A method and system for controlling interior wall tile glaze grinding based on intelligent analysis of energy consumption

Abstract

The present invention proposes a method and system for controlling interior wall tile glaze grinding based on intelligent analysis of energy consumption, and belongs to the technical field of ceramic tile process control. Aiming at the problems of high energy consumption and difficult quality control in the ceramic tile glaze process, the present invention designs a dual-channel control architecture for energy optimization and product quality. Each channel adopts a hybrid scheme in which a controller provides basic control, a network performs predictive modeling, and a controller implements optimized control. By designing a special channel coordination mechanism, the coordination of energy consumption optimization and quality control is achieved. Simulation results show that compared with traditional control, this method reduces the mean square error by 62% and energy consumption by 21.6% while ensuring the consistency of glaze layer thickness, proving its effectiveness.

Description

Technical Field

[0001] The invention relates to the field of ceramic tile process control, and in particular to an interior wall tile glaze grinding control method and system based on energy consumption intelligent analysis. Background Art

[0002] The ceramic tile industry has a wide range of applications in architecture, decoration, and infrastructure. The production process of ceramic tiles is complex, including multiple stages such as raw material preparation, molding, firing, glazing, and finishing. Among them, the glazing process is particularly important because it determines the final aesthetic and functional properties of the tile, such as color, gloss, texture, and wear resistance. In recent years, the emphasis on sustainable development and the need to improve productivity have driven innovation and efficiency improvements in the ceramic tile manufacturing industry. However, the glazing process, especially glazing through a rotating disk system, presents unique challenges due to its high energy consumption and the nonlinear characteristics of the process dynamics. The purpose of this invention is to design and implement a more robust, energy-efficient, and adaptive controller for the ceramic tile glazing process by adopting a novel hybrid approach.

[0003] The glazing process, like many other industrial processes, is complex and dynamic. The glaze application phase is an energy-intensive phase in the glazing process. Conventional glazing systems use a high-speed rotating disk, combined with compressed air, to disperse the glaze over the tile surface. This process has many factors that affect its performance: these factors include, but are not limited to, the speed of the rotating disk, the air pressure, the properties of the glaze, the distance between the nozzle and the tile surface, and the conveyor speed on which the tiles move. Due to the interaction of all these factors, the process is difficult to accurately model using traditional mathematical methods, which makes the design of an effective controller particularly challenging. In addition, the process exhibits strong nonlinear characteristics, where the relationship between the input (control parameters) and the output (glaze layer thickness and uniformity) changes with operating conditions. This inherent nonlinearity limits the effectiveness of traditional linear controllers such as the widely used controller), resulting in suboptimal energy consumption and inconsistent product quality. Although easy to implement, controllers often have difficulty achieving consistent results under different process conditions, requiring constant manual adjustments by skilled operators. This further leads to significant losses in productivity and further exacerbates energy waste.

[0004] One of the main challenges of the glaze grinding process is balancing the competing demands of energy efficiency and product quality. Minimizing energy consumption requires reducing the power usage of the rotating disk and the air compressor as well as the amount of glaze used, which can lead to insufficient or uneven glaze application. On the other hand, ensuring high product quality requires precise and uniform application of the glaze, which often leads to higher energy consumption and excessive glaze application. Therefore, an effective control system needs to address both aspects at the same time, namely achieving high energy efficiency and maintaining good product quality. This is a complex multi-objective optimization problem that cannot be well solved by traditional single-loop control structures, which is exactly the problem that the present invention aims to solve.

[0005] To address the above challenges, the present invention proposes a novel hybrid control approach that integrates the advantages of multiple control strategies. The architecture proposed in the present invention is based on a dual-channel approach. This dual-channel architecture separates energy optimization from product quality control and handles both objectives simultaneously. One channel is dedicated to minimizing energy consumption, while the other channel focuses on maintaining the required product quality, i.e., consistent glaze thickness (which is the most important quality indicator). The two channels will operate in coordination to ensure that the system is both efficient and effective, while also allowing simple parameter configuration adjustments for different objectives.

[0006] In this architecture, the present invention adopts three key control technologies, namely Control, Model Predictive Control ( )and .Tradition The controller is used as the base control layer to provide basic process stability. On this stable foundation, the Perform model-based optimization to improve control accuracy and performance. In addition, rely The network provides accurate and long-term predictions of process dynamics, thereby improving the adaptability of the controller. , hybrid controllers address the challenge of modeling complex nonlinear dynamics without resorting to complex mathematical modeling. This approach provides a powerful tool for identifying and predicting how different operating factors interact, enabling more advanced predictive control. Summary of the invention

[0007] In order to solve the above problems, on the one hand, the present invention provides a method for controlling the glaze grinding of interior wall tiles based on intelligent analysis of energy consumption. The flow chart of the method is as follows: Figure 1 As shown, through dual-channel operation, each channel uses - - Hybrid control scheme is committed to optimizing energy consumption and product quality. The following are the detailed steps of this control method:

[0008] Step 1: Real-time data acquisition and comprehensive pre-processing of sensor readings to achieve accurate input to the control system.

[0009] Step 1.1: Accurate sensor data collection of glazing process: The system first collects real-time data from various high-precision sensors, including glaze flow rate ( ), air pressure used to disperse glaze ( ), the disc speed that affects the atomization and dispersion of the glaze ( ), the thickness of the glaze layer on the surface of ceramic tiles as a key quality indicator ( ), and the conveyor belt speed of the bricks during the glazing process ( ), ensuring accurate measurement data for further processing and control.

[0010] Step 1.2: Advanced noise filtering and data cleaning: To remove high-frequency noise and other unwanted interference from sensor measurement errors, the raw sensor data is fed into the pre-processing module where sophisticated noise filtering techniques are applied such as moving average filters for smoothing short-term fluctuations, median filters for effectively removing impulse noise, and low-pass filters for blocking high-frequency noise to reduce noise and ensure signal clarity. Missing data points are then processed using techniques such as interpolation to fill in the gaps, and outliers are removed using methods such as standard deviation thresholding to eliminate erroneous measurements. All these steps ensure that the data is representative of the underlying process.

[0011] Step 1.3: Data Normalization for Stable and Efficient Processing: To ensure that all data are of comparable scale when training the model and to improve the stability of the control algorithm, data normalization techniques are used, such as scaling the features to a minimum-maximum scale in the range of 0 to 1, or standardization to ensure that each variable has a mean of 0 and a standard deviation of 1, to rescale all input variables to ensure consistent scales, which can achieve better stability and also make The training of the model and the performance of the controller are more stable.

[0012] Step 2: Energy-optimized channel control to reduce the overall energy consumption of the glazing process.

[0013] Step 2.1: Basic control of the controller: Provide a stable benchmark for the control system, The controller calculates the basic control signal based on the error of the current energy efficiency target , the controller expression is: ,in express The basic control output of the controller, It represents the error between the expected value of energy efficiency target and the measured process variable. represents the proportional gain used for instant adjustment, represents the integral gain used to eliminate steady-state error, Represents the derivative gain that predicts future errors based on the rate of change.

[0014] Step 2.2: Use Network for advanced energy consumption prediction: using trained Model predicts future energy consumption based on current and historical inputs such as air pressure, disc speed, and glaze flow , enabling the model to capture short-term and long-term dependencies in this data and make long-term predictions on how these variables affect energy consumption, thus providing Provides the necessary predictions to perform model-based control.

[0015] Step 2.3: Use model predictive control ( ) Optimize energy consumption: Model predictive controller ( )use Model-generated predicted energy consumption , combined with basic The control signal is used as input to formulate and minimize the cost function in the prediction time domain while complying with various operating constraints of the system, which can be expressed as: ,in represents the total cost to be minimized, represents the predicted system behavior based on the system state at time k, represents the expected reference value of the system output, Indicates changes in control actions. is the prediction horizon that defines the future prediction time window, is the control time domain that defines the window for calculating control actions, is the weight matrix for adjusting error sensitivity, is the weight matrix of the control effort. After minimizing the cost, the output will be the optimal air pressure. and disc speed , to minimize the energy consumption while considering all the operating constraints of the system.

[0016] Step 3: Product quality channel control to maintain the required glaze thickness on the brick surface.

[0017] Step 3.1: Use Perform basic quality control: provide basic control operations for glaze thickness, standard The controller calculates the basic control action using the same formula as in step 2.1, with a different target to meet the quality requirement.

[0018] Step 3.2: Use Network prediction of future glaze thickness: dedicated The model uses the conveyor speed, glaze flow rate and the control output from the energy optimization channel ( , ) historical data to train and predict the future glaze thickness , which enables the model to effectively understand and capture how all these variables interact to affect glaze thickness.

[0019] Step 3.3: Use model predictive control ( ) Optimize glaze application: The controller uses its dedicated Model-predicted glaze thickness , combined with the basic Control and control output from the energy optimization channel and , minimizing a cost function with the same formula as in step 2.3, subject to the operating constraints, the final output of the controller will be the optimal glaze flow rate to maintain the desired glaze layer thickness .

[0020] Step 4: Actuator control, which implements the calculated control action through physical changes in the system.

[0021] Step 4.1: Send control signals to the system actuators: The optimized control signal calculated by the controller is the air pressure from the energy optimization channel , disc speed , and glaze flow from the product quality channel , transmitted to their respective actuators in real time.

[0022] Step 4.2: Precise adjustment of physical parameters in the glazing process: These actuators physically adjust the operating parameters of the glazing process according to the control signals, including adjusting the air compressor to achieve the required air pressure, adjusting the rotary motor to change the speed of the rotating disc, and adjusting the glaze metering valve to change the glaze flow rate, realizing real-time control of physical variables based on computer control output.

[0023] Step 4.3: Real-time physical response based on computational decision: Through this action, the physical parameters of the glazing process will change accordingly, realizing the physical implementation of the control strategy calculated by the system, thus forming a closed-loop control system.

[0024] Step 5: Accurate dual-channel coordination to ensure coordinated control between energy saving and product quality.

[0025] Step 5.1: Seamless information sharing between control channels: Product quality channel The controller calculates the air pressure for the energy-optimizing channel and disc speed Seamlessly integrated into its own control calculations, this ensures that the Product Quality channel is fully aware of adjustments made to control parameters by the Energy Savings channel, allowing it to make control decisions to compensate for these changes and maintain consistent glaze thickness.

[0026] Step 5.2: Coordinate control actions to maintain quality while saving energy: This two-way communication enables the two channels to dynamically adjust to ensure a balanced optimization strategy. The product quality channel will compensate for changes made by the energy efficiency channel in real time, ensuring that energy savings do not negatively impact product quality.

[0027] Step 5.3: Achieve system coordination to meet conflicting requirements: Through the above mechanism, the two control channels work together to ensure that the system operates with minimum energy consumption and high product quality without negatively affecting each other, thus achieving a fully coordinated control architecture.

[0028] Step 6: Continuous iteration and feedback of glaze process adaptive control performance.

[0029] Step 6.1: Repeat the control loop execution to achieve dynamic operation: The entire control process, from sensor data acquisition (step 1) to actuator control (step 4), is continuously executed in a closed loop, which enables the system to dynamically react to any changes in the process state or external environment and ensures that the system maintains consistency in long-term operation.

[0030] Step 6.2: System Adaptation Based on Feedback Control Loop: By closing the loop and continuously executing the control strategy based on real-time sensor data, the system operates in a feedback manner to adapt to changes, allowing the system to maintain high performance in the face of various disturbances, while allowing the system to continuously track its control target through real-time adjustments and enabling the system to automatically correct errors.

[0031] Step 6.3: Ensure stable and consistent performance during long-term operation: The controller operates based on the above mechanism to achieve stable and consistent control during long-term operation, ensuring stable execution of the brick glazing process under different conditions and maximizing energy savings with little human intervention while maintaining the required product quality.

[0032] On the other hand, the present invention also proposes an interior wall tile glaze grinding control system based on energy consumption intelligent analysis. The structure diagram of the system is shown in FIG. Figure 2 As shown, it includes a data acquisition preprocessing module, an energy optimization control module, a product quality control module, an actuator control module, a channel coordination module and an adaptive optimization module.

[0033] Among them, the data acquisition preprocessing module includes multiple high-precision sensors for collecting process parameters such as glaze flow, air pressure, disc speed, glaze layer thickness and conveyor belt speed; it is also equipped with a noise filtering device and a data normalization processing unit, which are used to eliminate measurement noise and standardize data respectively.

[0034] Energy optimization control module consists of Basic controller, Predictor and The optimized controller consists of The controller provides basic control signals based on the energy efficiency target error. The predictor is responsible for predicting future energy consumption trends. The optimization controller combines the outputs of the two to calculate the optimal air pressure and disc speed control strategy.

[0035] The structure of the product quality control module is similar to that of the energy optimization control module and also includes Controller, Predictor and Optimize the controller, but focus on the control of glaze thickness. The controller controls based on the glaze thickness error. The predictor predicts the thickness variation trend, The optimization controller generates the optimal glaze flow control signal.

[0036] The actuator control module consists of three parts: air pressure actuator, speed actuator and flow actuator. The air pressure actuator is used to adjust the output pressure of the air compressor, the speed actuator controls the speed of the rotating disc, and the flow actuator adjusts the opening of the glaze metering valve, which together realize the physical adjustment of the process parameters.

[0037] The channel coordination module includes an information sharing unit and a dynamic compensation unit. The information sharing unit is responsible for transmitting the output of the energy optimization control module to the product quality control module, while the dynamic compensation unit ensures that the product quality control module can effectively compensate for the impact of energy optimization.

[0038] The adaptive optimization module includes a feedback control unit, a parameter adjustment unit and a performance evaluation unit. The feedback control unit continuously monitors the system operation status, the parameter adjustment unit dynamically optimizes the control parameters according to the feedback information, and the performance evaluation unit is responsible for evaluating the control effect and guiding the system optimization.

[0039] Each module exchanges information through the data bus. The output of the data acquisition preprocessing module serves as the input of the energy optimization control module and the product quality control module. The outputs of the two control modules are coordinated by the channel coordination module and then transmitted to the actuator control module. The adaptive optimization module continuously monitors and optimizes the operating status of the entire system to jointly achieve the dual optimization goals of energy saving and quality.

[0040] The beneficial technical effects of the present invention include: solving the dual goals of minimizing energy consumption and maintaining consistent product quality in the ceramic tile glazing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The flowchart of the method proposed in the present invention is shown in FIG.

[0042] Figure 2 It is a structural diagram of the system proposed by the present invention. DETAILED DESCRIPTION

[0043] The glazing stage has a significant impact on the final product quality, production costs and environmental impact. The traditional control systems widely used in the ceramic tile industry are mainly based on the use of Simple feedback control structure of the controller, such as for disk speed control Controllers, air pressure control Controller. The controller is easy to implement and tune, but has known limitations when applied to complex and nonlinear systems such as glazing processes. The main disadvantage is its reliance on a linear approximation of the process dynamics.

[0044] However, the glaze application process is highly nonlinear, and the system response can change dramatically depending on operating conditions, such as different flow rates, changes in glaze viscosity, or changes in environmental conditions such as air humidity. When encountering such a nonlinear system, The performance of the controller outside of its design point will be severely degraded, often resulting in oscillations, significant deviations from the set point, and inconsistent product quality. The adjustment of parameters (which is necessary in the field to meet changing operating conditions) is itself a tedious and time-consuming process that relies on a trial-and-error approach and often fails to achieve optimal performance. The process control methods of have a number of disadvantages, not to mention that they completely ignore energy efficiency.

[0045] The nonlinear nature of the glazing process encourages us to go beyond traditional A promising approach is to use model predictive control ( ). It is a control strategy that explicitly uses a dynamic model of the process to predict the future behavior of the system. Based on this prediction, the controller calculates the optimal control sequence within a finite prediction horizon while taking into account the control input constraints and output constraints. At each time step, the first control action in the calculated sequence is applied to the process. The optimization is then repeated at the next time step, forming a closed-loop control system. The use of a dynamic model enables the controller to predict future changes in the process and optimize control actions accordingly.

[0046] although It has many advantages, but its performance is heavily dependent on the accuracy of the underlying dynamic model. For systems with highly nonlinear dynamics, such as the glaze application process, it is difficult to develop an accurate mathematical model due to the complexity of the process and the various factors that affect the dynamics (such as flow, air pressure and ambient humidity). The accuracy of such models is almost always limited by the approximations made when modeling, and changes in the operation process, such as glaze composition, environmental changes, etc., can also greatly affect the model accuracy, resulting in reduced control performance and limited generalization ability.

[0047] recently, All areas of the network receive significant attention. It is a network architecture that relies entirely on the self-attention mechanism. Its core self-attention enables the network to better capture nonlinear relationships in time series, which is a major advantage over recursive-based models.

[0048] The network consists of an encoder and a decoder, each of which has a self-attention module. The network's ability to model complex dynamic systems, coupled with its computational efficiency due to parallel processing, makes it an ideal candidate for integration into predictive controllers. The network performs well on sequence prediction problems and is therefore suitable as a Predictive model in the controller.

[0049] However, these previous studies neither focus on the ceramic tile glazing process nor consider the dual optimization objectives of energy efficiency and product quality, which are of great significance in the ceramic tile industry. In addition, most of the control schemes proposed in previous studies focus on single-channel control and thus fail to address multi-objective optimization problems in ceramic tile glazing applications. To address these limitations of the current state of the art, the present work proposes a novel dual-channel hybrid control architecture, combining , and , achieving more energy-saving and robust control performance in ceramic tile glazing technology.

[0050] The core idea of ​​the approach is to decompose the energy optimization and product quality control tasks into two independent control channels. This approach allows independent optimization of the individual objectives while coordinating their actions to achieve synergistic effects. The first channel, called the "energy optimization channel", focuses on minimizing the energy consumption of the glazing process. The second channel, called the "product quality channel", focuses on maintaining the required glaze layer thickness on the ceramic tiles. This dual-channel architecture is specifically designed to address the above-mentioned trade-off between energy efficiency and product quality.

[0051] At the heart of each channel is a hybrid control scheme that combines three key components: traditional Controller, Model Predictive Controller ( )and Model. The controller is used as the underlying control layer to provide stable basic control operations. The controller implements advanced model optimization based on this Build on the basis of control. Model used as The prediction module in the controller. Use The network helps by providing accurate and long-term predictions of process dynamics Achieve better performance, which is critical for effective model-based predictive control.

[0052] The overall architecture of the control system proposed by the present invention consists of two independent control channels working in parallel: an energy optimization channel and a product quality channel.

[0053] The Energy Optimization channel aims to minimize the total energy consumption of the glaze application process by reducing the use of compressed air and the rotation speed of the disc. During the glazing process, a lot of energy is used for compressed air (for pushing the glaze through the nozzle) and for rotating the disc at high speed (for better dispersion). The channel receives the current state of the process, mainly including the glaze flow rate, air pressure and disc rotation speed. Based on these inputs, the channel generates control commands to adjust the air pressure and the rotation speed of the disc. These control parameters will directly affect the glaze application and thus the energy consumption and product quality. The channel tries to minimize the energy consumption while keeping the glaze flow rate within an acceptable range and ensuring that there is no significant deviation from the desired product quality target.

[0054] The Product Quality channel is primarily focused on maintaining the desired glaze thickness on the tile surface. This channel receives as input the current state of the glazing process, including the measured glaze thickness, conveyor speed, and control parameters (air pressure and speed) from the Energy Optimization channel. Based on these inputs, the Product Quality channel calculates control commands to adjust the glaze flow rate to ensure that the actual glaze thickness is as close to the target glaze thickness as possible. This channel ensures that product quality is maintained even when the control inputs from the Energy Efficiency channel change.

[0055] The two channels work in coordination with each other. Specifically, the product quality channel receives not only the current state of the glazing process, but also the control parameters from the energy optimization channel as input, enabling the product quality channel to compensate for any changes introduced by the energy optimization channel. This coordination allows the controller to maintain the desired product quality while optimizing energy efficiency.

[0056] In each channel, the most basic control layer is the standard Controller, provides basic stability for the entire control system. Controllers are designed to adjust the control output based on the error between the desired set point and the measured process variable. The controller works the same way in both channels. Control output of the controller It is expressed as:

[0057] ;in, express controller output of the controller; is the error between the set point and the current value of the target control variable; is the proportional gain; is the integral gain; is the differential gain. The specific gains of the controller may vary and may be adjusted manually or through some automatic algorithm. The output of the controller is then used as the next layer (the model predictive controller ) input. The controller is always the basic control of any control loop, ensuring the basic stability and tracking ability of the system.

[0058] In basic On top of the controller, each channel uses a model predictive controller ( ).here The main purpose of is to perform more refined and optimized control by considering system dynamic models and future predictions. It works by computing the optimal sequence of control actions over a finite prediction horizon, taking into account the current state of the process, the desired future behavior, constraints, and other factors. Unlike traditional feedback controllers, The future behavior of the system is considered when making control decisions, which yields better performance in terms of dynamic response while also allowing multi-objective optimization to be handled in a more systematic manner.

[0059] exist The optimization process in involves minimizing a cost function, which represents the trade-off between achieving the desired set point and minimizing the control effort. The control action is calculated by minimizing the following cost function:

[0060] ;in, is the cost function to be minimized; is the predicted output at time k+i given the system state at time k; is the expected reference output at time k+i; is the change in control action at time k+i-1; is the prediction horizon, i.e., the time range for predicting future system behavior; is the control time domain, i.e. the time range of the optimized control action; is the weight matrix of the output error, which determines the importance of matching the desired set point; is the weight matrix of the control inputs, which determines the importance of minimizing the change in the control output. The optimization problem is also subject to various constraints, such as bounds on the control inputs and control outputs.

[0061] Energy Optimization Channel Controller Usage The model predicts energy consumption based on the inputs of air pressure, disk speed, and frit flow rate. Calculates the optimal air pressure and speed to minimize the predicted energy consumption while respecting the operating constraints. The controller predicts the glaze thickness based on the conveyor speed and the output of the energy optimization channel. Calculate the optimal glaze flow rate to maintain the required glaze thickness on the brick surface, while also considering the changes in the process caused by the energy efficiency channel.

[0062] The innovation here is how the present invention converts Structure and The network combines and combines it with two channels, each channel optimizing a different objective. Unlike traditional control systems that focus only on output tracking and disturbance rejection, the hybrid approach of the present invention not only allows these objectives, but also adds the goal of energy efficiency at the same time. This allows optimization to be performed based on future predictions rather than reacting to past values, resulting in a smoother and more stable system. In addition, the use of predictive modeling makes this controller very adaptable to a variety of operating conditions, as these dynamic behaviors of the system can be determined by The model itself captures.

[0063] A key component of the control architecture proposed in this invention is the use of The network is used to simulate the dynamic characteristics of the glaze process. In the past, the process model is usually based on some mathematical equations, but considering the complexity of the glazing process, it is almost impossible to develop an accurate mathematical model. The network captures the nonlinear relationships between input and output variables directly from the process data, eliminating the need to manually develop complex mathematical models. The network is specifically designed for sequence modeling, making it suitable for predicting future process behavior based on historical data.

[0064] The method used in the present invention The model is based on an encoder-decoder structure and will be trained using data collected from the system. The encoder takes as input a sequence of recent historical values ​​of the input variables and maps it to a fixed-length embedding that encodes the temporal and inter-dependency information in the input sequence. The decoder receives as input the encoder’s embedding and the output of the previous time step to predict the output value for the next time step. This process is repeated until a complete sequence of predictions is generated. The predictions are used by the model predictive controller as an internal model to predict future control actions and their respective effects.

[0065] The self-attention mechanism is The core component of the model that enables the network to automatically learn the temporal dependencies between input variables. The self-attention mechanism works by computing the weighted sum of input features, with the weights determined by the correlations between different parts of the input. The self-attention mechanism can be expressed as:

[0066] ;in, is the query matrix, representing the input currently being evaluated; is the key matrix, representing the past inputs compared with the current input; is the value matrix, which is the input itself weighted according to the attention score; is the dimension of the key matrix, which helps in scaling the attention matrix.

[0067] Energy Optimization Channel The model takes historical data of air pressure, disc speed, and glaze flow rate as input, and then outputs a forecast of future energy consumption. The model receives as input historical data such as conveyor speed, glaze flow rate, and control outputs from the energy optimization channel, and then outputs a prediction of future glaze thickness. The parameters of the model will be learned from the training data, and the learned model will be used It has very accurate long-term prediction capabilities. This combination of data-driven deep learning and classical model predictive control enables the proposed method to cope with model uncertainty and strong nonlinear characteristics in the system.

[0068] The training of the network requires the collection of real data from the ceramic tile glazing process. The accuracy and robustness of the final controller directly depends on the quality and diversity of this training data. Therefore, the data collection strategy is crucial. The data should cover the full range of operating conditions that the glazing process is expected to encounter in real scenarios. For the energy optimization channel, this includes data on glaze flow, air pressure, disc speed and corresponding energy consumption values, as well as external environmental variables such as temperature and humidity. The collected data should have a wide range of values ​​so that The model generalizes well to the full range of operating conditions. For the product quality channel, the data should include glaze thickness measurements, conveyor speed, and control outputs from the energy optimization channel.

[0069] The sensors for data collection must be carefully selected to ensure high accuracy and reliability, as any inaccurate measurements will introduce errors into the training data. For measuring glaze flow, a high-precision flow meter should be used. For measuring air pressure, an accurate pressure sensor with high sensitivity should be used. For measuring rotational speed, a tachometer should be used. For measuring the thickness of the glaze layer on the brick surface, an optical sensor can be used. All sensors need to be carefully calibrated before collecting data. These sensors must be connected to a data acquisition system to collect and store data. It is important to note that multiple sensors must be synchronized when performing real-time data collection, especially when performing control tasks. It is also important to note that data should be collected when the system is operating under normal and abnormal conditions (such as the system warm-up period) to collect a more complete data set for training.

[0070] To ensure data quality, noise filtering and data preprocessing steps are performed, including handling missing data, handling outliers, removing measurement noise, and data normalization. Missing data points can be filled based on nearby values ​​or through interpolation techniques. Outliers can be removed by removing data that deviates from the mean by a number of standard deviations. Raw sensor data may also be affected by measurement noise, which can be reduced by moving average filters, median filters, or low-pass filters. Finally, data normalization can be performed to ensure that all features are consistent during training. This is very important because some sensor readings have completely different magnitudes, which is problematic for network training. After the data has been cleaned, processed, and normalized, they can be divided into training, validation, and test datasets for Network training.

[0071] The model is trained using the prepared dataset. The training process can be performed offline using the collected historical data. This process includes architecture definition, hyperparameter tuning, loss function definition and optimization. The training process is divided into two stages: first, the energy efficiency channel is trained. model, and then train the product quality channel Model. Two The training process of the models is similar, the only difference is that each The input data received by the model is different.

[0072] The architecture of the model involves determining the number of encoder and decoder layers, the size of the attention heads, the hidden dimension of the fully connected layer, and the embedding dimension. The present invention uses 2 encoder layers and 2 decoder layers, 4 attention heads per layer, a hidden dimension of 128, and an embedding size of 64. These parameters will be determined empirically based on the validation set performance.

[0073] The training process uses the mean squared error (MSE) loss between the predicted output and the true value:

[0074] ;in, is the mean square error; is the total number of data points used for training; is the true value of the training point; is the predicted value of the model. During training, the parameters are learned using gradient descent with the goal of minimizing the above loss.

[0075] The optimization algorithm uses an optimizer with a learning rate of 0.001 and a batch size of 32. Early stopping is also implemented during training, i.e., if the validation error does not decrease after a certain number of epochs, the training is terminated. This can be achieved by monitoring the validation set performance and stopping the training when overfitting is observed. Once the training process is completed, the trained The model is Controller for online operation.

[0076] For product quality channels, The controller uses the trained The model predicts the future glaze thickness based on current measurements of conveyor speed, glaze flow and control outputs of the energy optimization channel. The controller optimizes the glaze flow rate to achieve the desired glaze thickness target while taking into account system constraints.

[0077] - The detailed steps of integration are as follows:

[0078] Sensor data collection: The system continuously collects sensor data including glaze flow, air pressure, disc speed, glaze thickness and conveyor speed. All these sensor data are , and Before inputting into the controller, the data is pre-processed by filtering and normalization to ensure data quality.

[0079] Control: Each control channel uses a The controller calculates the control signal. The output of the controller forms the basic control output of the system, providing initial stability for the system.

[0080] based on Prediction: Controller parallel, The model receives the recent history of sensor readings and generates predictions of the future behavior of the controlled variables. Specifically, the energy optimization channel Predict future energy consumption, and product quality channel Predict future glaze thickness.

[0081] optimization: Controller Usage The model's predictions, combined with The controller output is used to calculate the optimal control action by minimizing a cost function (such as the error between the setpoint and the predicted value) and taking into account the control effort and operating constraints. Outputs the control signal to the actuator.

[0082] Actuator control: The control signals calculated by the controller are used to physically act on the system by adjusting control parameters such as air pressure, disc speed, and glaze flow rate through actuators. The system then continues from step one by collecting new sensor readings.

[0083] The interactions between the three main components are: The controller provides baseline stability, Models provide accurate dynamic predictions of system behavior, while These predictions are used to achieve optimal control performance in terms of energy consumption and product quality. This integration is performed within each control channel to achieve synergy between all three components. Although the above describes the interaction as sequential steps, in actual real-time operation, this process is carried out concurrently. The controller continuously senses the environment, calculates control actions, and the actuators act simultaneously.

[0084] The two channels of the control system work together to achieve the dual goals of minimizing energy consumption and controlling quality. The coordination between the channels is achieved by inputting the control output of the energy optimization channel into the product quality channel. The controller uses the air pressure and disc speed control outputs of the Energy Optimizing Channel controller as part of its input when calculating the optimal glaze flow rate. By doing this, the Product Quality Channel can compensate for any changes in the glaze application process caused by adjustments made by the Energy Optimizing Channel. For example, if the Energy Optimizing Channel decides to reduce air pressure to save energy, the Product Quality Channel will adjust the glaze flow rate accordingly to maintain the desired glaze layer thickness.

[0085] The dual-channel architecture makes the control system more flexible and robust. This decoupling of the two control objectives simplifies the control design by focusing on different aspects of the system simultaneously. For example, the parameters in the energy optimization channel (i.e., their The weights in the controller can be adjusted to prioritize energy conservation without affecting the product quality channel and its objectives. Similarly, the parameters in the product quality channel can be adjusted to prioritize quality aspects without sacrificing energy efficiency. In practice, this dual-channel structure provides a simple way to optimize a specific goal by adjusting only a small number of parameters, rather than having to adjust a large number of parameters as in a single-channel control system.

[0086] The real-time operation process of the system is as follows: first, the sensor data is acquired and preprocessed in real time, and then the two control channels start to calculate their respective control signal, after which each channel The prediction model generates future predictions, and then each channel The calculations are applied in real time to the final control output of the actuator. This procedure is executed at each control cycle, forming a closed-loop control system. The coordination between the two channels means that the product quality channel is always aware of any control variable changes from the energy efficiency channel, effectively compensating for these changes by appropriately adjusting its control output, achieving stable operation of the entire system.

[0087] The simulations were performed in a numerical simulation environment that models the dynamics of the ceramic tile glazing process, including nonlinear effects, process constraints, and the impact of control parameters on energy consumption and product quality. The simulations were designed to verify the effectiveness of the proposed approach in achieving the twin goals of minimizing energy consumption while maintaining the desired product quality, specifically consistent glaze thickness on ceramic tiles.

[0088] The simulation environment is designed to be as close to the real glazing process as possible. The model consists of three parts: glaze application model, energy consumption model and sensor model.

[0089] Glaze Application Model: Models the relationship between control inputs (air pressure, disk speed, and glaze flow) and outputs (i.e., glaze layer thickness). The model is based on a combination of physical equations and empirical approximations to represent the nonlinear dynamics of the process. Nonlinearities are modeled through a series of lookup tables that capture how the process dynamics change as operating parameters vary. Glaze Layer Thickness The expression is:

[0090] ;in, Indicates glaze flow rate; Indicates air pressure; Indicates the disc speed; Indicates the conveyor belt speed; Represents a nonlinear function. In actual simulation, the model parameters will be obtained from data collected on site, and different parameters can be assigned to reflect different operating conditions, different glaze materials, and different brick surfaces. Function Capturing the nonlinear relationships between these variables can be approximated using spline interpolation from a series of lookup tables based on field collected data.

[0091] Energy consumption model: Calculates the energy consumed by the air compressor and disc rotation system, which is mainly related to air pressure, disc speed and running time. Total energy consumption It can be approximated as:

[0092] ;in, and are coefficients related to the energy consumption of the air compressor and the disk rotation system, respectively, depending on the specific machine. The numerical integration is performed in the simulation program using discrete time steps.

[0093] Sensor model: simulates sensor measurements and adds Gaussian noise to represent the measurement error of the actual sensor. The Gaussian noise added to the sensor value can be expressed as ,in is the standard deviation of the noise level, the present invention will select different The robustness of the controller under different noise levels is evaluated.

[0094] Simulation parameters such as prediction time domain, control time domain, weight matrix and as well as The controller gains are empirically determined through a series of simulation runs. The simulation is performed for a fixed time, and the present invention analyzes various indicators, such as the mean square error (MSE) of the glaze thickness, the average energy consumption during the period, and the dynamic response of different control parameters.

[0095] In order to evaluate the performance of the hybrid control system proposed in this invention, the traditional The control methods were benchmarked. This single channel The controller adjusts the glaze flow rate only according to the error between the desired glaze thickness and the actual glaze thickness. The control equation of the controller is as follows:

[0096] ;in, Indicates the controller output (glaze flow control), Indicates the error between the set point and the current measured glaze thickness. , and express Gain. The controller gains were chosen empirically through a trial and error process to give a controller with relatively good tracking performance and stability. This controller is a good representation of controllers used in industrial settings.

[0097] The execution time of the simulation is the same as that of the dual-channel controller. In the simulation, the target glaze thickness is set to a constant value. The performance of the methods is compared. The present invention focuses on two key performance indicators: the consistency of the glaze layer thickness (measured using mean square error) and the total energy consumption during the simulation.

[0098] The simulation results show that the dual-channel - - Hybrid control system outperforms traditional The results are shown in Table 1 below.

[0099] Table 1: Dual-channel hybrid control system and traditional Controller consistency comparison table:

[0100] ; The results show that the proposed method achieves a significant reduction in the mean square error of the glaze thickness, a 62% reduction compared to the baseline, which means that the actual glaze thickness in the proposed method is closer to the target set point. It also achieves significantly lower total energy consumption, a 21.6% reduction compared to the baseline. These results clearly show that the proposed method outperforms the traditional method in both objectives, achieving better control accuracy while consuming less energy.

[0101] The reason why the controller of the present invention performs better is that the present invention has a The prediction model makes Ability to think about future events, rather than In that way only the current error is reacted. In addition, the present invention not only optimizes the glaze thickness control, but also optimizes the energy consumption, thus achieving multi-objective control.

[0102] This paper proposes a method and system for controlling interior wall tile glaze grinding based on intelligent energy consumption analysis, which is used to optimize the ceramic tile glazing process and solve the two key challenges of achieving energy efficiency and consistent product quality. The proposed method takes advantage of the advantages of three control strategies: Controller, Model Predictive Controller ( )and The innovation lies in the architecture of the dual-channel control system, which separates the energy optimization task from the product quality control task and achieves concurrent optimization of both objectives.

[0103] The key to the control method proposed by the present invention is to use The network is used to capture the complex nonlinear dynamics of the glazing process. This eliminates the need for an exact mathematical model while also making it more robust to changes in process conditions. The network learns dynamic behaviors from real data and then ,These learned behaviors are used to predict future system responses to make better control actions, thereby providing more robust control performance.,The dual-channel structure ensures that even if the controller parameters of the energy efficiency channel are adjusted,,the controller of the product quality channel will compensate for these changes to maintain consistent,product quality, thus eliminating the cross-interference between the two objectives, which is a common problem in many existing control,architectures.

[0104] The structure of the control system includes standard controller, providing basic stability. On top of this basic stability, the present invention uses based on Future predictions of the network add optimized control behavior. The dynamic prediction model of the proposed method uses a self-attention mechanism, which enables the network to learn long-term and short-term temporal dependencies from historical data without complex recurrent connections. This enables the controller to better predict the future behavior of the system and therefore implement control strategies in advance rather than reacting to current values.

[0105] The simulation results show that the control system proposed by the present invention is significantly better than the traditional one in terms of energy efficiency and glaze thickness consistency. Specifically, the proposed method achieved a reduction of more than 20% in energy consumption, while also reducing the mean square error of glaze thickness by more than 60%. These results strongly support the effectiveness of the proposed method in solving multi-objective control problems in the nonlinear dynamics of glazing processes, which is a problem that traditional single-objective, single-channel The simulation also shows that the proposed controller is robust to noise measurements, which is crucial for practical implementation.

Claims

1. An energy consumption intelligent analysis method for controlling interior wall tile glaze grinding, characterized in that: The method specifically comprises the following steps: Step 1: Real-time data acquisition and pre-processing of sensor readings to achieve accurate input to the control system; Step 2: Energy consumption optimization channel control is used to reduce the overall energy consumption of the glazing process, including: Step 2.1: Utilize The controller performs basic control: The controller expression is: ,in express The basic control output of the controller, It represents the error between the expected value of energy efficiency target and the measured process variable. represents the proportional gain used for instant adjustment, represents the integral gain used to eliminate steady-state error, represents the derivative gain that predicts future errors based on the rate of change; Step 2.2: Use Network for advanced energy consumption prediction: using trained Model predicts future energy consumption based on current and historical inputs ; Step 2.3: Optimize energy consumption using model predictive control: The model predictive controller utilizes Model-generated predicted energy consumption , combined with basic The control signal is taken as input, and the cost function is formulated and minimized in the prediction domain: ,in represents the total cost to be minimized, represents the predicted system behavior at time k+i based on the system state at time k, represents the expected reference value of the system output at time k+i, represents the change of control action at time k+i-1, is the prediction horizon that defines the future prediction time window, is the control time domain that defines the window for calculating control actions, is the weight matrix for adjusting error sensitivity, is the weight matrix of the control effort. After minimizing the cost, the output will be the optimal air pressure. and disc speed ; Step 3: Product quality channel control, used to maintain the required glaze thickness on the brick surface, specifically including: Step 3.1: Use Perform basic quality control: provide basic control operations for glaze thickness, standard The controller calculates the basic control action using the same formula as step 2.1; Step 3.2: Use Network prediction of future glaze thickness: dedicated The model uses the conveyor speed, glaze flow rate and the control output from the energy optimization channel ( , ) historical data to train and predict the future glaze thickness , which enables the model to effectively understand and capture how all these variables interact to affect glaze thickness; Step 3.3: Optimizing glaze application through model predictive control: The model predictive controller uses its dedicated The thickness of the glaze layer predicted by the model , combined with the basic Control and control output from the energy optimization channel and , minimizing a cost function with the same formula as in step 2.3 while respecting the operating constraints, the final output of the controller will be the optimal glaze flow rate to maintain the desired glaze layer thickness ; Step 4: Actuator control, implementing the calculated control action through physical changes in the system; Step 5: Accurate dual-channel coordination to ensure coordinated control between energy saving and product quality; Step 6: Continuous iteration and feedback of glaze process adaptive control performance.

2. According to the energy consumption intelligent analysis method for controlling interior wall tile glaze grinding according to claim 1, it is characterized in that: Step 1 specifically includes: Step 1.1: Accurate sensor data collection of glazing process: First, real-time data is collected from various high-precision sensors, including glaze flow rate indicating the amount of glazing applied per unit time, air pressure used to disperse glaze, disc speed affecting glaze atomization and dispersion, thickness of glaze layer on the surface of ceramic tiles as a key quality indicator, and conveyor belt speed of bricks during glazing process, to ensure accurate measurement data for further processing and control; Step 1.2: Noise filtering and data cleaning; Step 1.3: Data Normalization: Scale features to a minimum-maximum scale between 0 and 1, or standardize to ensure that each variable has a mean of 0 and a standard deviation of 1, to rescale all input variables to ensure consistent scale.

3. The method for controlling interior wall tile glaze grinding by intelligent energy consumption analysis according to claim 1 is characterized in that: The step 4 specifically includes: Step 4.1: Send control signals to the system actuators: The optimized control signal calculated by the controller is the air pressure from the energy optimization channel , disc speed , and glaze flow from the product quality channel , transmitted to the respective actuators in real time; Step 4.2: Precise adjustment of physical parameters in glazing process: The actuator physically adjusts the operating parameters of the glazing process according to the control signal, including adjusting the air compressor to achieve the required air pressure, adjusting the rotary motor to change the speed of the rotating disc, and adjusting the glaze metering valve to change the glaze flow rate, realizing real-time control of physical variables based on computer control output; Step 4.3: Real-time physical response based on computational decision.

4. The method for controlling interior wall tile glaze grinding by intelligent energy consumption analysis according to claim 1 is characterized in that: The step 5 specifically includes: Step 5.1: Seamless information sharing between control channels: The model predictive controller of the product quality channel calculates the gas pressure of the energy optimization channel and disc speed Seamlessly integrated into its own control calculations; Step 5.2: Coordinate control actions to maintain quality while saving energy; Step 5.3: Systems collaborate to satisfy conflicting demands.

5. An interior wall tile glaze grinding control system based on intelligent energy consumption analysis, characterized in that: The system is used to execute the method described in any one of claims 1 to 4, and the system includes a data acquisition preprocessing module, an energy optimization control module, a product quality control module, an actuator control module, a channel coordination module and an adaptive optimization module. The data acquisition preprocessing module includes multiple sensors for collecting glaze flow, air pressure, disc speed, glaze thickness and conveyor belt speed; it is also equipped with a noise filtering device and a data normalization processing unit, which are used to eliminate measurement noise and standardize data respectively; Energy optimization control module consists of Basic controller, The predictor and the model predictive control optimization controller are composed of The controller provides basic control signals based on the energy efficiency target error. The predictor is responsible for predicting the future energy consumption trend, and the model predictive control optimization controller combines the outputs of the two to calculate the optimal air pressure and disk speed control strategy; The structure of the product quality control module includes Controller, Predictor and model predictive control optimization controller, focusing on the control of glaze thickness; The controller controls based on the glaze thickness error. The predictor predicts the thickness variation trend, and the model predictive control optimization controller generates the optimal glaze flow control signal; The actuator control module includes three parts: air pressure actuator, speed actuator and flow actuator; the air pressure actuator is used to adjust the output pressure of the air compressor, the speed actuator controls the speed of the rotating disc, and the flow actuator adjusts the opening of the glaze metering valve to achieve physical adjustment of process parameters; The channel coordination module includes an information sharing unit and a dynamic compensation unit. The information sharing unit is responsible for transmitting the output of the energy optimization control module to the product quality control module, while the dynamic compensation unit ensures that the product quality control module can effectively compensate for the impact of energy optimization. The adaptive optimization module includes a feedback control unit, a parameter adjustment unit and a performance evaluation unit. The feedback control unit continuously monitors the system operation status, the parameter adjustment unit dynamically optimizes the control parameters according to the feedback information, and the performance evaluation unit is responsible for evaluating the control effect and guiding the system optimization. Each module exchanges information through the data bus. The output of the data acquisition preprocessing module serves as the input of the energy optimization control module and the product quality control module. The outputs of the two control modules are coordinated by the channel coordination module and then transmitted to the actuator control module. The adaptive optimization module continuously monitors and optimizes the operating status of the entire system.

Images