Adaptive intelligent control method for powder coating bonding machine based on multi-source information fusion

Abstract

The application is based on a multi-source information fusion powder coating bonding machine self-adaptive intelligent control method, specifically as follows: real-time data acquisition through the sensor arranged on the powder coating bonding machine, using CodeSys interactive platform to configure the communication link of the sensor and the bonding machine, establishing variable mapping to store the collected data; based on Kalman filtering algorithm, a high-precision filtering model suitable for the bonding machine is constructed; based on recursive least squares method, a glass transition degree estimation algorithm is constructed, and the glass transition degree of the powder coating is calculated in real time through recursive updating mode; based on the estimation result of the glass transition degree, an adaptive impedance control model is constructed, and the impedance control parameters of the adaptive impedance control model are dynamically adjusted according to the glass transition degree; through the segmented PID-fuzzy compound temperature control algorithm, the blade rotating speed of the bonding machine and the temperature supply of the hot and cold blocks of the bonding barrel and the single length of the bonding production are adjusted in real time. The application solves the problems of low control precision, slow response and poor coordination of the existing bonding machine.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control, and in particular to a powder coating bonding machine adaptive intelligent control method based on multi-source information fusion. BACKGROUND

[0002] As a green and environmentally friendly coating, powder coating is widely used in the fields of household appliances, automobile hubs, electronic products, etc. Among them, metal powder coating has a growing market demand due to its protective performance and metal flash effect. In actual production, the preparation methods of metal powder coating mainly include dry mixing method, melt extrusion method and bonding method. Among them, the bonding method can effectively avoid the defects of dry mixing method and melt extrusion method. The prepared powder coating has excellent color stability and vivid and lifelike metal flash effect when sprayed. The performance of some high-end products can even be comparable to that of metal liquid coating. The key of the bonding method lies in accurately controlling the bonding temperature (close to the glass transition temperature of the powder coating) and the torque to avoid powder caking or insufficient bonding.

[0003] The existing bonding machine control technology objectively has the following shortcomings: low data processing accuracy, large torque and temperature data noise interference caused by device vibration and temperature fluctuation during the bonding process, traditional filtering methods (such as mean filtering) cannot effectively eliminate multi-source noise, resulting in parameter measurement error exceeding 2℃, affecting the glass state judgment; glass estimation lag, the existing technology relies on offline testing (such as differential scanning calorimetry) to obtain glass parameters, which cannot track the glass dynamic changes of the powder coating in the bonding process in real time, resulting in control parameter adjustment lag and difficulty in breaking through the metal pigment addition amount of 12%; control mode switching is cumbersome, traditional force / position hybrid control needs to frequently switch modes between free space and contact space, and when facing the dynamic changes of the glass state of the powder coating, it is easy to have overshoot exceeding large and long response time, which further affects the bonding effect; temperature control and torque control are disconnected, the existing temperature control mostly uses single PID algorithm, which does not adjust in coordination with the glass state reflected by the torque, resulting in large temperature difference error in the bonding barrel and easy caking of the powder in some areas.

[0004] Therefore, the present application proposes a powder coating bonding machine adaptive intelligent control method based on multi-source information fusion to solve the above problems. SUMMARY

[0005] The present application is aimed at the shortcomings of the prior art and develops a powder coating bonding machine adaptive intelligent control method based on multi-source information fusion. The present application can solve the problems of low control accuracy, slow response and poor coordination of the existing bonding machine by fusing multi-source information, estimating the glass degree in real time and dynamically adjusting the control parameters.

[0006] The technical scheme for solving the technical problems of the present application is a powder coating bonding machine self-adaptive intelligent control method based on multi-source information fusion, comprising the following steps:

[0007] S1. Real-time acquisition of multi-source data: real-time acquisition of data through sensors arranged on the powder coating bonding machine, configuration of the communication link between the sensors and the bonding machine using the CodeSys interaction platform, establishment of variable mapping to store the acquired data;

[0008] S2, multi-source data preprocessing: based on the Kalman filtering algorithm, a high-precision filtering model suitable for the bonding machine is constructed, the acquired data is input into the filtering model for denoising, and accurate data after denoising is obtained;

[0009] S3, on-line estimation of the glass transition degree of powder coating: based on the recursive least squares method, a glass transition degree estimation algorithm is constructed, the accurate data after denoising is taken as input, and the glass transition degree of the powder coating is calculated in real time through recursive updating;

[0010] S4, self-adaptive impedance control parameter adjustment: based on the estimation result of the glass transition degree, a self-adaptive impedance control model is constructed, the self-adaptive impedance control model is based on the "mass-damping-spring" physical model, and the impedance control parameters of the self-adaptive impedance control model are dynamically adjusted according to the glass transition degree;

[0011] S5, bonding process closed-loop control: based on the adjusted impedance control parameters, combined with the segmented PID-fuzzy compound temperature control algorithm, the blade rotation speed of the bonding machine and the temperature supply of the cold-hot segmented bonding barrel and the single time length of the bonding production are adjusted in real time.

[0012] S1 is as follows:

[0013] (1) Data acquisition: the sensors include a high-precision six-dimensional force / torque sensor, a temperature sensor and an integrated speed sensor;

[0014] The high-precision six-dimensional force / torque sensor is arranged at the connection end of the bonding machine stirring shaft and the driving motor, and simultaneously acquires the dynamic torque signal, the axial / radial contact force signal and the blade displacement signal in the stirring process;

[0015] The temperature sensors are arranged in the cold-hot segmented block area of the bonding barrel, the core area of the powder material and the inlet and outlet of the blade cooling hole, and acquire the real-time temperature of each place;

[0016] The integrated speed sensor is arranged on the blade driving shaft to acquire the real-time speed of the blade;

[0017] (2) Communication link and data storage:

[0018] Based on the CodeSys platform, the communication link between the sensor and the programmable logic controller of the bonding machine is configured. The industrial Ethernet protocol is used to dynamically map the collected data, build the coupling relationship between the data, and store the data through a dynamic data caching mechanism.

[0019] S2 is as follows:

[0020] A high-precision filtering model suitable for bonding machines is constructed based on the Kalman filtering algorithm. The torque, force, displacement and temperature data in the collected data are used as the input of the filtering model. The noise caused by equipment vibration and environmental interference is eliminated by constructing state equations and observation equations. Then the data is verified and corrected, and the denoised accurate data is output.

[0021] S2.1 Constructing a high-precision filtering model:

[0022] The state equations are as follows:

[0023] ,

[0024] The specific observation equations are as follows:

[0025] ,

[0026] In the formula, express A state vector composed of moment-to-moment torque, force, displacement, and temperature data. express A state vector composed of moment-to-moment torque, force, displacement, and temperature data; express The process noise matrix at each time step; This represents the state transition matrix, which is essentially a diagonal matrix defined based on the dynamic characteristics of the bonded machine. This represents the control input matrix, which is essentially the influence coefficient of blade speed on the state. express The blade rotation speed at any given moment; express Raw data collected by the time sensor; This represents the observation matrix, where the observation moments are essentially the identity matrix. express The observation noise matrix at time step;

[0027] in, Each element in the set follows a mean of 0 and a covariance of . Gaussian distribution, Each element in the set follows a mean of 0 and a covariance of . A Gaussian distribution is used, and the variance of the original data is calculated by adjusting the sliding window size, which then dynamically corrects the variance. and covariance coefficient and ;

[0028] S2.2, checksum correction of data:

[0029] (1) Correct the temperature sensor. If the data deviation between two temperature sensors in the same cold-hot block area exceeds the set threshold, the two temperature sensors are judged to be abnormal, the faulty temperature sensor data is removed, and the removed data is replaced by a sliding average;

[0030] (2) Correct the torque data. Based on the coupling relationship between the blade speed and the torque, the filtered torque data is linearly corrected.

[0031] S3 is as follows:

[0032] S3.1, based on the denoised accurate data to construct the vitrification degree correlation model, the calculation formula is as follows:

[0033] ,

[0034] In the formula, represents the vitrification degree, , when it represents complete non-vitrification, represents complete vitrification; , , and represent the weight coefficient, ; represents the normalized function of torque ; represents the normalized function of temperature ; represents the correction function of contact force ; represents the correction function of displacement :

[0035] S3.2, the weight coefficient in the vitrification degree calculation is updated in real time by using the recursive least square method, and the updated vitrification degree is obtained.

[0036] First, initialize the data, and take the normalized function of torque , the normalized function of temperature , the correction function of contact force and the correction function of displacement as the initial output vector , , and the weight coefficient As the initial weight coefficient vector ,correspond Then, recursively update the data based on the input vector from the previous moment. Calculate the augmented matrix at the current time step. Then, based on the augmented matrix at the current moment and the weight coefficient vector of the previous moment Calculate the current weight coefficient vector ;

[0037] The calculation formula is as follows:

[0038] ,

[0039] ,

[0040] ,

[0041] in, , express The torque normalization function at time t, express The temperature normalization function at time t, express Contact force correction function at time t, express The displacement correction function at time t. Indicates transpose. This represents the forgetting factor that changes according to the bonding process. This represents a reference value for the degree of vitrification based on offline experiments. Represents the identity matrix.

[0042] S4 is as follows:

[0043] Based on the dynamic characteristics of the bonding machine stirring system, a physical model of "mass-damping-spring" for impedance control is constructed, where "mass" refers to the equivalent mass. The moment of inertia corresponding to the agitator shaft and blades of the bonding machine; "damping" refers to the damping coefficient. It is positively correlated with the updated degree of glassiness, and "spring" corresponds to the spring stiffness. The contact force is positively correlated with the updated glass transition level; the desired contact force is determined based on the updated glass transition level. Then, based on the actual contact force Angular displacement of the blade angular velocity angular acceleration The physical model is determined, and the calculation formula is as follows:

[0044] ;

[0045] Wherein, the updated glassification degree is compared with the preset glassification degree threshold And The equivalent mass , damping coefficient And spring stiffness are dynamically adjusted, ;

[0046] When , corresponding to low glassification degree, increase the damping coefficient, reduce the spring stiffness, and the equivalent mass remains unchanged;

[0047] When , corresponding to medium glassification degree, linearly adjust the damping coefficient and spring stiffness, and correct the equivalent mass;

[0048] When , corresponding to high glassification degree, reduce the damping coefficient, increase the spring stiffness, and correct the equivalent mass.

[0049] S5 is as follows:

[0050] S5.1, segmented PID-fuzzy compound temperature control algorithm:

[0051] (1) In the stage of large temperature deviation, fuzzy control is adopted, taking temperature deviation and deviation change rate as input, and according to the preset fuzzy rule table, rapid heating or cooling is carried out;

[0052] The temperature deviation is equal to the difference between the collected temperature and the glassification temperature;

[0053] (2) In the stage of small temperature deviation, PID control is adopted to continue to shorten the temperature deviation;

[0054] S5.2, adaptive adjustment:

[0055] (1) Adjust the blade speed:

[0056] When , increase the speed;

[0057] When , keep the medium speed;

[0058] When , reduce the speed;

[0059] (2) Adjust the single bonding time:

[0060] According to the time of the glassification degree to reach the target glassification degree, correct the single time of the next batch , , Indicates the reserved stabilization time;

[0061] If the deviation of 3 batches in succession meets the condition, then it is fixed ; if it does not meet the condition, then it is recalculated.

[0062] The effects provided in the summary are only the effects of the embodiments, not all the full effects of the application, and the above technical solutions have the following advantages or beneficial effects:

[0063] The application adopts multiple sensors to collect multi-source data, realizes synchronous collection of multi-source data, and further comprehensively reflects the bonding state; through the communication link of the configured sensor and the bonding machine, a dynamic variable mapping mechanism is constructed to build a data coupling relationship, avoid parameter calculation deviation caused by multi-source data timestamp mismatch, and lay a data foundation for subsequent accurate control; for multi-band noise caused by vibration and environmental interference of the bonding machine, a Kalman filter model of the frequency band parameter is constructed, through a dynamic correction process noise covariance Q and observation noise covariance R, the error of temperature and torque can be reduced, at the same time, combined with temperature sensor redundancy check and torque-speed coupling correction, the data reliability is further improved, and the composite noise can be effectively eliminated; an association model of torque, temperature, contact force, displacement and glassification degree is constructed, the glassification degree is estimated by using the recursive least square method, the hysteresis problem can be solved, and the change of the glassification degree can be obtained in time and accurately; based on the dynamic adjustment of the equivalent mass, the damping coefficient and the spring stiffness according to the glassification degree, the cumbersome switching under the hybrid control condition can be avoided, and the stability and rapidity of the adaptive impedance regulation and control are improved; the segmented PID-fuzzy composite temperature control and closed-loop cooperative algorithm are adopted, the temperature can be accurately adjusted, the problem of slow response and large overshoot of single temperature regulation and control is avoided, at the same time, combined with the linkage adjustment of the blade speed and the single bonding time according to the glassification degree, the bonding temperature error can be reduced, and the local powder caking phenomenon can be prevented.

[0064] In summary, through the deep combination of multi-source information fusion and adaptive intelligent control, the whole process optimization from data collection, preprocessing, state estimation to control execution can be formed, the data accuracy and real-time performance can be improved, the dynamic control parameter adaptive adjustment and the multi-vacuum link cooperative linkage can be realized, the problems of disconnection of temperature control and torque control and cumbersome mode switching are solved, and the equipment operation stability and product quality are improved. BRIEF DESCRIPTION OF DRAWINGS

[0065] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, together with the embodiments of the application, to explain the application, and do not constitute a limitation on the application.

[0066] Figure 1A flow chart of the method of the present application. DETAILED DESCRIPTION

[0067] In order to clearly illustrate the technical features of the present application, the present application will be described in detail below with reference to the specific embodiments and the accompanying drawings.

[0068] Embodiment 1

[0069] A powder coating bonding machine self-adaptive intelligent control method based on multi-source information fusion, comprising the following steps:

[0070] S1. Real-time acquisition of multi-source data: real-time acquisition of data through sensors arranged on the powder coating bonding machine, configuration of the communication link between the sensors and the bonding machine using the CodeSys interaction platform, establishment of variable mapping to store the acquired data;

[0071] S2, multi-source data preprocessing: based on the Kalman filter algorithm, a high-precision filter model suitable for the bonding machine is constructed, the acquired data is input into the filter model for denoising, and accurate data after denoising is obtained;

[0072] S3, on-line estimation of the glass transition degree of powder coating: based on the recursive least squares method, a glass transition degree estimation algorithm is constructed, the accurate data after denoising is taken as input, and the glass transition degree of the powder coating is calculated in real time through recursive updating;

[0073] S4, self-adaptive impedance control parameter adjustment: based on the estimation result of the glass transition degree, a self-adaptive impedance control model is constructed, the self-adaptive impedance control model is based on the "mass-damping-spring" physical model, and the impedance control parameters of the self-adaptive impedance control model are dynamically adjusted according to the glass transition degree;

[0074] S5, bonding process closed-loop control: based on the adjusted impedance control parameters, combined with the segmented PID-fuzzy compound temperature control algorithm, the blade rotation speed of the bonding machine and the temperature supply of the cold-hot segmented bonding barrel and the single time length of the bonding production are adjusted in real time.

[0075] In the specific embodiment, S1 is specifically as follows:

[0076] (1) Data acquisition: the sensors include a high-precision six-dimensional force / torque sensor, a temperature sensor, and an integrated speed sensor, the temperature sensor can be a PT100 platinum resistance temperature sensor;

[0077] A high-precision six-dimensional force / torque sensor is arranged at the connection end of the bonded machine stirring shaft and the driving motor to synchronously collect dynamic torque signals, axial / radial contact force signals and blade displacement signals during stirring; the axial / radial contact force signals can reflect the interaction between the powder and the blade, the blade displacement signals can monitor the vibration deviation of the stirring system, thereby ensuring the synchronization of multi-channel data time stamps and avoiding parameter mismatch caused by time delay;

[0078] Temperature sensors are arranged at the cold-hot partition area of the bonded barrel, the core area of the powder material and the inlet and outlet of the blade cooling hole to collect real-time temperatures; the temperature sensors arranged at the cold-hot partition area of the bonded barrel can monitor the barrel wall temperature and the material contact temperature, the pre-embedded sensors arranged at the core area of the powder material can collect the internal temperature of the material to avoid the deviation between the surface temperature and the internal temperature, and the sensors arranged at the inlet and outlet of the blade cooling hole can feedback the cooling system efficiency;

[0079] An integrated rotating speed sensor is arranged at the blade driving shaft to collect the real-time rotating speed of the blade;

[0080] (2) Communication link and data storage:

[0081] The communication link of the sensor and the programmable logic controller of the bonded machine is configured based on the CodeSys platform, industrial Ethernet protocol is adopted to perform dynamic variable mapping on the collected data, the coupling relationship between the data is constructed, and the data is stored through a dynamic data caching mechanism;

[0082] The CodeSys platform can realize 100ms-level data transmission, thereby ensuring the real-time performance of data transmission.

[0083] In the specific implementation, S2 is specifically as follows:

[0084] A high-precision filtering model suitable for the bonded machine is constructed based on the Kalman filtering algorithm, the torque, force, displacement and temperature data in the collected data are taken as the input of the filtering model, the state equation and the observation equation are constructed to eliminate the noise caused by device vibration and environmental interference, and then the data is checked and corrected to output the accurate data after noise reduction;

[0085] The construction of the high-precision filtering model suitable for the bonded machine based on the Kalman filtering algorithm can design filtering parameters in different frequency bands according to the different frequency bands of the vibration noise and the temperature fluctuation noise of the bonded machine, thereby better reducing the data noise;

[0086] S2.1, constructing a high-precision filtering model:

[0087] The state equation is specifically as follows:

[0088] ,

[0089] The observation equation is specifically as follows:

[0090]

[0091] In the formula, represents a state vector composed of torque, force, displacement and temperature data at the moment, represents a state vector composed of torque, force, displacement and temperature data at the moment; represents a process noise matrix at the moment; represents a state transition matrix, which is essentially a diagonal matrix based on the characteristics of the bonding motor; represents a control input matrix, which is essentially the influence coefficient of the blade speed on the state; represents the blade speed at the moment; represents the original data collected by the sensor at the moment; represents an observation matrix, which is essentially a unit matrix,

[0092] wherein, each element in the formula is subject to a Gaussian distribution with a mean of 0 and a covariance of each element in the formula is subject to a Gaussian distribution with a mean of 0 and a covariance of , the variance of the original data is calculated by a sliding window size, and the covariance coefficients of and are dynamically corrected;

[0093] S2.2, verification and correction of data:

[0094] (1) The temperature sensor is corrected. If the data deviation between two temperature sensors in the same cold and hot block area exceeds the set threshold, the two temperature sensors are judged to be abnormal, the faulty temperature sensor data is removed, and the removed data is replaced by a sliding average;

[0095] (2) The torque data is corrected. Based on the coupling relationship between the blade speed and the torque, the filtered torque data is linearly corrected.

[0096] In the specific implementation, S3 is specifically as follows:

[0097] ​​​​​​​​​​​S3.1, Constructing a glassification degree correlation model based on the denoised accurate data, and the calculation formula is as follows:

[0098] ,

[0099] In the formula, represents the glassification degree, , when represents complete non-glassification, represents complete glassification; , , and represent weight coefficients, and the setting , , , , ;

[0100] represents a normalized function of the torque ; ;

[0101] represents a normalized function of the temperature ; ;

[0102] represents a correction function of the contact force ; ;

[0103] represents a correction function of the displacement ;

[0104] ,

[0105] wherein, represents the initial value of the torque, represents the torque corresponding to the glassification temperature , represents the ambient temperature, represents the optimal bonding contact force, and represent the minimum and maximum values of the displacement, represents the optimal bonding displacement, represents the agglomeration correction coefficient;

[0106] S3.2, Real-time updating the weight coefficients in the glassification degree calculation by using the recursive least square method to obtain the updated glassification degree;

[0107] First, data initialization is performed, and the normalized function of the torque , the normalized function of the temperature Normalized function, contact force Correction function and displacement The correction function is used as the initial output vector. , Weighting coefficients As the initial weight coefficient vector ,correspond Then, recursively update the data based on the input vector from the previous moment. Calculate the augmented matrix at the current time step. Then, based on the augmented matrix at the current moment and the weight coefficient vector of the previous moment Calculate the current weight coefficient vector ;

[0108] The calculation formula is as follows:

[0109] ,

[0110] ,

[0111] ,

[0112] in, , express The torque normalization function at time t, express The temperature normalization function at time t, express Contact force correction function at time t, express The displacement correction function at time t. Indicates transpose. This represents the forgetting factor that changes according to the bonding process. This represents a reference value for the degree of vitrification based on offline experiments. Represents the identity matrix.

[0113] In a specific implementation, S4 is as follows:

[0114] Based on the dynamic characteristics of the bonding machine stirring system, a physical model of "mass-damping-spring" for impedance control is constructed, where "mass" refers to the equivalent mass. The moment of inertia corresponding to the agitator shaft and blades of the bonding machine; "damping" refers to the damping coefficient. It is positively correlated with the updated degree of glassiness, and "spring" corresponds to the spring stiffness. The contact force is positively correlated with the updated glass transition level; the desired contact force is determined based on the updated glass transition level. Then, based on the actual contact force Angular displacement of the blade angular velocity angular acceleration The physical model is determined, and the calculation formula is as follows:

[0115] ;

[0116] Among them, the updated glassiness The value of is related to the preset glass transition threshold. and Comparison, for equivalent quality Damping coefficient and spring stiffness Make dynamic adjustments. ,set up , ;

[0117] when At this time, corresponding to a low degree of glass transition, the damping coefficient is increased and the spring stiffness is decreased, while the equivalent mass remains unchanged;

[0118] when At that time, corresponding to the degree of glass transition, the damping coefficient and spring stiffness are linearly adjusted to correct the equivalent mass;

[0119] when At this time, corresponding to a high degree of glass transition, the damping coefficient is reduced, the spring stiffness is increased, and the equivalent mass is corrected.

[0120] In a specific implementation, S5 is as follows:

[0121] S5.1, Segmented PID-Fuzzy Composite Temperature Control Algorithm:

[0122] (1) When the temperature deviation is large, fuzzy control is adopted. The temperature deviation and the rate of change of deviation are used as inputs, and the temperature is rapidly increased or decreased according to the preset fuzzy rule table.

[0123] Temperature deviation is equal to the difference between the collected temperature and the glass transition temperature. ;

[0124] Deviation change rate ;

[0125] The fuzzy rule table has seven states corresponding to the temperature deviation and the rate of change of deviation: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. The corresponding outputs are seven control modes: strong cooling, medium cooling, weak cooling, hold, weak heating, medium heating, and strong heating.

[0126] (2) When the temperature deviation is small, PID control is used to further shorten the temperature deviation;

[0127] S5.2, Adaptive adjustment:

[0128] (1) Adjust the blade speed:

[0129] When , increase the speed;

[0130] When , keep the medium speed;

[0131] When , reduce the speed;

[0132] (2) Adjust the single bonding time:

[0133] According to the time of the glassification degree reaching the target glassification degree, correct the single time of the next batch , , Indicates the reserved stabilization time;

[0134] If the deviation of the last 3 batches meets the conditions, fix ; If it does not meet the conditions, recalculate.

[0135] Example 2

[0136] The method in the application is applied, and the application process is as follows:

[0137] (1) Device configuration

[0138] The main body of the bonding machine adopts HM200 CM800 bonding machine, equipped with cold-hot segmented bonding barrel (inner diameter 500mm, barrel height 800mm, cold-hot segmented volume ratio 1:1.2), which meets the demand of multi-zone temperature control;

[0139] Sensor: Six-axis force / torque sensor (model: ATI Nano17, measurement range 0-50N•m); PT100 temperature sensor is respectively arranged in the cold-hot segmented area of the bonding barrel, the core area of the powder material, and the inlet and outlet of the blade cooling hole, and collects real-time temperature-200℃-600℃; integrated speed sensor, model E6B2-CWZ6C, arranged on the blade driving shaft;

[0140] Control platform: CodeSys V3.5 development environment, matched with Simulink R2022b simulation platform and industrial PC (CPU i7-12700, memory 16GB).

[0141] (2) Implementation steps

[0142] Multi-source data acquisition: Configure the sensor and the EtherCAT communication of the bonder machine through the CodeSys platform, set the data acquisition period Δt=0.2 seconds, and collect torque, contact force, displacement and temperature in real time; Establish a dynamic variable mapping table to map the collected data to internal variables in the PLC;

[0143] Kalman filter processing:

[0144] (1) Construct the state equation and the observation equation;

[0145] (2) Data verification and correction: When the temperature sensor deviation in the same cold and hot block area is more than 0.5℃, the faulty data is removed in combination with the trend of the adjacent area, and the moving average is used instead; Based on the torque-speed coupling relationship, the torque is corrected, and the noise of the filtered torque is ±0.2N•m and the noise of the temperature is ±0.1℃

[0146] Glassification degree estimation: Construct a correlation model, set the forgetting factor λ=0.9 of the recursive least squares method, update the estimated value every 0.2 seconds, set =80℃, When the temperature reaches , the glass transition progress estimation value converges to 0.9±0.05;

[0147] Adaptive impedance control parameter adjustment: Construct a "mass-damping-spring" model, establish a relationship between the glassification degree, and adjust the mass-damping-spring according to the set glassification degree threshold;

[0148] Bonder process closed-loop control: When the temperature deviation is 3℃, use fuzzy control to increase the hot oil supply ratio from 50% to 70%, and reduce the deviation to 1.5℃ within 10 seconds; Then switch to PID control, P=6.5, I=0.2, D=0.8 to stabilize the temperature at 80±0.3℃, and adjust the blade rotation speed from 300r / min to 450r / min.

[0149] (3) Implementation effect

[0150] Through this embodiment, the temperature difference error in the bonding process is ±0.8℃, the single bonding time is 18 minutes, the metal pigment addition amount is 17.2%, the gloss effect of replacing electroplated silver reaches 720 degrees, and the equipment runs stably without powder caking phenomenon.

[0151] Example 3

[0152] In order to prove the beneficial effects of the present application, the present application is compared with the traditional PID method (control group 1) and the "offline + hybrid control" method (control group 2) for comparison test;

[0153] Test object: Select the same batch of metal powder coating (composition: epoxy resin 60%, curing agent 8%, metal pigment 15%, auxiliary agent 2%), respectively, using the method of the application (experimental group), traditional PID control method (control group 1), offline vitrification test + force / position hybrid control method (control group 2) for bonding experiment, each group of experiment is repeated 5 times, and the average value is taken.

[0154] Test equipment: HM200 CM800 bonder is used in three groups of experiments, the sensor configuration is consistent (the experimental group is deployed according to the application, and the control group 1 / 2 only retains the temperature and torque sensors), and the control platform is CodeSys + application algorithm (experimental group), traditional PLC + single PID (control group 1), and PLC + offline test software (control group 2).

[0155] Determine evaluation index: temperature control accuracy (±℃), torque overshoot (%), vitrification degree estimation lag time (seconds), upper limit of metal pigment addition amount (%), single bonding time (minutes), product pass rate (%), no caking and glossiness ≥ 650 degrees are qualified.

[0156] The experimental results are shown in Table 1.

[0157] Table 1 Comparison results of the application and the control group

[0158]

[0159] As shown in Table 1, the temperature control accuracy of the experimental group (±0.3℃) is much higher than that of the control group 1 (±1.5℃), and is slightly better than that of the control group 2 (±0.8℃), so that the segmented PID-fuzzy compound temperature control combined with multi-source data filtering can effectively suppress temperature fluctuation; the torque overshoot (2.5%) is significantly lower than that of the control group 1 (8.2%) and the control group 2 (5.7%). In terms of real-time and flexibility, the vitrification degree estimation lag of the experimental group is only 0.2 seconds, and dynamic tracking is realized, while the control group 2 relies on offline test and cannot adapt to the dynamic changes of the bonding process; the control group 1 has no online estimation function and can only be controlled according to fixed parameters, resulting in a low upper limit of metal pigment addition amount (10.5%), while the experimental group breaks through to 17.2% through real-time estimation. In terms of production efficiency and product quality, the single bonding time of the experimental group (18 minutes) is shortened by 29% compared with the control group 1 (25.3 minutes) and by 21% compared with the control group 2 (22.7 minutes), because the closed-loop cooperative control optimizes the speed and time; the product pass rate (98%) is much higher than that of the control group 1 (72%) and the control group 2 (85%), which proves that the comprehensive performance of the method of the application is better.

[0160] In conclusion, the present application is superior to the prior art in control precision, real-time performance, production efficiency and product quality through multi-source information fusion and adaptive intelligent control.

[0161] The above describes the specific embodiments of the application in combination with the drawings, but is not a limitation on the protection scope of the application, and various modifications or changes made by those skilled in the art on the basis of the technical solutions of the application without creative labor are still within the protection scope of the application.

Claims

1. A multi-source information fusion-based powder coating bonder self-adaptive intelligent control method, characterized in that, The method comprises the following steps: S1. Real-time acquisition of multi-source data: real-time acquisition of data through sensors arranged on a powder coating bonding machine, configuration of a communication link between the sensors and the bonding machine using a CodeSys interactive platform, establishment of variable mapping to store the acquired data; S2, multi-source data preprocessing: based on the Kalman filtering algorithm, a high-precision filtering model suitable for the bonding machine is constructed, the collected data is input into the filtering model for denoising, and accurate data after denoising is obtained; S3, on-line estimation of the glass transition degree of the powder coating: based on the recursive least squares method, a glass transition degree estimation algorithm is constructed, the accurate data after denoising is taken as input, and the glass transition degree of the powder coating is calculated in real time through recursive updating; S4, adaptive impedance control parameter adjustment: based on the estimation result of the glass transition degree, an adaptive impedance control model is constructed, the adaptive impedance control model is based on the "mass-damping-spring" physical model, and the impedance control parameters of the adaptive impedance control model are dynamically adjusted according to the glass transition degree; S5, closed-loop control of the bonding process: based on the adjusted impedance control parameters, combined with the segmented PID-fuzzy compound temperature control algorithm, the blade rotation speed of the bonding machine and the temperature supply of the cold-hot segmented blocks of the bonding barrel and the single length of the bonding production are adjusted in real time.

2. The multi-source information fusion based adaptive intelligent control method for a powder coating applicator according to claim 1, characterized in that, S1 is as follows: (1) Data acquisition: the sensors include a high-precision six-dimensional force / torque sensor, a temperature sensor, and an integrated rotation speed sensor; The high-precision six-dimensional force / torque sensor is arranged at the connection end of the bonding machine stirring shaft and the driving motor, and simultaneously acquires dynamic torque signals, axial / radial contact force signals, and blade displacement signals during stirring; The temperature sensor is arranged in the cold-hot segmented block area, the powder material core area, and the inlet and outlet of the blade cooling hole of the bonding barrel, and acquires real-time temperatures at each location; The integrated rotation speed sensor is arranged on the blade driving shaft, and acquires real-time rotation speed of the blade; (2) Communication link and data storage: Based on the CodeSys platform, the communication link between the sensors and the programmable logic controller of the bonding machine is configured, industrial Ethernet protocol is adopted, dynamic variable mapping is performed on the collected data, the coupling relationship between the data is constructed, and the data is stored through a dynamic data caching mechanism.

3. The multi-source information fusion based adaptive intelligent control method for powder coating bonding machines according to claim 2, characterized in that, S2 is as follows: Based on the Kalman filtering algorithm, a high-precision filtering model suitable for the bonding machine is constructed, the torque, force, displacement, and temperature data in the collected data are taken as the input of the filtering model, the noise caused by device vibration and environmental interference is eliminated by constructing a state equation and an observation equation, and the accurate data after denoising is output after the data is verified and corrected; S2.1, construction of a high-precision filtering model: The state equation is as follows: ， The observation equation is as follows: ， In the formula, represents State vector composed of torque, force, displacement and temperature data at time t, represents State vector composed of torque, force, displacement and temperature data at time t; represents Process noise matrix at time t; represents State transition matrix, which is essentially a diagonal matrix based on the setting of the bonding dynamics characteristics; represents Control input matrix, which is essentially the influence coefficient of blade speed on state; represents Blade speed at time t; represents Original data collected by sensors at time t; represents Observation matrix, which is essentially a unit matrix, represents Observation noise matrix at time t; wherein, each element of the vector obeys a Gaussian distribution with mean 0 and covariance each element of the vector obeys a Gaussian distribution with mean 0 and covariance each element of the vector obeys a Gaussian distribution with mean 0 and covariance the covariance coefficients of and are calculated by a sliding window size to calculate the variance of the original data, and dynamically correct and the covariance coefficients of and and ; S2.2, verification and correction of data: (1) Correct the temperature sensor, if the data deviation between two temperature sensors in the same cold-hot segmented block area exceeds the set threshold, perform abnormal judgment on the two temperature sensors, remove the faulty temperature sensor data according to the temperature variation trend of the adjacent area, and replace the removed data with a sliding average; (2) Correct the torque data, based on the coupling relationship between the blade speed and the torque, linear correction of the filtered torque data.

4. The multi-source information fusion based self-adaptive intelligent control method of a powder coating applicator according to claim 3, wherein S3 Specific as follows: S3.1, based on the denoising of accurate data to construct the glassification degree correlation model, the calculation formula is as follows: ， wherein represents the degree of vitrification, , represents complete non-vitrification, represents complete vitrification; , , and represents a weighting factor, ; represents a normalized function of the torque ; represents a normalized function of the temperature ; represents a correction function of the contact force ; represents a correction function of the displacement ; S3.2, using recursive least squares method to update the weight coefficient in the calculation of glassification degree in real time, get the updated glassification degree; First, data initialization is performed, and a normalization function of torque , a normalization function of temperature , a correction function of contact force , and a correction function of displacement are taken as an initial output vector , , and a weight coefficient is taken as an initial weight coefficient vector corresponding ; then, recursive update is performed, an augmented matrix at the current time is calculated from an input vector at the previous time, and a weight coefficient vector at the current time is calculated from the augmented matrix at the current time and a weight coefficient vector at the previous time. The calculation formula is as follows: ， ， ， wherein , denotes a torque normalization function at time instant, denotes a temperature normalization function at time instant, denotes a contact force correction function at time instant, denotes a displacement correction function at time instant, denotes the transpose, denotes a forgetting factor that varies according to a forgetting process, denotes a glassiness reference value based on offline experiments, denotes the identity matrix.

5. The multi-source information fusion based adaptive intelligent control method for powder coating bonding machines according to claim 4, characterized in that, S4 is as follows: Based on the kinetic characteristics of the bonded machine stirring system, a "mass-damping-spring" physical model of impedance control is constructed, "mass" is the equivalent mass , corresponding to the rotational inertia of the bonded machine stirring shaft and the blade, "damping" is the damping coefficient , and "spring" corresponds to the spring stiffness , which is positively correlated with the updated glass transition degree; the expected contact force is determined according to the updated glass transition degree , and then the actual contact force , the angular displacement of the blade , the angular velocity , and the angular acceleration are determined to determine the physical model, and the calculation formula is as follows: ； Wherein the updated glassification degree is compared with a preset glassification degree threshold And The equivalent mass , damping coefficient And spring stiffness Dynamic adjustment, ; When corresponding low glass transition, increase the damping coefficient, reduce the spring stiffness, equivalent mass remains unchanged; When corresponding to the degree of vitrification, linearly adjusting the damping coefficient and the spring stiffness, and correcting the equivalent mass; When corresponding high glass transition, reduce the damping coefficient, increase the spring stiffness, correction equivalent mass.

6. The multi-source information fusion based self-adaptive intelligent control method of a powder coating applicator according to claim 5, characterized in that, S5 is as follows: S5.1, piecewise PID-fuzzy compound temperature control algorithm: (1) Temperature deviation is larger stage, using fuzzy control, temperature deviation and change rate of deviation as input, according to the preset fuzzy rule table to carry out rapid heating or cooling; Temperature deviation is equal to the difference between the collected temperature and the glass transition temperature; (2) Temperature deviation is smaller stage, using PID control, continue to shorten the temperature deviation; S5.2, adaptive adjustment: (1) Adjust the blade speed: When the rotational speed is increased; When the intermediate speed is maintained; When the rotational speed is reduced; (2) Adjust the single bonding time: According to the time to reach the target vitrification level , correct the single duration of the next batch , , denotes the reserved stabilization time; If 3 consecutive batches If the deviation meets the conditions, then fix it. If the conditions are not met, the calculation will be repeated.

Images