Design Of Electronic Control System For Multi-leaf Collimator Based On FPG

Radiation therapy is widely used in the treatment of malignant tumors,which works by using high-energy radiation to kill cancer cells or prevent their growth.However,high-energy radiation can also affect normal cells,so precise control of the irradiation range is particularly important during radiotherapy.Currently,most radiotherapy devices use a Multi-Leaf Collimator(MLC)as the control device for the irradiation range,which involves large-scale motor control and requires high parallelism.Therefore,this paper fully utilizes the parallel advantages of FPGA and develops an FPGA-based MLC electrical control system to achieve efficient control of the irradiation range.In order to achieve efficient control,this paper first builds a software and hardware experimental platform based on FPGA according to the existing mechanical structure,and then designs the blade motion control strategy.The BP neural network is used to optimize the blade speed control,and the coordinated control of the position and speed of 120 blades is achieved.Experimental results show that the system makes the hardware structure lightweight and can efficiently complete the initialization,reset,and formation of 120 blades.The communication cycle with the host computer is less than 20 ms,the blade positioning accuracy is less than0.25 mm,and the blades can quickly and stably reach 18mm/s,which meets the requirements of the medical MLC design.The main work of this paper includes four aspects:(1)explaining the system composition of MLC and its working principle in radiation therapy,based on which the overall design framework of the MLC system is proposed;(2)the hardware and logical design of the FPGAbased MLC electrical control system;(3)in response to the difficulty of selecting PID parameters for DC brushed servo motors,this paper uses BP neural network optimization to obtain the parameters;(4)debugging and analysis of the FPGA-based MLC electrical control system.In the hardware design part,the system hardware is divided,and a six-layer FPGA main control board composed of a core board and a base board,and a four-layer motor connection board with a leaky air structure are designed.In the logic design module,a modular design approach is used for module partitioning,and the design of system communication,MLC data pre-processing,anti-collision,and blade motion control algorithms are emphasized.Regarding the BP neural network training process,the fixed learning rate is changed to a variable learning rate to optimize the blade performance during movement,as slow speed and multiple extreme points were a problem.In the verification stage,verification methods were formulated for communication,single blade control and multiple blade assembly,and software and instruments such as Vivado’s ILA,network communication tools,and high-precision laser rangefinders were used to conduct system function testing and performance analysis.