caused by coating supply pulsations and mechanical rotational imbalances. The second is the random fluctuation frequency band energy, which is a wider high-frequency range significantly higher than the fundamental frequency, for example, from 100 Hz to 500 Hz. This random fluctuation frequency band energy mainly characterizes the intensity of random fluctuations generated by physical processes such as droplet breakup and air turbulence during atomization. The atomization stability index is calculated based on the calculated main fluctuation frequency band energy and random fluctuation frequency band energy. The specific calculation process for the atomization stability index is as follows: First, the random fluctuation frequency band energy is added to a normal coefficient, and simultaneously, the main fluctuation frequency band energy is added to the same normal coefficient. The normal coefficient is a very small positive number, such as 10 to the power of -10, used to prevent division by zero errors or undefined logarithmic operations when the frequency band energy value is close to zero, ensuring the numerical stability of the calculation. Second, the sum of the former is divided by the sum of the latter to obtain a basic quotient. Next, the logarithm of this basic quotient is taken to the base 10. Finally, the logarithmic result is multiplied by 10, and the product is the atomization stability index. This atomization stability index is measured in decibels. An increase in its value indicates a stronger advantage of random fluctuation energy over periodic fluctuation energy, usually corresponding to a better and more stable atomization state. A decrease in its value may indicate a decline in atomization quality, such as the appearance of pulses, droplet coarsening, or uneven atomization. The real-time calculated value of this atomization stability index is compared with a pre-determined expected atomization state value representing the optimal atomization state, obtained through calibration experiments, to obtain the atomization state deviation. The expected atomization state value is determined in the experiment... Under standard operating conditions, when the atomization quality is evaluated as optimal, the atomization stability index value is measured and averaged multiple times; for example, this value might be 20 decibels. This atomization state deviation is then input into a proportional-integral controller. The specific rules for gain adjustment can be as follows: a deviation threshold is set. When the absolute value of the atomization state deviation is greater than the threshold, a set of pre-set larger proportional gain and integral gain is used. When the absolute value of the deviation is less than or equal to the threshold, a set of smaller gain values ​​is switched. The proportional gain and integral gain of the proportional-integral controller are adaptively adjusted according to the magnitude of the absolute value of the atomization state deviation. A larger gain is used when the absolute value of the deviation is large, and a smaller gain is used when the absolute value of the deviation is small. The output of the proportional-integral controller is the feedback compensation amount for the atomization air pressure and the feedback compensation amount for the coating pressure. This feedback mechanism can sense and compensate for atomization quality fluctuations caused by unmodeled factors such as slight nozzle wear and environmental airflow disturbances in real time. It complements the feedforward control in step S3, jointly ensuring the stability of the atomization state. The specific process of fine-tuning the multidimensional process knowledge base by comparing the actual final control command with the corresponding effective dynamic viscosity estimate of the coating and the effective data composed of the coating temperature is as follows: During continuous system operation, a control calmness index reflecting the system's control calmness is calculated in real time. The specific calculation process for this index is as follows: Within a fixed number of control cycles over the past period, the absolute values ​​of the feedback compensation for atomized air pressure and the absolute values ​​of the feedback compensation for coating pressure are statistically analyzed. These two absolute values ​​for the same cycle are added together, and the arithmetic mean of all these sums is calculated. The result is the control calmness index. This fixed number is defined as the moving average window length, for example, set to twenty control cycles. A normal coefficient is set as the control calmness threshold, which is typically set to 0.5% of the rated pressure. For example, if the rated atomized air pressure is 0.5 MPa, then the control calmness index is... The threshold can be set to 2.5 kPa. When the control calmness index is below the control calmness threshold for a preset number of consecutive cycles, for example, if this condition is met for five consecutive control cycles, it is determined that the current system is running smoothly and the optimal atomization parameter feedforward setting has a high degree of matching with the actual demand. At this time, the combination of the final control command that is actually effective at the current moment, the corresponding estimated effective dynamic viscosity of the coating, and the coating temperature is marked as a high-confidence valid data pair. When fine-tuning is triggered, the estimated effective dynamic viscosity of the coating and the coating temperature in the valid data pair are first located in the two-dimensional grid of the multi-dimensional process knowledge base to determine the four nearest neighbor grid vertices affected by it, and the spatial interpolation weights from the data point to these four vertices are calculated. This weight calculation method is as follows: The weight function and weight decay coefficient used in step S3 are consistent with those used in the interpolation query. Then, for each affected grid vertex, a recursive weighted average is applied to update the stored optimal atomized air pressure or optimal coating pressure using spatial interpolation weights, a global small learning rate, the currently stored optimal atomized air pressure or optimal coating pressure value of that vertex, and the atomized air pressure and coating pressure values ​​contained in the final control command in the effective data pair. The global learning rate is a very small constant between zero and one. The update rule is as follows: the updated optimal atomized air pressure or optimal coating pressure value is equal to the corresponding optimal pressure value before the update minus the corresponding optimal pressure value before the update, the learning rate, and the spatial interpolation. The difference in the weighted product, plus the product of the learning rate, spatial interpolation weights, and the corresponding pressure value contained in the final control command; the mathematical meaning of this update rule is to adjust the old optimal pressure value of the vertex towards the current high-confidence actual pressure command value. The adjustment magnitude is jointly determined by the product of the learning rate and the spatial interpolation weights. The larger the weight (indicating that the vertex is more affected by the current data point) or the larger the learning rate, the larger the adjustment magnitude, but the overall adjustment process is very slow and smooth. This update process slowly adjusts the multidimensional process knowledge base with a very small magnitude, so that the multidimensional process knowledge base can perform gradual and smooth self-optimization based on the high-confidence experience accumulated in actual operation, while avoiding sudden changes in the multidimensional process knowledge base due to a single disturbance or abnormal data;This online fine-tuning mechanism, based on confidence criteria and a minimal learning rate, enables the process knowledge base to adapt to performance drift after long-term equipment operation and subtle differences between different batches of coatings. This allows for continuous self-learning and performance improvement of the system without manual intervention or recalibration.

[0024] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0025] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0026] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0027] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0028] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0029] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0030] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An automated anti-corrosion spraying method for bridge pylons based on machine vision, characterized in that, Specifically, the steps include the following: Step S1: Pressure sensors are installed at the outlet of the paint supply pump and the inlet of the atomizing nozzle in the spraying system, respectively. Flow meters and temperature sensors are installed in the paint supply pipeline. The paint supply pump outlet pressure and atomizing nozzle inlet pressure measured by the pressure sensors, the paint volume flow rate measured by the flow meters, and the paint temperature measured by the temperature sensors are collected synchronously at a fixed sampling period. Based on the paint supply pump outlet pressure, atomizing nozzle inlet pressure, and paint volume flow rate, the real-time total flow resistance characteristic is calculated. The real-time total flow resistance characteristic represents the flow resistance from the paint supply pump to the atomizing nozzle. Step S2: Input the real-time total flow resistance characteristic, the real-time collected paint volume flow rate and paint temperature into the preset parameterized flow resistance model, and use the recursive parameter estimation algorithm to perform real-time inversion and update of the effective dynamic viscosity parameters of the paint in the parameterized flow resistance model to obtain the estimated value of the current effective dynamic viscosity of the paint. Step S3: Take the estimated effective dynamic viscosity of the coating and the real-time coating temperature as input, query the preset multi-dimensional process knowledge base, and output a set of optimal atomization parameter feedforward settings through interpolation calculation. The optimal atomization parameter feedforward settings include at least the atomization air pressure setting and the coating pressure setting. Step S4: The optimal atomization parameter feedforward setpoint is sent as the basic control quantity to the corresponding pressure regulating actuator. At the same time, the indirect feedback signal reflecting the actual atomization state is collected, and a feedback compensation quantity is generated based on the deviation between the indirect feedback signal and the preset atomization state expectation value. The optimal atomization parameter feedforward setpoint and the feedback compensation quantity are superimposed to obtain the final control command. During spraying, the effective final control command and the effective data pair consisting of the corresponding effective dynamic viscosity estimate of the coating and the coating temperature are used to fine-tune the multi-dimensional process knowledge base.

2. The automatic anti-corrosion spraying method for bridge pylons based on machine vision according to claim 1, characterized in that: In step S1, the specific operations of arranging pressure sensors at the outlet of the paint supply pump and the inlet of the atomizing nozzle of the spraying system, and arranging flow meters and temperature sensors in the paint supply pipeline are as follows: A first pressure sensor is installed on a straight pipe section downstream of the fluid outlet flange of the paint supply pump to measure the outlet pressure of the paint supply pump; a second pressure sensor is installed on the fluid inlet connection of the atomizing nozzle to measure the pressure at the inlet of the atomizing nozzle; a flow meter is installed on a straight pipe section with stable flow velocity in the main pipeline between the paint supply pump and the atomizing nozzle to measure the volumetric flow rate of the paint flowing through it; a temperature sensor is installed on the outer wall of the pipeline near the flow meter or inside the paint storage container to measure the paint temperature. The same central control unit synchronously collects the measured values ​​of the first pressure sensor, the second pressure sensor, the flow meter, and the temperature sensor at a fixed sampling period, and packages them into data packets with the same timestamp.

3. The automatic anti-corrosion spraying method for bridge towers based on machine vision according to claim 2, characterized in that: The specific process for calculating the real-time total flow resistance characteristic is as follows: First, the measured value of the first pressure sensor is subtracted from the measured value of the second pressure sensor to obtain the real-time pressure difference between the pump and the nozzle. Secondly, the measured values ​​of paint volumetric flow rate collected synchronously are subjected to first-order low-pass filtering to obtain smoothed paint volumetric flow rate values. Finally, the real-time total flow resistance characteristic is calculated as follows: multiply the real-time pressure difference by the nominal density constant of the coating at standard temperature and pressure to obtain a first intermediate product; multiply the smoothed volumetric flow rate of the coating by the nominal dynamic viscosity constant of the coating at standard temperature and pressure to obtain a second intermediate product; divide the first intermediate product by the second intermediate product, and the quotient is the real-time total flow resistance characteristic.

4. The automatic anti-corrosion spraying method for bridge pylons based on machine vision according to claim 3, characterized in that: In step S2, the preset parameterized flow resistance model is constructed as follows: Based on the equivalent pipeline geometric constants from the paint supply pump to the atomizing nozzle, the nominal physical property constants of the paint, and the empirical correction coefficients determined through previous calibration experiments, a parametric flow resistance model is established that expresses the real-time total flow resistance characteristic as a function of the estimated effective dynamic viscosity of the paint, the paint temperature, and the smoothed paint volumetric flow rate. The parametric flow resistance model includes a linear term based on the classical laminar flow law and a comprehensive empirical correction term. The comprehensive empirical correction term is composed of the product of the empirical constant, the exponential part which is the product of the negative temperature influence correction coefficient and the difference between the paint temperature and the standard temperature, and the power part which is the base of the ratio of the smoothed paint volumetric flow rate to the reference flow rate and the exponential part which is the flow nonlinearity correction exponent. The linear term based on the classical laminar flow law is specifically the first coefficient obtained by dividing the equivalent pipe length by the fourth power of the equivalent hydraulic radius, and then multiplying it by the ratio of the estimated effective dynamic viscosity of the coating to the nominal dynamic viscosity of the coating.

5. The automatic anti-corrosion spraying method for bridge towers based on machine vision according to claim 4, characterized in that: The specific process of using a recursive parameter estimation algorithm to perform real-time inversion and updating of the effective dynamic viscosity parameters of the coating in the parameterized flow resistance model is as follows: Within each control cycle synchronized with the data acquisition cycle, firstly, the estimated effective dynamic viscosity of the coating obtained from the previous control cycle, along with the coating temperature and smoothed volumetric flow rate values ​​acquired in real time during the current cycle, are input into the parameterized flow resistance model to calculate the real-time total flow resistance characteristic predicted by the model for the current cycle. Secondly, the real-time total flow resistance characteristic predicted by the model is subtracted from the real-time total flow resistance characteristic measured and calculated in step S1 for the current cycle to obtain the model prediction error. Finally, using the recursive least squares method, the gain is dynamically calculated based on the model prediction error and the sensitivity of the parameterized flow resistance model to the effective dynamic viscosity parameter of the coating. This gain is then used to calculate the correction increment, which is added to the estimated effective dynamic viscosity of the coating from the previous cycle to obtain and output the estimated effective dynamic viscosity of the coating for the current cycle. This process is continuously repeated to achieve real-time tracking of the changes in the actual physical properties of the coating by the estimated effective dynamic viscosity of the coating.

6. The automatic anti-corrosion spraying method for bridge towers based on machine vision according to claim 5, characterized in that: In step S3, the construction process of the multidimensional process knowledge base is as follows: In the laboratory, spraying tests are conducted on one or more coatings intended for use at multiple discrete effective dynamic viscosity values ​​covering the working range, and at multiple discrete coating temperatures covering the working range. By adjusting the atomizing air pressure and coating pressure, and using atomization quality assessment methods, the optimal values ​​of atomizing air pressure and coating pressure are determined for each set of discrete combinations of effective dynamic viscosity values ​​and coating temperatures to produce the best atomization effect. Each set of discrete combinations of effective dynamic viscosity values ​​and coating temperatures is treated as a coordinate point, and the corresponding optimal values ​​of atomizing air pressure and coating pressure are stored together as a set of data in a structured database, thus forming a multidimensional process knowledge base with effective dynamic viscosity and coating temperature as a joint index.

7. The automatic anti-corrosion spraying method for bridge towers based on machine vision according to claim 6, characterized in that: The specific process of calculating and outputting a set of optimal atomization parameter feedforward setpoints through interpolation is as follows: In each control cycle, firstly, the received effective dynamic viscosity estimate of the coating and the real-time coating temperature are used as query points to locate in the two-dimensional grid of the multi-dimensional process knowledge base. The four nearest grid vertices surrounding the query point are determined, and the optimal values ​​of atomized air pressure and coating pressure stored in these four vertices are read. Secondly, calculate the distance from the query point to these four grid vertices; then, based on the calculated four normalized process distances, use a Gaussian weighting function to calculate the weight coefficient corresponding to each grid vertex. The calculation process is as follows: use the product of a negative two times weight decay coefficient and the square of the normalized process distance as the exponent to calculate the value of the natural exponential function, and obtain the initial weight of the grid vertex; sum the initial weights of the four grid vertices to obtain the total weight. Divide the initial weight of each grid vertex by the total weight to get the final weight coefficient of that grid vertex. Finally, the optimal atomized air pressure values ​​stored at each of the four grid vertices are summed using a weighted average, and the summation result is output as the atomized air pressure setpoint for this cycle. The optimal paint pressure values ​​stored at each of the four grid vertices are also summed using a weighted average, and the summation result is output as the paint pressure setpoint for this cycle. The atomized air pressure setpoint and the paint pressure setpoint together constitute a set of optimal atomization parameter feedforward setpoints.

8. The automatic anti-corrosion spraying method for bridge pylons based on machine vision according to claim 7, characterized in that: The calculation process for the normalized process distance is as follows: The difference between the estimated effective dynamic viscosity of the coating at the query point and the effective dynamic viscosity of the coating at any grid vertex is calculated. This difference is then divided by the average grid spacing of the effective dynamic viscosity dimension of the coating in the multidimensional process knowledge base to obtain the first normalized difference. Calculate the difference between the coating temperature at the query point and the coating temperature at the same grid vertex. Divide this difference by the average grid spacing of the coating temperature dimension in the knowledge base to obtain the second normalized difference. Add the square of the first normalized difference to the square of the second normalized difference, and take the square root of the sum. The result is the normalized process distance from the query point to the corresponding grid vertex.

9. The automatic anti-corrosion spraying method for bridge pylons based on machine vision according to claim 8, characterized in that: In step S4, the specific process of acquiring an indirect feedback signal reflecting the actual atomization state and generating a feedback compensation amount based on the deviation between the indirect feedback signal and the preset expected atomization state is as follows: In each control cycle, the operating current signal of the motor driving the atomizing turbine or the motor driving the paint supply pump in the pressure regulating actuator is synchronously collected as an indirect feedback signal. The operating current signal is preprocessed to remove the DC component and perform bandpass filtering, while retaining the frequency band that reflects the dynamic characteristics of the atomization process. A short-time Fourier transform is performed on the preprocessed operating current signal to convert it from the time domain to the frequency domain, thus obtaining the current signal spectrum within the control cycle. In the current signal spectrum, the energy of two characteristic frequency bands is defined and calculated: the first is the energy of the main fluctuation frequency band, whose frequency range is a narrow band interval around the fundamental frequency of the atomizing turbine rotation. The second is the random fluctuation frequency band energy, which is a relatively wide high-frequency range significantly higher than the rotating fundamental frequency. The atomization stability index is calculated based on the calculated main fluctuation frequency band energy and the random fluctuation frequency band energy. The specific calculation process for this atomization stability index is as follows: First, the energy of the random fluctuation frequency band is added to a normal coefficient, and the energy of the main fluctuation frequency band is also added to the same normal coefficient. Second, the sum of the former is divided by the sum of the latter to obtain a basic quotient. Next, the logarithm of this basic quotient is calculated to the base 10. Finally, the logarithm result is multiplied by 10, and the product is the atomization stability index. The real-time calculated value of this atomization stability index is compared with a pre-determined expected value of the atomization state, representing the optimal atomization state, to obtain the atomization state deviation. This atomization state deviation is input into a proportional-integral controller. The output of this proportional-integral controller is the feedback compensation amount for the atomizing air pressure and the feedback compensation amount for the coating pressure.

10. The automatic anti-corrosion spraying method for bridge pylons based on machine vision according to claim 9, characterized in that: The specific process of fine-tuning the multidimensional process knowledge base by comparing the actual final control command with the corresponding effective dynamic viscosity estimate of the coating and the coating temperature is as follows: During operation, a control calmness index reflecting the system's control calmness is calculated in real time. The specific calculation process for this index is as follows: Within a fixed number of control cycles over the past period, the absolute values ​​of the feedback compensation for atomized air pressure and the absolute values ​​of the feedback compensation for coating pressure are statistically analyzed. These two absolute values ​​for the same cycle are added together, and the arithmetic mean of all these sums is calculated. The result is the control calmness index. This fixed number is defined as the moving average window length, and a normal coefficient is set as the control calmness threshold. When the control calmness index exceeds a preset number of cycles but remains below the control calmness threshold, the system is considered to be operating smoothly, and the optimal atomization parameter feedforward setpoint has a high degree of matching with actual requirements. At this point, the final control index that is actually effective at the current moment is set... The combination of the corresponding effective dynamic viscosity estimate of the coating and the coating temperature is marked as a high-confidence effective data pair. When fine-tuning is triggered, the effective dynamic viscosity estimate of the coating and the coating temperature in the effective data pair are first located in the two-dimensional grid of the multidimensional process knowledge base to determine the four nearest neighbor grid vertices affected by it, and the spatial interpolation weights from the data point to these four vertices are calculated. Then, for each affected grid vertex, the optimal value of the atomized air pressure or the optimal value of the coating pressure stored at the vertex is updated by recursion weighted average using the spatial interpolation weights, a global small learning rate, the optimal value of the atomized air pressure or the optimal value of the coating pressure currently stored at the vertex, and the atomized air pressure value and the coating pressure value contained in the final control command in the effective data pair.

Images