Research and application of tin furnace temperature control system based on neural network and DSP

Modern electronic component assembly requires a more stable soldering temperature in the tin furnace, and the difficulty of high temperature control of the tin furnace is also increased. With the continuous deepening of the application research of predictive neural networks, due to its large amount of computational data and slow convergence, its application is limited by hardware, and there are not many practical applications. However, the DSP high-speed digital signal processing speed is getting faster and faster, and the online real-time control capability is getting stronger and stronger, which has a significant effect in the application of neural networks. This paper uses the predictive ability of neural networks to learn and adjust temperature parameters, and combines digital signal processing (DSP) modules for control and calculation to achieve high-speed calculation and processing control, and finally realizes online real-time compensation heating control of the tin furnace temperature control system.
1 Neural network control structure
Neural network is a good data processing technology for event prediction. Rules are discovered during the learning process, and the parameters of the control function are learned and adjusted by combining prediction and DSP calculation and processing control. The control system structure based on neural network and DSP is shown in Figure 1. The control system consists of two parts: predictive neural network and DSP digital signal processing calculation control. These two parts have a common input signal, namely the network temperature error e. The prediction neural network predicts and evaluates the sampled temperature samples and the pre-set samples. After the prediction value is properly corrected by the influencing factors of the external environment control, the weight of the prediction neural network and the parameters of the control function are adjusted. The DSP operation and processing control center receives and issues various control commands and executes them through high-speed data operation and processing based on the control temperature error and the neural network prediction value, including real-time temperature display, temperature control output, temperature limit alarm and other input and output parameters. The actuator is the object of the control module, and the final object of the temperature control system is the heater. Therefore, the tin furnace temperature control system takes the change of temperature as the entire control core, which is converted by the temperature sensor and effectively controlled through the prediction of the neural network and digital signal processing (DSP).

The prediction of this control system adopts BP neural network [1]. Its characteristic is that only the neurons between the two adjacent layers are connected to each other, and the output neuron outputs the predicted value. The prediction neural network structure is shown in Figure 2. The network structure is divided into three layers, namely the input layer, hidden layer and output layer. The input layer is responsible for receiving data and does not perform calculations. Among them, x0 is the initial value of the activation function, a random number between [-1, 1], and x1 is the temperature error e of the network control system, and x2 and x3 are the voltage and current detection values ​​of the heater respectively.

In practical applications, w _ij is the connection weight of each layer, and the control parameter net of the activation function f(net) is used to control the convergence range of the network system, which is beneficial to ensure the stability of the entire system.

Therefore, through the experiment of the algorithm, the adaptive factor (1-β) is used to flexibly change the weight modification amount as needed, and the learning of the network convergence speed is realized.
2 DSP system implementation
The tin furnace heating temperature control system can implement PID control on the key heating control components, thereby accurately realizing the temperature heating compensation control of the tin furnace temperature change, increasing the effective application of thermal efficiency, and is conducive to improving the utilization rate of electric energy, achieving energy saving, and at the same time improving the qualified rate of circuit board welding. The system structure includes DSP processing [4], temperature sensor detection, voltage and current detection, temperature display, temperature keyboard parameter setting, temperature alarm, control output and other functions. The system operation is simple, the display is intuitive, and the control is convenient. The core design of the control system is the DSP digital signal operation and processing controller. The DSP chip TMS320FL2407 is used as the control center, and the peripheral circuit is used to realize the system control.
2.1 System Hardware Circuit
The control system mainly uses the DS1002 system board and its multi-channel I/O board. The system is mainly composed of the digital signal processing chip TMS320FL2407 center and the corresponding interface peripheral circuits, including the controllable drive circuit of the heating device, the temperature acquisition A/D circuit, the temperature setting keyboard input, the temperature display circuit and the alarm circuit. The hardware circuit principle of the real-time control system is shown in Figure 3.

(1) DSP digital processing circuit. It processes the digital signal converted from the received temperature sensor detection, completes the processing of the host computer's predicted state, and outputs the execution control signal;
(2) Controllable drive circuit of the heating device. The actuator of the heater can adopt the traditional relay type and thyristor type, but its control is simple and its performance is poor. This heating temperature control system adopts PID method, and the analog quantity automatically adjusts the voltage phase angle, which can continuously control the temperature, solve the accuracy and stability requirements of temperature control, and achieve the control temperature accuracy of ±2 ℃. According to the sampling data provided by the temperature sensor, the PWM control signal output by the DSP controls the actuator of the heater to achieve the purpose of online real-time control of the tin furnace temperature;
(3) Temperature, voltage and current acquisition A/D circuit. When the system is initialized, it continuously collects the current real-time temperature of the tin furnace through the sensor. After the system board completes the A/D conversion of the analog data and temperature error collected by the temperature sensor through the I/O board, it is sent to the DSP control center through the high-speed path. At the same time, it combines the voltage and current data, compares the collected data with the set value in the comparison register, and performs D/A conversion after DSP high-speed calculation processing, and enters the control execution mechanism;
(4) Temperature setting keyboard input. The input of the system's temperature control parameters is realized by 8 input touch buttons through the communication between the serial port and the DSP system board;
(5) Display circuit and alarm circuit. The display circuit component unit is composed of a liquid crystal module board, an indicator light-emitting diode and a corresponding drive circuit. The liquid crystal module board displays the operating parameters of the system, mainly displaying the real-time temperature of the tin furnace in large fonts, and displaying the current voltage and current of the heater in small fonts. When working normally, the power indicator diode emits red light and the working indicator diode emits green light; when the working indicator diode emits yellow light, it indicates that the system circuit has a fault and the tin furnace temperature may be abnormal. It is necessary to disconnect the control system for maintenance or restart the system and reset it. At the same time, an alarm prompt sound is given, and the DSP control output terminal sends a control voltage to disconnect the silicon circuit to stop the heater from continuing to work. The alarm circuit consists of a speaker and its driving amplifier circuit, and the alarm sound is realized by software programming.

2.2 System Software TMS320LF2407
is selected as the system chip . The algorithm developed in C language is compiled and linked by the host computer, converted into the target file format (COFF) that can be received by DSP, and loaded into DSP for execution control. Figure 4 is the flow chart of the control system software. The analog parameters of the collected temperature sensor are converted into digital signals through A/D. The control algorithm written in C language is compiled through the CL30 development tool and transmitted to the DSP system board for calculation and control to obtain the control drive signal. In the DSP control center, in the process based on the neural network prediction algorithm, the host PC captures the status information through RAM, monitors the parameter performance of the control system in real time, provides a strong analysis basis for the control algorithm of the temperature control system, and completes the debugging of the system.

After the system is initialized, the control system opens interrupts and performs various function tests. The temperature parameter sampling is converted into digital signals by A/D and sent to the DSP operation center for digital processing. The output control signal is converted by D/A to drive the actuator. Timer T0 is used as the sampling setting cycle. Each sampling cycle completes a sampling and operation analysis process until the prediction neural network learning is completed and the interrupt is exited and stopped.
3 System application results
Temperature, voltage and current detection devices are installed in the circuit. After circuit experiments, the results of the prediction neural network algorithm are implanted in the DSP to realize temperature PID control. The temperature control curve is shown in Figure 5.

The method of combining predictive neural network and DSP high-speed computing processing is applied to the temperature PID control system of the tin furnace, which greatly improves the stability and accuracy of the control system temperature, provides reliable soldering temperature for electronic components during assembly, and reduces the possibility of damaging components and circuit boards due to high temperature.
References
[1] Jiang Zongli. Introduction to Artificial Neural Networks [M]. Beijing: Higher Education Press, 2001: 39-52.
[2] Wang Yongji, Tu Jian. Neural Network Control [M]. Beijing: Machinery Industry Press, 1999: 1-68.
[3] Liu Tienan. A new second-order learning algorithm for multi-layer forward neural networks [J]. Control Theory and Applications, 2000, 17(5): 721-724.
[4] Zhu Zhifu. Application of DSP in boiler feed water frequency conversion control system [J]. Journal of Henan University of Science and Technology (Natural Science Edition), 2007, 28(4): 21-32.