Raymond mill is an important grinding equipment. It is very important to design and improve it scientifically and effectively, so that it can feed automatically, which is conducive to improving production accuracy and efficiency, and also conducive to saving resources. This paper designs a controller of Raymond mill automatic feeder based on STM32 single chip computer and machine vision technology. The parameters are automatically set by using neural network technology. It can converge quickly for some simple use scenarios, and has been widely used in engineering.
In the non-linear field, adaptive dynamic programming has a very strong point. It plays an important role. In this paper, it is discussed to lay a solid foundation for its further development. Due to the high cost of the automatic feeder system and the difficulty of its transformation, the automatic feeder system has not fully covered the factories. In view of the fact that some factories are not well renovated and the corresponding investment is limited, it is necessary to design a set of low cost, convenient installation and renovation, practical and reliable automatic feeder control system. Under the above conditions, it can also provide reliable help for automatic feeder control and energy saving, reduce power waste and improve the utilization rate of automatic feeder. It only needs to design an automatic feeder control system by using single-chip computer. Through intelligent control, it can reduce the waste of power and the short service life of equipment caused by manual control. At present, China’s equipment and energy crisis has begun to affect people’s daily production and life, rising prices, population unemployment and other social realities are closely related to it. Especially in recent years, some parts of the country have been over-exploited. In order to pursue the rapid development of regional economy, it has caused a serious waste of resources and energy consumption. The design of traditional Raymond abrasive feeder system has many shortcomings, such as complex wiring, waste of energy, backward management, low comfort and short service life, which can not meet people’s needs in many aspects. Therefore, with the development of technology and the improvement of people’s demand, how to design a new feeder control system with low energy consumption, easy management and control, and high comfort is the key content of industrial construction in the new era.
SCM power-on reset, self-check peripheral chips and sensors, error correction, display display shows the cumulative running time of Raymond mill, control the blower motor, spindle motor, feeding motor relay suction, often closed contacts disconnected, to ensure that Raymond mill starts in the correct order. After the start-up sequence of Raymond mill is completed, the current of spindle motor and fan is collected, and the feeding speed is controlled by integration, dynamic neural network operation. Using image recognition to calculate speed data, inverse proportional integral and calculation, the control data is obtained to trigger the conduction angle of thyristor, and the analysis speed is stabilized by changing the magnetic field produced by the excitation coil. When the speed of the current and analyzer of the spindle motor and blower motor exceeds the set upper/lower limit data and the temperature of the spindle motor exceeds the set value, the alarm relay sucks in, the sound and light alarm stops running, and the display screen displays the corresponding fault code.
When the current of the spindle motor is less than the lower limit parameter set by the dynamic neural network, the material in the grinding room is less, the alarm relay is absorbed, and the alarm is reset when the current of the spindle restores. Press auto stop press and twist, stop feeding, 30 seconds later when the current of the spindle is less than the lower limit parameters set by dynamic neural network, stop the spindle motor, fan, analyzer in order to prevent material from clogging the roller grinding device. When the running time exceeds the set time parameters after parking, the display time will be out of time, otherwise the accumulated running time will be displayed.
In this paper, OpenCV is used as a test platform for machine vision module. According to the actual use of Raymond Mill automatic feeder, the resolution of OpenCV with 300,000 pixels and 320*240-640*480 resolution can meet the requirements. In addition, thermostatic element OpenCV Cam M7 is powered by 216 MHz ARM Cortex M7 processor and equipped with IO port, which can drive various types of motors. The positioning system of Raymond mill automatic feeder uses STM32F765VI as the control module of Raymond mill automatic feeder, and with TB6612 motor driving board to realize the control of Raymond mill automatic feeder. STM32F765VI can simulate its movement through TB6612 motor driving board. The mill control module uses the control board with STM32F103 to control the steering gear. In this paper, two steering gears are used to simulate the real working conditions. In actual production, Raymond mill automatic feeder has strict technical parameters. The machine vision module, Raymond mill automatic feeder control module and grinding chamber control module are composed of STM32F765VI, TB6612 motor drive board and STM32F103. The control module of Raymond mill automatic feeder is responsible for controlling the movement of Raymond mill automatic feeder.
And send and collect information to the grinding room control module and machine vision module. The USB module on the module is responsible for loading the program, and the power module is responsible for supplying power to the control module. Machine vision module is responsible for processing the camera image and extracting the features. At the same time, the image features are compared with those in the machine vision library. The grinding room control module is responsible for controlling the tool replacement and cleaning operation of the grinding room. The AprilTag marking method is embedded in STM32 single chip computer, and a micro Python interpreter and OV2640 image sensor are mounted on OpenCV open source micro-machine. Through Python script language programming, the camera can be controlled to focus and judge the object’s spatial position information, extract data features and color tracking, and guide Raymond mill to auto-feed. Material machine tracking object and other functions. This enables OpenCV to quickly obtain image information and make the positioning system of Raymond mill automatic feeder get more location information.
However, in the actual production and life of Raymond mill automatic feeder, because the working environment is outdoor, the ambient light intensity changes obviously from the start of work in the morning to the end of work in the afternoon and is extremely vulnerable to environmental impact.
。 AprilTag marking method is a method that uses Tag markers to paste onto objects and then judge the distance of objects, so it is very easy to invalidate due to the change of light intensity in the environment. In the process of experiment, the change of ambient light will directly lead to the reduction of black and white pixel difference and the error of mark recognition. If it is applied to the Raymond mill automatic feeder, the positioning failure will probably occur. Therefore, the camera visual angle ranging method can also be used as an alternative to the AprilTag marking method when the ambient light is weak.
In software aspect, the dynamic programming algorithm based on neural network is mainly used. It mainly makes the whole optimization strategy more and more close to the actual demand through successive approximation of functions, so as to get the optimization index. Usually, two neural networks are used to optimize the control of functions, so that the adaptive dynamic programming in non-linear problems can be effectively carried out.
Solve. Many experts have done extensive and in-depth research on optimal control in non-linear systems, and put forward the viewpoint of dynamic programming. So adaptive dynamic programming mainly uses neural network to solve optimization problems. Because neural network can adjust its weight, it can be solved by several mathematical models. These problems are solved optimally, assisting managers to make judgments and decisions, and improving the efficiency of the whole project. Therefore, to a great extent, it has a strong ability of self-learning and adaptation. However, because of the need to initialize any control scheme, the strategy we calculated is not necessarily the optimal allowable control scheme, so we need value iteration algorithm to optimize it. In the calculation of the whole value iteration algorithm, the requirement for the internal dynamic characteristics of the system is very high, which needs to be calculated in every step of calculation. At the same time, we should try our best to make the state function more approximate to the modern price function and finally get the optimal situation. When the number of iteration steps approaches infinity, we can do so. In order to find the optimal cost function, we can analyze the optimal strategy. If the control system of Raymond mill automatic feeder is widely used in the market, its automatic production can be realized and the production efficiency can be greatly improved. With the sustainable development strategy and the direction of social development put forward by our government, the environmental protection and energy issues in our country have attracted wide attention from all walks of life. The efforts made by all walks of life in saving resources and developing new energy sources are obvious to all. We also need to respond to the call of the state, constantly reform and innovate technology and industrial structure, reduce resource consumption, improve control effect, and strive for higher economic returns.