Affiliations: [a] Key Laboratory of Dependable Service Computing in Cyber Physical Society, MOE, Chongqing, China | [b] School of Automation, Chongqing University, Chongqing, China | [c] School of Software Engineering, Chongqing University, Chongqing, China
Correspondence:
[*]
Corresponding author: Qingyu Xiong, School of Software Engineering, Chongqing University, Chongqing, China. E-mail:cquxqy@163.com
Abstract: Microwave heating has been widely used for its advantage
of volumetric heating. For the material exposed by two input sources, due to
microwave superposition property, different phase between the input sources
can generate different power distribution. To achieve a uniform temperature
rising process, a sliding mode neural network combined with Cuckoo Search
algorithm is constructed. By regarding the power dissipation at each
sampling point as an independent actuator, a sliding mode neural network
control method is used to get the suitable inputs for those sampling points
under uncertain process parameters. As the constraint that only three inputs
variables, which are two input powers and the phase difference between the
two sources, can be controlled, Cuckoo Search is used to get the best fit
values to approximate the calculated inputs at those sampling points. Tap
water and White Pudding are chosen as the heating material to show the
correctness and the feasibility of this algorithm, and a uniform temperature
rising process can be obtained. Compared with genetic algorithm, Cuckoo
Search uses less time and has a better tracking result. The algorithm is
also verified in the test of microwave source output powers randomly jumping
around 100 ± 40% of the calculated values, sampling temperature with
the error of -0.3 ∼ 0.3 K, long sampling period of 4 seconds, and uniform
temperature rising processes can be also achieved. The average temperature
of those sampling points can well follow the reference trajectory.
Keywords: Power and phase control, sliding mode control, neural network, cuckoo search
