Intelligent temperature control system and method for tunnel kiln

Abstract

The invention discloses an intelligent temperature control system and method for a tunnel kiln. The intelligent temperature control system and method are high in sintered product quality stability, high in environment change adaptability and capable of improving production efficiency. The intelligent temperature control system for the tunnel kiln comprises a multi-modal data acquisition layer used for acquiring the temperature, the environment and the loading capacity in the tunnel kiln; the control layer is used for collecting data of the multi-modal data acquisition layer and realizing frequency conversion control, combustion system adjustment and wind speed and wind pressure dynamic response adjustment based on a PLC (Programmable Logic Controller) and an industrial personal computer; the intelligent AI decision analysis layer comprises an embedded AI chip and adopts a deep reinforcement learning model to perform intelligent decision; and the data interaction layer is used for realizing human-computer interaction and information backtracking of the intelligent AI decision analysis layer and the control layer through butt joint of an HMI or a mobile terminal or an MES. The invention is applied to the technical field of tunnel kilns.

Description

Technical Field

[0001] The present invention relates to an intelligent temperature control system and method for a tunnel kiln, and particularly to an intelligent temperature control system and method for a tunnel kiln based on multi-modal data fusion and deep reinforcement learning. Background Art

[0002] A tunnel kiln is a kiln similar to a tunnel made of refractory materials, thermal insulation materials, and building materials, with transport tools such as kiln cars installed inside, and is a modern continuous firing thermal equipment.

[0003] Tunnel kilns are widely used in the roasting production of resin abrasive products. Resin abrasive production is a crucial field in manufacturing, and the quality of its products directly affects the performance and safety of downstream industries. As the core process, the curing of the tunnel kiln determines the strength, stability, and production efficiency of the products. However, the existing curing control methods have significant defects and are difficult to meet the production requirements of high precision and high efficiency. Traditional tunnel kilns rely on PLC control with fixed programs, which are difficult to adapt to environmental changes or loading differences, resulting in large temperature deviations, high energy consumption, many defects, and a curing cycle as long as 24 hours, restricting the improvement of production capacity. These limitations pose severe challenges to production enterprises in terms of quality consistency and delivery efficiency.

[0004] In this context, the accuracy and dynamic adaptability of temperature control have become the core issues that need to be solved urgently. Due to the lack of real-time multi-source data perception, existing systems are difficult to accurately capture the temperature distribution and environmental changes inside the tunnel kiln, resulting in a large deviation between the actual furnace temperature and the set value, and further leading to unstable product quality. The inaccurate temperature control further exacerbates the insufficient adaptability of the system to the loading volume and environmental changes. The fixed control strategy cannot be dynamically adjusted according to the actual production status, causing problems such as too slow or too fast heating and long holding time. These problems are interrelated and jointly limit the efficiency and safety of the curing process. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide an intelligent temperature control system and method for a tunnel kiln with high sintered product quality stability, strong environmental change adaptability, and improved production efficiency.

[0006] The technical solution adopted by the present invention is: The present invention includes an intelligent temperature control system and method for a tunnel kiln. The intelligent temperature control system for a tunnel kiln includes: A multi-modal data acquisition layer: used to collect the temperature, environment, and loading volume inside the tunnel kiln; A control layer: used to collect the data of the multi-modal data acquisition layer and implement frequency conversion control, combustion system adjustment, and dynamic response adjustment of wind speed and air pressure based on a PLC and an industrial control computer; Intelligent AI Decision Analysis Layer: It includes an embedded AI chip and uses a deep reinforcement learning model for intelligent decision-making; Data Interaction Layer: It is docked through HMI or mobile terminal or MES to achieve human-machine interaction and information traceability between the intelligent AI decision analysis layer and the control layer.

[0007] Furthermore, the multi-modal data acquisition layer includes a temperature sensor component, an environmental sensor, and a loading capacity sensor.

[0008] Furthermore, the temperature sensor component includes several infrared thermal imagers and thermocouples installed at different positions in the tunnel kiln; the environmental sensors include several gas composition sensors, humidity sensors, and flow rate sensors installed at different positions in the tunnel kiln; the loading capacity sensor is a weighing sensor installed on the kiln cart during transportation, and the weighing sensor is a resistance strain type weighing sensor or an inductive weighing sensor or a capacitive weighing sensor.

[0009] Even further, the intelligent temperature control method for the tunnel kiln includes: Obtain the real-time data streams of the temperature sensor group, environmental sensors, and loading capacity sensor in the kiln, and fuse them to generate a multi-modal data set including a temperature distribution matrix, environmental parameters, and loading capacity; Extract the temperature distribution matrix from the multi-modal data set, use a convolutional neural network to extract the temperature values at predefined coordinate points, and calculate the deviation vectors between each coordinate point and the target temperature to generate state data; If the maximum deviation value in the deviation vector exceeds the preset threshold T1, use a fuzzy logic algorithm to calculate the power correction coefficient for each heating zone to generate a power distribution instruction; According to the power distribution instruction and the loading capacity change rate, use a rolling horizon prediction algorithm to calculate the heating rate curve and holding time parameters; If the correlation coefficient between the slope change amount of the heating rate curve and the current environmental temperature and humidity data exceeds the threshold T2, use a model predictive control algorithm to update the heating power matrix and ventilation coefficient; According to the updated heating power matrix and ventilation coefficient, use the NSGA-II algorithm to optimize the Pareto front solution set of energy consumption and curing time to generate comprehensive control parameters; Drive the heating element group and the fan array according to the comprehensive control parameters, collect the temperature distribution data after execution, and perform a matching degree analysis with the quality index database; If there is an abnormal temperature area that does not match the quality index, use a deep Q-network algorithm combined with historical control records to update the weights of the convolutional neural network to generate a new control strategy.

[0010] Furthermore, the multi-modal data set includes: Obtain the real-time data streams of the temperature sensor, environment sensor, and loading capacity sensor, and store them in a preset database through the data acquisition module to obtain a sensor data set; If there are missing values in the sensor data set, use the mean imputation method to fill in the missing values to obtain a complete sensor data set; According to the temperature sensor data in the complete sensor data set, use a two-dimensional interpolation algorithm to generate a temperature distribution matrix to obtain temperature distribution matrix data; Extract the environment sensor data from the complete sensor data set, and generate an environment parameter set through standardization processing to obtain environment parameter data; Extract the loading capacity sensor data from the complete sensor data set, and generate a loading capacity data set through smoothing filtering processing to obtain loading capacity data; If the timestamps of the temperature distribution matrix data, environment parameter data, and loading capacity data are consistent, merge the three through a data fusion algorithm to obtain a multi-source data set; According to the multi-source data set, use the principal component analysis algorithm to extract the main features to obtain a multi-source data set with reduced dimensions.

[0011] The beneficial effects of the present invention are as follows: Through multi-source data perception, accurately capture the temperature distribution and environmental changes in the tunnel kiln, thereby reducing the deviation between the actual furnace temperature and the set value, and further ensuring the stability of product quality. 2. Precise temperature control further improves the adaptability of the system to changes in loading capacity and environment. 3. The flexible control strategy can be dynamically adjusted according to the actual production status to avoid problems such as too slow or too fast heating and long holding time. Description of the Drawings

[0012] Figure 1 It is a flowchart of the intelligent temperature control method for the tunnel kiln. Detailed Embodiment

[0013] In this embodiment, the present invention includes an intelligent temperature control system and method for a tunnel kiln. The intelligent temperature control system for the tunnel kiln includes: Multi-modal data acquisition layer: used to collect the temperature, environment, and loading capacity in the tunnel kiln; Control layer: used to collect the data of the multi-modal data acquisition layer, and implement frequency conversion control, combustion system adjustment, and dynamic response adjustment of wind speed and wind pressure based on PLC and industrial control computer; Intelligent AI decision analysis layer: includes an embedded AI chip and uses a deep reinforcement learning model for intelligent decision-making, and historical data can be fed back for training the deep reinforcement learning model to achieve self-learning and continuous optimization; Data interaction layer: dock through HMI or mobile terminal or MES to realize human-computer interaction and information backtracking between the intelligent AI decision analysis layer and the control layer.

[0014] In this embodiment, the multimodal data acquisition layer includes a temperature sensor assembly, an environmental sensor, and a loading amount sensor.

[0015] In this embodiment, the temperature sensor assembly includes a plurality of infrared thermal imagers and thermocouples installed at different positions in the tunnel kiln; the environmental sensors include a plurality of gas composition sensors, humidity sensors, and flow rate sensors installed at different positions in the tunnel kiln; the loading amount sensor is a weighing sensor installed on the kiln car during transportation, and the weighing sensor is a resistance strain type weighing sensor or an inductive weighing sensor or a capacitive weighing sensor; a temperature inductor is also arranged outside the tunnel kiln, which is used to automatically reduce the heating power in the tunnel kiln in high temperature weather to prevent over-temperature explosion and combustion.

[0016] In this embodiment, as Figure 1 shown, the intelligent temperature control method for the tunnel kiln includes the following steps: S101. Obtain the real-time data streams of the temperature sensor group, environmental sensor, and loading amount sensor in the kiln, and fuse them to generate a multimodal data set including a temperature distribution matrix, environmental parameters, and loading amount; S102. Extract the temperature distribution matrix from the multimodal data set, use a convolutional neural network to extract the temperature values at predefined coordinate points, and calculate the deviation vectors between each coordinate point and the target temperature to generate state data; S103. If the maximum deviation value in the deviation vector exceeds the preset threshold T1, use a fuzzy logic algorithm to calculate the power correction coefficient for each heating zone to generate a power distribution instruction. The first step of the fuzzy logic algorithm is to convert the accurate deviation vector data exceeding the preset threshold T1 into a fuzzy set. Specifically, it is to calculate the membership degrees of the input data to each fuzzy set according to the membership functions of the predefined fuzzy sets; S104. According to the power distribution instruction and the loading amount change rate, use a rolling horizon prediction algorithm to calculate the heating rate curve and the holding time parameter. The core idea of the rolling horizon prediction algorithm is to consider the control actions in the future for a period of time and calculate the heating rate curve and the holding time parameter; S105. If the correlation coefficient between the slope change amount of the heating rate curve and the current ambient temperature and humidity data exceeds the threshold T2, use a model predictive control algorithm to update the heating power matrix and the ventilation coefficient. The core idea of the model predictive control algorithm is to use the mathematical model of the system to predict future behaviors and achieve the desired control objectives by optimizing the control input; S106. According to the updated heating power matrix and ventilation coefficient, use the NSGA-II algorithm to optimize the Pareto front solution set of energy consumption and curing time to generate comprehensive control parameters; S107. Drive the heating element group and the fan array according to the comprehensive control parameters, collect the temperature distribution data after execution, and perform a matching degree analysis with the quality index database; S108. If there is an abnormal temperature area that does not match the quality index, use the deep Q-network algorithm combined with historical control records to update the weights of the convolutional neural network to generate a new control strategy.

[0017] In this embodiment, the multi-modal data set includes: Obtain the real-time data streams of the temperature sensor, the environmental sensor, and the loading amount sensor, and store them in a preset database through the data acquisition module to obtain a sensor data set; If there are missing values in the sensor data set, use the mean imputation method to fill in the missing values to obtain a complete sensor data set; According to the temperature sensor data in the complete sensor data set, use a two-dimensional interpolation algorithm to generate a temperature distribution matrix to obtain temperature distribution matrix data; Extract the environmental sensor data from the complete sensor data set, and generate an environmental parameter set through standardization processing to obtain environmental parameter data; Extract the loading amount sensor data from the complete sensor data set, and generate a loading amount data set through smoothing filtering processing to obtain loading amount data; If the timestamps of the temperature distribution matrix data, the environmental parameter data, and the loading amount data are consistent, merge the three through a data fusion algorithm to obtain a multi-source data set; According to the multi-source data set, use the principal component analysis algorithm to extract the main features to obtain a multi-source data set with reduced dimensions.

[0018] The present invention is applied to the technical field of tunnel kilns.

[0019] Although the embodiments of the present invention are described with actual solutions, they do not constitute a limitation on the meaning of the present invention. For those skilled in the art, modifications to its implementation solutions according to this specification and combinations with other solutions are obvious.

Claims

1. An intelligent temperature control system for a tunnel kiln, characterized in that, It includes: Multi-modal data acquisition layer: used to collect the temperature, environment and loading volume inside the tunnel kiln; Control layer: used to collect the data of the multi-modal data acquisition layer, and realize frequency conversion control, combustion system regulation, and dynamic response regulation of wind speed and air pressure based on PLC and industrial control computer; Intelligent AI decision analysis layer: includes an embedded AI chip and uses a deep reinforcement learning model for intelligent decision-making; Data interaction layer: docked through HMI or mobile terminal or MES to realize human-computer interaction and information backtracking between the intelligent AI decision analysis layer and the control layer.

2. The intelligent temperature control system for a tunnel kiln according to claim 1, wherein: The multi-modal data acquisition layer includes a temperature sensor component, an environment sensor, and a loading volume sensor.

3. An intelligent temperature control system for a tunnel kiln according to claim 1, characterized in that: The temperature sensor component includes several infrared thermal imagers and thermocouples installed at different positions inside the tunnel kiln; the environment sensors include several gas composition sensors, humidity sensors, and flow rate sensors installed at different positions inside the tunnel kiln; the loading volume sensor is a weighing sensor installed on the transported kiln car, and the weighing sensor is a resistance strain type weighing sensor or an inductive weighing sensor or a capacitive weighing sensor.

4. A temperature control method applied to the tunnel kiln intelligent temperature control system according to any one of claims 2-3, characterized in that, It includes: Obtain the real-time data streams of the temperature sensor group, environment sensor and loading volume sensor inside the kiln, and fuse them to generate a multi-modal data set containing a temperature distribution matrix, environmental parameters and loading volume; Extract the temperature distribution matrix from the multi-modal data set, use a convolutional neural network to extract the temperature values of predefined coordinate points, and calculate the deviation vector between each coordinate point and the target temperature to generate state data; If the maximum deviation value in the deviation vector exceeds the preset threshold T1, use the fuzzy logic algorithm to calculate the power correction coefficient of each heating zone to generate a power distribution instruction; According to the power distribution instruction and the loading volume change rate, use the rolling horizon prediction algorithm to calculate the heating rate curve and the holding time parameter; If the correlation coefficient between the slope change amount of the heating rate curve and the current environmental temperature and humidity data exceeds the threshold T2, use the model predictive control algorithm to update the heating power matrix and the ventilation coefficient; According to the updated heating power matrix and ventilation coefficient, use the NSGA-II algorithm to optimize the Pareto front solution set of energy consumption and curing time to generate comprehensive control parameters; Drive the heating element group and the fan array according to the comprehensive control parameters, collect the temperature distribution data after execution, and perform a matching degree analysis with the quality index database; If there is an abnormal temperature area that does not match the quality index, use the deep Q network algorithm combined with historical control records to update the weights of the convolutional neural network to generate a new control strategy.

5. The intelligent temperature control method for a tunnel kiln according to claim 4, wherein: The multi-modal data set includes: Obtain the real-time data streams of the temperature sensor, environment sensor and loading volume sensor, store them in a preset database through a data acquisition module to obtain a sensor data set; If there are missing values in the sensor data set, use the mean imputation method to fill the missing values to obtain a complete sensor data set; According to the temperature sensor data in the complete sensor data set, use a two-dimensional interpolation algorithm to generate a temperature distribution matrix to obtain temperature distribution matrix data; Extract environmental sensor data from the complete sensor data set, generate a set of environmental parameters through normalization processing, and obtain environmental parameter data; Extract load sensor data from the complete sensor data set, generate a set of load data through smoothing filtering processing, and obtain load data; If the timestamps of the temperature distribution matrix data, environmental parameter data, and load data are consistent, merge the three through a data fusion algorithm to obtain a multi-source data set; According to the multi-source data set, use the principal component analysis algorithm to extract the main features and obtain the multi-source data set after dimensionality reduction.

Images