CN-121988216-A - Online batching system for powder materials and control method thereof

CN121988216ACN 121988216 ACN121988216 ACN 121988216ACN-121988216-A

Abstract

The invention discloses an online batching system for powder materials and a control method thereof, belonging to the technical field of industrial automation and intelligent manufacturing. The system comprises a material unit, a feeding unit, a weighing unit, a conveying unit and a control unit, wherein the method comprises the steps of collecting sensor data of a batching system, preprocessing the sensor data, and predicting optimal parameters of a controller of the feeding unit based on the preprocessed data. The invention effectively filters weighing signal noise through CEEMDAN-improved wavelet threshold algorithm, provides reliable data basis for control, adopts the TCN-LSTM-Attention network optimized by DE-GWO to predict the optimal PID parameter in real time, realizes the self-adaptive adjustment of the controller parameter, improves the batching precision, shortens the adjusting time, obviously improves the batching speed, precision and stability, reduces the dependence on the experience of operators, and realizes the intelligent closed-loop control in real sense.

Inventors

SUN CHAOBO
XU JIANGANG

Assignees

北京马赫天诚科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260302

Claims (8)

1. An on-line dosing system for powder material, comprising: the material unit is used for storing powder materials; The feeding unit is connected with the material unit and is used for conveying powder materials; the weighing unit is arranged below the discharging end of the feeding unit and is used for measuring the weight of the received materials in real time; The conveying unit is arranged below the weighing unit and is used for conveying the proportioned materials to the next working procedure; the control unit, respectively with the feed unit with weighing unit electric connection, the control unit includes: The data acquisition module is used for acquiring weight signals of the weighing unit, rotating speed signals of the feeder and environmental signals in real time; the signal processing module is used for carrying out noise reduction pretreatment on the weight signals of the weighing unit and calculating real-time multidimensional feature vectors based on the pretreated data; and the deep learning intelligent decision module is used for outputting optimal control parameters of a controller of the feeding unit in the next control period according to the real-time multidimensional feature vector.
2. The system of claim 1, wherein the control unit is integrated in a control cabinet, the hardware of which comprises: the PLC is used as a hardware carrier of the data acquisition module and the signal processing module; the embedded industrial computer is used as a hardware carrier of the deep learning intelligent decision module; the human-computer interface is connected with the PLC through an Ethernet or a communication line; wherein, PLC and embedded industrial computer pass through industrial Ethernet communication connection.
3. A control method for an on-line batching system for powder materials, applied to the on-line batching system for powder materials according to any one of claims 1-2, characterized by comprising: Step 1, collecting sensor data of a batching system; step 2, preprocessing sensor data; and 3, carrying out optimal parameter prediction on a controller of the feeding unit based on the preprocessed data.
4. A control method according to claim 3, characterized in that the sensor data of the dosing system comprise a weight signal of the weighing cell, a feeder rotational speed signal and an environmental signal.
5. The control method according to claim 4, wherein in step 2, the sensor data is preprocessed, specifically: Processing the weight signal of the weighing unit based on CEEMDAN-an improved wavelet threshold noise reduction algorithm; And carrying out real-time feature vector calculation based on the feeder rotating speed signal and the processed weight signal, and constructing a real-time multidimensional feature vector based on a calculation result and an environment signal.
6. The control method according to claim 5, characterized in that the weight signal of the weighing cell is processed based on CEEMDAN-modified wavelet threshold noise reduction algorithm, in particular: Decomposing a weight signal of the weighing unit into a plurality of IMF components by adopting an adaptive noise complete set empirical mode decomposition algorithm; Calculating the pearson correlation coefficient of each IMF component and the weight signal of the weighing unit, and dividing the IMF component into a noise dominant component, a mixed component and a useful signal dominant component according to a preset correlation threshold; Noise reduction is carried out on the mixed components by adopting an improved wavelet threshold function; And linearly reconstructing the noise-reduced mixed component, the dominant component of the useful signal and the residual component to obtain a processed weight signal.
7. The control method of claim 6, wherein the real-time multidimensional feature vector includes a weight deviation, a rate of change of the weight deviation, a rate of change of a feeder rotational speed, a moving average of the processed weight signal, and an environmental signal.
8. The control method according to claim 7, wherein in step 3, the optimal parameter prediction is performed on the controller of the feeding unit based on the preprocessed data, specifically: constructing an optimal parameter prediction model based on a DE-GWO-TCN-LSTM-Attention network structure; training an optimal parameter prediction model based on a preset data set; And inputting the real-time multidimensional feature vector into an optimal parameter prediction model to obtain the optimal parameters of the controller of the feeding unit.

Description

Online batching system for powder materials and control method thereof Technical Field The invention relates to the technical field of industrial automation and intelligent manufacturing, in particular to an online batching system for powder materials and a control method thereof. Background In the industrial fields of chemical industry, food, pharmacy and the like, the accurate batching of powder materials is one of the core processes for guaranteeing the quality of the final products. At present, an online batching system commonly adopted in the industry is mostly based on a traditional PID control algorithm, and the rotating speed of a feeder (such as a screw feeder) is regulated through the deviation between a weight signal fed back by a weighing sensor and a set value. However, the powder material itself has complex characteristics of poor flowability, easy moisture absorption, easy generation of 'bridging' or 'jet', and the like, which results in obvious nonlinearity, large hysteresis and random interference in the conveying and weighing processes. The traditional PID controller is difficult to dynamically adapt to the complex working conditions due to fixed parameters, so that the problems of low batching precision, long adjusting time, easy overshoot and the like generally exist. Although some researches introduce fuzzy control or self-adaptive PID, the adjusting capability is limited, and expert experience is seriously relied on, so that deep dynamic characteristics of the whole batching process cannot be fundamentally learned and modeled. In addition, the weighing signal is extremely easy to be polluted by complex noise such as mechanical vibration and electromagnetic interference in the acquisition process, and the traditional filtering method (such as low-pass filtering and moving average) has limited effect on the non-stationary noise processing overlapped with the useful signal frequency band, so that noise is filtered, the abrupt change point of the real weight can be blurred, hysteresis is introduced, and the improvement of control precision is further restricted. In recent years, the deep learning technology has great potential in time sequence prediction and nonlinear system control, but how to integrate the deep learning technology into an industrial batching control loop with high real-time requirements effectively, overcome noise interference and realize prospective and self-adaptive precise control is still a technical problem to be solved currently. In view of the above, there is a need for an online batching system for powder materials and a control method thereof, which solve the above problems of the conventional method and realize online batching of powder materials. Disclosure of Invention The invention aims to provide an on-line batching system for powder materials and a control method thereof, which realize the self-adaptive setting of controller parameters, improve batching precision, shorten adjusting time, remarkably improve batching speed, precision and stability, reduce dependence on experience of operators and realize intelligent closed-loop control in a true sense. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: An on-line batching system for powder materials, comprising: the material unit is used for storing powder materials; The feeding unit is connected with the material unit and is used for conveying powder materials; the weighing unit is arranged below the discharging end of the feeding unit and is used for measuring the weight of the received materials in real time; The conveying unit is arranged below the weighing unit and is used for conveying the proportioned materials to the next working procedure; the control unit, respectively with the feed unit with weighing unit electric connection, the control unit includes: The data acquisition module is used for acquiring weight signals of the weighing unit, rotating speed signals of the feeder and environmental signals in real time; the signal processing module is used for carrying out noise reduction pretreatment on the weight signals of the weighing unit and calculating real-time multidimensional feature vectors based on the pretreated data; and the deep learning intelligent decision module is used for outputting optimal control parameters of a controller of the feeding unit in the next control period according to the real-time multidimensional feature vector. Further, the control unit is integrated in a control cabinet, and the hardware of the control unit comprises: the PLC is used as a hardware carrier of the data acquisition module and the signal processing module; the embedded industrial computer is used as a hardware carrier of the deep learning intelligent decision module; the human-computer interface is connected with the PLC through an Ethernet or a communication line; wherein, PLC and embedded industrial computer pass through industrial Ethernet communicati