Wide vermicelli processing remote control system based on Internet of Things

Abstract

The invention relates to the technical field of intelligent control of a starch product processing process, in particular to a wide vermicelli processing remote control system based on Internet of Things. Comprising a data acquisition module used for acquiring initial biochemical characteristic parameters of raw materials and monitoring dynamic process parameters in real time; the first processing module is used for determining a starch state vector through a preset hybrid physical information neural network model; the second processing module is used for calculating a coordination index according to the starch state vector and a preset state coordination limit value vector; the third processing module is used for performing prospective prediction through a hybrid physical information neural network model and determining a trend vector; and the parameter correction module is used for determining a corrected process control parameter vector through a multivariable state feedback control law. The system solves the problem that the traditional process depends on indirect parameters and artificial experience to carry out fuzzy judgment, the scientificity and transparency of process monitoring are improved, and a solid mechanism model foundation is laid for realizing fine control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of starch product processing, in particular to a wide powder processing remote control system based on the Internet of Things. BACKGROUND

[0002] Wide powder processing is a complex biochemical process, the core of which is to accurately control the state evolution of starch under the action of water and heat, which directly determines the taste, texture and quality of the final product. In the existing wide powder processing technology, a key challenge is that the internal core state of starch, such as gelatinization degree, molecular chain structure change and water activity, cannot be quantified in real time and accurately; the process control of the production process therefore relies heavily on the experience of operators based on indirect parameters such as temperature and torque; the disadvantage of this control method is that when the batch characteristics of raw materials fluctuate or the working conditions are disturbed, the control adjustment is often delayed and non-optimal, making it difficult to ensure the high consistency of product quality; in addition, the lack of forward-looking prediction ability of state evolution trend also makes the system unable to avoid process instability risks in advance, thereby limiting the further improvement of production efficiency. Therefore, there is an urgent need for a solution that can accurately evaluate, dynamically predict and automatically correct the state of starch to solve the problems of poor quality consistency, low production efficiency and strong dependence on human experience in the current wide powder processing technology.

[0003] The above information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0004] To solve the above technical problems, the present application discloses a wide powder processing remote control system based on the Internet of Things, in particular, the technical solution of the present application comprises: A data acquisition module for obtaining initial biochemical characteristic parameters of raw materials and monitoring dynamic process parameters in real time; A first processing module for determining a starch state vector based on the initial biochemical characteristic parameters and the dynamic process parameters through a preset hybrid physical information neural network model; A second processing module for calculating a coordination index according to the starch state vector and a preset state coordination limit vector; A third processing module for determining a trend vector through a hybrid physical information neural network model based on the starch state vector for forward-looking prediction; A parameter correction module for determining a corrected process control parameter vector through a multivariable state feedback control law according to the trend vector and the current process control parameter vector.

[0005] Preferably, the dynamic process parameters include motor torque, temperature field and ultrasonic wave speed.

[0006] Preferably, the first processing module determines a starch state vector, comprising: determining a gelatinization degree factor, a molecular chain orientation index and a free water content factor; based on the gelatinization degree factor, the molecular chain orientation index and the free water content factor, the starch state vector is constituted.

[0007] Preferably, the second processing module calculates a coordination index, comprising: determining the coordination index by calculating the ratio of the Euclidean norm of the starch state vector to the norm of the state coordination limit vector.

[0008] Preferably, further comprising: an evaluation module for comparing the coordination index with a preset coordination index threshold value, and dividing the current process state into a qualified level, a first warning level or a second warning level.

[0009] Preferably, the third processing module determines a trend vector, comprising: taking the starch state vector at the current time as input, a predicted state vector at a future time step is predicted; based on the starch state vector at the current time and the predicted state vector, a gelatinization trend factor is calculated; based on the starch state vector at the current time and the predicted state vector, a thickening trend index is calculated; based on the starch state vector at the current time and the predicted state vector, a dehydration trend index is calculated; the gelatinization trend factor, the thickening trend index and the dehydration trend index are constituted into a trend vector.

[0010] Preferably, the parameter correction module determines a corrected process control parameter vector, comprising: calling a control gain matrix representing the influence gain of each state evolution trend on each control variable; based on the current process control parameter vector, subtracting the product of the control gain matrix and the trend vector to determine the corrected process control parameter vector.

[0011] Preferably, further configured to execute a hierarchical correction strategy, comprising: taking the corrected process control parameter vector as input, performing forward propagation through the hybrid physical information neural network model to obtain a predicted coordination index; when the predicted coordination index is in the first warning interval, the system automatically applies the corrected process control parameter vector for parameter fine-tuning; When the prediction coordination index is in the secondary warning interval, the system sends a warning to the remote operator to request manual intervention based on the application of the revised process control parameter vector; When the prediction coordination index is in the qualified interval, the system maintains the current process control parameter vector.

[0012] Compared with the prior art, the present application has the following beneficial effects: 1. The system innovatively quantifies the difficult-to-directly-measure starch state into a multi-dimensional state vector composed of paste degree factor, molecular chain orientation index and free water content factor by constructing a hybrid physical information neural network model. This method first realizes precise and interpretable online characterization of the core process state of wide flour processing, solves the problem of traditional process relying on indirect parameters and manual experience for fuzzy judgment, greatly improves the scientificity and transparency of process monitoring, and lays a solid mechanism model foundation for realizing fine control.

[0013] 2. The system uses the forward-looking prediction ability of the neural network model to generate a trend vector containing multiple dimensions such as pasting, thickening and dewatering by calculating the difference between the current state and the future short-time-step predicted state. This design enables the control system to have foresight, changing from passive response to existing state deviation to active feedforward intervention according to the upcoming change trend. This control strategy based on change rate can inhibit disturbances earlier, making the control process more stable and responsive, and significantly improving the dynamic performance of the system in response to process changes.

[0014] 3. The system uses a parameter revision module based on a multivariate state feedback control law, which intelligently and decoupledly maps the deviation trend of the multi-dimensional state to the precise adjustment amount of multiple process control parameters such as screw speed and steam valve opening through a pre-set control gain matrix. This method avoids the system shock that may be caused by traditional single-variable control, realizes stable and closed-loop automatic revision of complex coupled systems, ensures the accuracy and robustness of control, significantly reduces the dependence on manual operation skills, and improves the automation level of the entire production process.

[0015] 4. The system not only calculates the deviation degree of the current state from the ideal state, but also innovatively introduces a pre-judgment mechanism before control. After calculating the revised parameters, the system first predicts the future coordination index by applying the revised parameters, and executes a graded revision strategy according to the risk level of the prediction result. This mechanism avoids ineffective or excessive control actions, realizes the intelligent combination of automatic fine-tuning and manual intervention, ensures efficient and automated control, and also provides safety redundancy for high-risk scenarios, comprehensively improving the intelligence and safety of the system. BRIEF DESCRIPTION OF DRAWINGS

[0016] The application will be further explained in connection with the accompanying drawings and embodiments: Figure 1 is a system structure diagram of the application. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the application more clear and explicit, the application will be further described in detail below in connection with specific embodiments.

[0018] Embodiment 1: Please refer to Figure 1 A wide flour processing remote control system based on the Internet of Things, comprising: A data acquisition module is configured to obtain initial biochemical characteristic parameters of raw materials and to monitor dynamic process parameters in real time. A first processing module is configured to determine a starch state vector based on the initial biochemical characteristic parameters and the dynamic process parameters by using a preset hybrid physical information neural network model. A second processing module is configured to calculate a coordination index based on the starch state vector and a preset state coordination limit vector. A third processing module is configured to determine a trend vector by using the hybrid physical information neural network model based on the starch state vector for prospective prediction. A parameter correction module is configured to determine a corrected process control parameter vector by using a multivariable state feedback control law based on the trend vector and a current process control parameter vector. The embodiment provides a wide flour processing remote control system based on the Internet of Things, which aims to solve the technical problems of low production efficiency, poor quality consistency and process control relying on manual experience due to the difficulty in real-time and accurate quantification of starch state in the existing wide flour processing process. By constructing a digital twin model deeply integrated with data driving and physical mechanism, the system realizes accurate evaluation, prospective prediction and automatic closed-loop correction control of the processing process. In a specific implementation scenario, the system is deployed in an automatic wide flour production line. The overall architecture of the system includes the following modules: A data acquisition module provides comprehensive and real-time input data for subsequent modeling, evaluation and control. This module is the basis of the perception layer of the entire system, ensuring the accuracy and real-time performance of the data source for model operation. In this embodiment, the data acquisition module is implemented by a multi-source heterogeneous sensor network. At the starting end of the production process, the initial biochemical characteristic parameters of the raw material batch are collected non-invasively by a hyperspectral imager, such as the initial content and distribution of starch, protein and moisture. These parameters serve as static initial conditions for the subsequent physical information fusion model, providing prior knowledge of batch differences in raw materials; A distributed sensor array is deployed within and along the processing equipment, such as an extruder, to monitor dynamic process parameters in real time; This array specifically includes: a motor torque sensor for measuring screw drive force, a distributed fiber optic temperature sensor for obtaining the temperature distribution of the material along the device axis, and an ultrasonic sensor for monitoring internal structural changes of the material; The first processing module is responsible for converting the physically meaningful and dispersed sensor data obtained by the data acquisition module into a unified index that can represent the core biochemical state of the starch; In this embodiment, the core of this module is a pre-trained hybrid physical information neural network model H-PINN; Hybrid physical information neural network model refers to a deep learning model whose training process is not only driven by real data but also strongly constrained by known physical laws; Its role is to accurately infer the complete internal state of the system even when industrial data is sparse or some states cannot be directly measured; The module receives the initial biochemical characteristic parameters provided by the data acquisition module as the initial boundary conditions of the model, and continuously receives real-time dynamic process parameters such as torque and temperature field as dynamic inputs; Through the forward inference calculation of the hybrid physical information neural network model, the module can accurately estimate a multi-dimensional vector that cannot be directly measured by any single sensor but can fully represent the core state of the starch slurry, i.e., the starch state vector, denoted as ; The second processing module is responsible for reducing the multi-dimensional starch state vector output by the first processing module to generate a single, intuitive scalar evaluation index for real-time judgment of the goodness of the current process state; In this embodiment, the operational logic of this module is based on vector space analysis; The state coordination limit vector refers to a pre-set reference vector representing the ideal starch state when the product quality is optimal, denoted as ; Its role is to provide a benchmark for the actual production state; Based on a large amount of historical optimal process batch production data, it is pre-calibrated through statistical analysis or expert experience; This module determines a dimensionless coordination index by calculating the norm ratio between the current starch state vector and the pre-set state coordination limit vector ; This index intuitively quantifies the deviation of the current process state from the ideal state; A third processing module, whose core function is to utilize the predictive ability of the model to realize the forward-looking judgment of the process state evolution trend, thereby providing the decision basis for the feedforward control; in the embodiment, the module utilizes a hybrid physical information neural network model to realize the forward-looking prediction; specifically, it takes the current time's starch state vector and the current process parameters as inputs, drives the model to perform a virtual evolution for a short time step, for example, 1 second, thereby predicting the state vector at the future time; By comparing the difference between the current state and the predicted state, the module can calculate a series of rate indicators representing the dynamic evolution trend of the system, and combine these indicators into a trend vector ; the trend vector refers to a vector composed of the change rates of multiple state components; its role is not only to indicate the direction of system state deviation, but also to quantify the urgency of deviation, thereby providing the control system with an advance amount; the time difference calculation is performed on the short-term prediction result of the hybrid physical information neural network model; A parameter correction module, which is configured to automatically calculate the optimal process parameter adjustment scheme according to the deviation trend predicted by the third processing module, thereby forming a closed-loop control; in the embodiment, the module adopts a multivariable state feedback control law to realize this function; the multivariable state feedback control law refers to a control algorithm derived from modern control theory, which directly maps the state change trend of the system to the adjustment amount of multiple control actuators through a control gain matrix, thereby realizing the decoupling and precise control of the multiple-input multiple-output (MIMO) system; its role is to avoid the system oscillation caused by traditional single-variable control, and to realize smooth and precise parameter correction; The control model is optimized and calculated based on system identification or a large amount of historical production data; the module receives the trend vector output by the third processing module , and combines the current process control parameter vector , for example, the set values of the steam valve opening degree and the screw rotation speed, to calculate a corrected process control parameter vector ; the is then sent to the programmable logic controller (PLC) of the production line for execution, thereby completing the automatic and feedforward correction of the process; The above modules work cooperatively to construct a complete intelligent control system from data acquisition, state understanding, dynamic evaluation, trend prediction to closed-loop correction; it first concretizes and computable expresses the abstract starch state through the starch state vector, and solves the problem of unmeasurable key state by utilizing the hybrid physical information neural network model; It should be particularly pointed out that, the attached​​​​​Figure 1 The flow in the figure shows the main data flow path of the system, and the final closed-loop control decision is made by the parameter correction module, which internally integrates a forward-looking evaluation link; this link judges the predicted coordination index after the execution of the correction parameter according to the state evaluation logic established by the second processing module and the evaluation module, so as to realize the hierarchical correction strategy; this constitutes a complete intelligent decision-making closed loop from state evaluation to prediction correction, ensuring the intelligence and safety of control; Compared with the traditional control mode relying on artificial experience, the system can improve the automation level and intelligent degree of the wide flour processing process, ensure the high consistency of product quality, reduce the dependence on the skills of operators, and effectively avoid process instability through forward-looking prediction and correction, thereby improving the production line efficiency and raw material utilization rate.

[0019] Embodiment 2: The dynamic process parameters include motor torque, temperature field, and ultrasonic speed. In this embodiment, the dynamic process parameters are defined as at least including: motor torque, temperature field, and ultrasonic speed, based on embodiment 1. The motor torque refers to the torque required to drive the screw of the extrusion equipment to rotate; it indirectly reflects the macroscopic viscosity and flow resistance of the starch slurry inside the equipment, and is a key external representation of rheological properties; it is obtained by real-time measurement through the torque sensor installed on the main drive motor; The temperature field refers to the temperature data sequence of the material distributed along the axial direction inside the equipment, denoted as ; its function is to accurately describe the energy input and heat transfer in the gelatinization process, which is the core parameter for controlling the kinetics of gelatinization reaction; it is obtained through the distributed fiber temperature sensor array laid along the inner wall of the equipment; The ultrasonic speed refers to the speed of ultrasonic wave propagation in the starch slurry, denoted as ; it sensitively reflects the changes in the microstructure of the material, such as particle dissolution, molecular chain rupture and recombination, providing information in the acoustic dimension for judging the degree of gelatinization; it is measured in real time through the ultrasonic transmitting-receiving transducer pair installed on the equipment; By taking the three parameters of motor torque, temperature field, and ultrasonic speed, which are different in physical dimension but internally related, as dynamic process parameters, this embodiment can provide more rich and comprehensive input information for the mixed physical information neural network model of the first processing module; the fusion of such multi-modal data enables the model to better understand the complex evolution process of the starch state, thereby improving the accuracy and robustness of the starch state vector estimation, and laying a more solid data foundation for the accurate evaluation and control of the entire system.

[0020] Embodiment 3: The first processing module determines the starch state vector, including: determining a gelatinization degree factor, a molecular chain orientation index, and a free water content factor; constructing a starch state vector based on the gelatinization degree factor, the molecular chain orientation index, and the free water content factor; The embodiment is based on embodiment 1, and the first processing module determines the specific description of the starch state vector process; the starch state vector is innovatively defined as a three-dimensional vector , each component of which is a dimensionless normalized index; The first processing module determines the following three core state components through the internal node output of the hybrid physical information neural network model: a gelatinization degree factor : refers to the degree of irreversible transition of starch particles from an ordered crystal structure to an amorphous paste after being heated and absorbing water; its role is to quantify the progress of the gelatinization reaction, which is the most core index for evaluating the taste and digestion characteristics of the product; a normalized value between 0 and 1 is calculated by the H-PINN model according to the input temperature field, torque, etc., and combined with the internal constraint Arrhenius-type kinetic equation; a molecular chain orientation index : refers to the degree of order of the arrangement of long-chain starch molecules along the flow direction under the action of extrusion shear flow; reflects the micro-anisotropy of the material, which is directly related to the mechanical strength, toughness and surface smoothness of the final product; a normalized index is calculated by the H-PINN model according to the input motor torque and the material velocity field solved by the internal fluid mechanics equation of the model; a free water content factor : refers to the proportion of free water in the starch slurry that is not bound by hydrogen bonds and other actions of starch molecules; represents the activity of water molecules in the system, which is crucial for controlling the reaction rate, preventing wall sticking, and affecting the final rehydration and shelf life of the product; calculated by the H-PINN model according to the input ultrasonic velocity, which is strongly related to the free water content of the medium, and the material balance equation; The module combines the above calculated gelatinization degree factor , molecular chain orientation index , and free water content factor in a predetermined order to construct a starch state vector ; By decomposing the abstract starch state into three orthogonal dimensions of gelatinization degree, molecular chain orientation and free water content, the embodiment enhances the scientificity and interpretability of state description; compared with a single index such as viscosity, the multi-dimensional state vector expression can more comprehensively and profoundly reveal the internal quality evolution of the material in the process, so that the subsequent evaluation and control can be based on a deeper understanding of the mechanism, thereby realizing unprecedented fine regulation and control.

[0021] Embodiment 4: The second processing module calculates the coordination index, comprising: The coordination index is determined by calculating the ratio of the Euclidean norm of the starch state vector to the norm of the state coordination limit vector; Further comprising: The evaluation module is configured to compare the coordination index with a preset coordination index threshold value, and divide the current process state into a qualified level, a first warning level or a second warning level; Based on embodiment 1, the second processing module calculates the coordination index and further develops the application thereof; the embodiment further comprises an evaluation module; The second processing module calculates the coordination index The specific steps are as follows: In order to reduce the multi-dimensional starch state vector output by the first processing module to a single, intuitive scalar evaluation index, the coordination index is introduced; the real-time calculation formula is: ; Wherein, : the coordination index, which is a dimensionless scalar, is calculated by the module in real time; : the Euclidean norm of the current state vector , which is calculated as , the value of which reflects the distance of the current state point from the origin in the state space, and comprehensively represents the overall strength of the current state, and the component parameters are calculated by the first processing module; : the norm of the preset limit vector , which is a constant value representing the strength of the ideal state; The design motivation of the formula comes from the vector deviation degree measurement method in vector space analysis; by comparing the norm of the current state vector with that of the ideal reference vector, the overall coordination of the current process can be intuitively and robustly quantified; The evaluation module is configured to compare the coordination index calculated by the second processing module Assigning explicit process state levels to provide direct trigger basis for subsequent hierarchical control strategies; in this embodiment, the evaluation module will coordinate index is compared with the preset coordination index threshold in real time, and the current process state is divided into the following three levels: Qualified level: when , it means that the length of the current state vector does not exceed the ideal state, the process runs inside the ideal state space, and the state is stable; First warning level: when , it means that the process state has deviated from the ideal region, but has not reached a dangerous level, and the system should start automatic fine tuning; Second warning level: when , it means that the process state has deviated seriously, and there is a higher risk of causing product quality unqualified or equipment running instability, so the system needs to take more powerful intervention measures; The determination method of the above threshold values 1.0 and 1.5 is: through statistical analysis of a large amount of historical production data, combined with the final product quality inspection results, the critical values are calibrated which can effectively distinguish the evolution interval of process state from normal to risk; By introducing the calculation and evaluation module of coordination index and hierarchical level, this embodiment successfully converts the complex multi-dimensional state vector into a process state signal that is easy for operators to understand and intuitive; This design not only realizes real-time visual monitoring of production state, provides a clear and reliable logical judgment basis for subsequent implementation of automatic hierarchical correction control strategies with different response intensities, making the control system more intelligent and safe.

[0022] Embodiment 5: The third processing module determines the trend vector, comprising: Taking the current starch state vector as input, the predicted state vector after the future time step is predicted; Based on the current starch state vector and the predicted state vector, a gelatinization trend factor is calculated; Based on the current starch state vector and the predicted state vector, a thickening trend index is calculated; Based on the current starch state vector and the predicted state vector, a dehydration trend index is calculated; The gelatinization trend factor, thickening trend index and dehydration trend index are combined to form a trend vector; This embodiment is based on embodiment 1, and the third processing module determines the trend vector process; the trend vector is defined as , each component of which represents the change rate of the corresponding component in the starch state vector ; The steps of determining the trend vector by the third processing module are as follows: The module takes the current time 's starch state vector as input, makes a quick forward-looking prediction through the trained hybrid physical information neural network model, and obtains the predicted state vector after one short time step in the future. Based on the current state and the predicted state, the module approximates the first derivative of the state variable with respect to time by time difference, thereby calculating each trend component: Calculate the gelatinization trend factor : ; Calculate the thickening trend index : ; Calculate the dehydration trend index : ; where, : the trend index of each state component, dimensionless, calculated by this module in real time; The subscripts and represent the values at the current time and the predicted time, respectively, both of which are calculated by the hybrid physical information neural network model. The module combines the calculated gelatinization trend factor , thickening trend index and dehydration trend index to form a trend vector . This embodiment introduces the prediction of future states and the calculation of trend vectors, enabling the control system to have foresight; instead of passively responding to the state deviation that has already occurred, it can actively intervene according to the changing trend that is about to occur; this feedforward control idea based on the rate of change can suppress disturbances earlier than feedback control based only on state deviation, making the control process more stable and the response more rapid, and improving the dynamic performance of the system in response to raw material fluctuations and changes in working conditions. Embodiment 6:

[0023] The parameter correction module determines the corrected process control parameter vector, including: Calling the control gain matrix representing the influence gain of each state evolution trend on each control variable; Based on the current process control parameter vector, subtract the product of the control gain matrix and the trend vector to determine the corrected process control parameter vector; ​Also used to execute the hierarchical correction strategy, comprising: The corrected process control parameter vector is input, forward propagation is performed through the hybrid physical information neural network model, and a predicted coordination index is obtained; When the predicted coordination index is in the first early warning interval, the system automatically applies the corrected process control parameter vector for parameter fine tuning; When the predicted coordination index is in the second early warning interval, the system applies the corrected process control parameter vector and sends a warning to the remote operator to request manual intervention; When the predicted coordination index is in the qualified interval, the system maintains the current process control parameter vector; This embodiment is based on embodiment 1, the parameter correction module determines the corrected process control parameter vector, and the integrated description of the subsequent hierarchical correction strategy is executed; The parameter correction module determines the corrected process control parameter vector The process is as follows: The core of this module is a multivariate state feedback control law, which is mathematically expressed as: ; Wherein, : The corrected process control parameter vector, which has the same dimension as , is calculated by this module and delivered to the actuator; : The current process control parameter vector, for example, a vector containing the steam valve opening degree in percentage and the screw speed in rpm, is read in real time from the PLC or control system; : Trend vector, a column vector provided by the third processing module; : Control gain matrix, a matrix, is the dimension of the control vector , which is calculated offline by system identification method based on a large amount of historical production data, and can also be adjusted online adaptively; The control gain matrix is a core parameter matrix, each element of which represents the ideal influence gain of the th state trend such as thickening trend index on the th control variable such as screw speed; its function is to decouple each independent deviation trend and accurately distribute it to the most suitable control actuator for correction; through dimensional analysis, the dimension of element in matrix is the dimension of the th state trend such as thickening trend index , and the dimension of the th control variable such as screw speed is ; a unit of control quantity multiplied by time , thus ensuring that the product of the dimension of the control gain matrix ; Based on the current process control parameter vector , subtract the product of the control gain matrix and the trend vector to determine the corrected process control parameter vector ; After calculating , the system further implements a hierarchical correction strategy: The corrected process control parameter vector is input again to pass through the hybrid physical information neural network model for a fast forward propagation to obtain a predicted coordination index ; this index indicates how the future process state will be if is immediately executed; According to the predicted coordination index , a hierarchical judgment and operation are performed: When the predicted coordination index is in the qualified interval, i.e. , it indicates that the current control parameters do not need to be adjusted or that a small disturbance can be self-absorbed by the system, so the system maintains the current process control parameter vector without correction, avoiding unnecessary control actions; When the predicted coordination index is in the first warning interval, i.e. , it indicates that a slight deviation state is predicted; at this time, the system automatically applies the corrected process control parameter vector for parameter fine-tuning to kill the deviation trend in the embryonic state; When the predicted coordination index is in the second warning interval, i.e. , it indicates that a serious deviation is predicted; at this time, the system applies the corrected process control parameter vector and simultaneously sends a high-risk warning to the remote operator to request manual confirmation or intervention; The control logic of the embodiment has double advancement; through the control gain matrix The product of the two, realizes the intelligent decoupling control of the multivariable system, ensures the accuracy and stability of the control; introduces a hierarchical strategy based on the modified effect prediction, which is an intelligent decision-making mechanism before control; it avoids invalid or excessive control operations, and can take different measures according to the predicted risk level, while ensuring the efficiency of automatic control, also introduces manual supervision as a safety redundancy in high-risk situations, improves the intelligent level, safety and comprehensive performance of the whole control system; To further ensure the stability and safety of the system, the parameter correction module also includes output saturation limitation and input data validity discrimination functions; after calculating the modified process control parameter vector , the system will check and limit each component in according to the preset safe operation range of each actuator, such as screw speed , steam valve opening , to prevent the output from exceeding the physically feasible domain control instruction; at the same time, the data acquisition module is built-in with data validity discrimination algorithm, which can identify and handle abnormal situations such as sensor data mutation, loss or exceeding normal range, so as to guarantee the data quality of the input to the H-PINN model, and improve the robustness of the whole control system.

[0024] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. An Internet of Things-based wide powder processing remote control system, characterized by, The method comprises the following steps: a data acquisition module is used to acquire initial biochemical characteristic parameters of raw materials and to monitor dynamic process parameters in real time; a first processing module is used to determine a starch state vector based on the initial biochemical characteristic parameters and the dynamic process parameters through a preset hybrid physical information neural network model; a second processing module is used to calculate a coordination index according to the starch state vector and a preset state coordination limit vector; a third processing module is used to determine a trend vector through the hybrid physical information neural network model based on the starch state vector for prospective prediction; a parameter correction module is used to determine a corrected process control parameter vector through a multivariable state feedback control law according to the trend vector and a current process control parameter vector.

2. The wide powder processing remote control system based on the Internet of Things according to claim 1, characterized in that, The dynamic process parameters include motor torque, temperature field and ultrasonic speed.

3. The wide powder processing remote control system based on the Internet of Things according to claim 1, characterized in that, The first processing module determines the starch state vector, which comprises the following steps: a gelatinization degree factor, a molecular chain orientation index and a free water content factor are determined; the starch state vector is constituted based on the gelatinization degree factor, the molecular chain orientation index and the free water content factor.

4. The wide powder processing remote control system based on the Internet of Things according to claim 1, characterized in that, The second processing module calculates the coordination index, which comprises the following steps: the coordination index is determined by calculating the ratio of the Euclidean norm of the starch state vector to the norm of the state coordination limit vector.

5. The wide powder processing remote control system based on the Internet of Things according to claim 4, characterized in that, The method further comprises the following steps: an evaluation module is used to compare the coordination index with a preset coordination index threshold value, and divide the current process state into a qualified level, a first warning level or a second warning level.

6. The wide powder processing remote control system based on the Internet of Things according to claim 1, characterized in that, The third processing module determines the trend vector, which comprises the following steps: a predicted state vector after a future time step is predicted by taking the starch state vector at the current time as input; a gelatinization trend factor is calculated based on the starch state vector at the current time and the predicted state vector; a thickening trend index is calculated based on the starch state vector at the current time and the predicted state vector; a dehydration trend index is calculated based on the starch state vector at the current time and the predicted state vector; the gelatinization trend factor, the thickening trend index and the dehydration trend index are constituted into the trend vector.

7. The wide powder processing remote control system based on the Internet of Things according to claim 1, characterized in that, The parameter correction module determines the corrected process control parameter vector, which comprises the following steps: a control gain matrix representing the influence gain of each state evolution trend on each control variable is called; the corrected process control parameter vector is determined by subtracting the product of the control gain matrix and the trend vector from the current process control parameter vector. 8.The wide powder processing remote control system based on the Internet of Things according to claim 1 or 7, characterized in that, The method is further used to execute a hierarchical correction strategy, which comprises the following steps: the corrected process control parameter vector is taken as input to perform forward propagation through the hybrid physical information neural network model to obtain a predicted coordination index; when the predicted coordination index is in the first warning interval, the system automatically applies the corrected process control parameter vector for parameter fine-tuning; when the predicted coordination index is in the second warning interval, the system applies the corrected process control parameter vector and sends a warning to a remote operator to request manual intervention; when the predicted coordination index is in the qualified interval, the system maintains the current process control parameter vector.

Images