Design Of Several Memristive Neural Networks And Switching Power Supply Controllers

Memristor, a two terminal passive electronic device, is used to describe the relationship of magnetic flux and charge. Its nonlinear characteristics and natural memory ability make it have unique advantages in nonlinear system and neural network. At first, the behavior and mechanism of memristor is very similar to the synapses of human brain. So it can be used as electronic synapses in artificial neural networks, which can increase neurons’ connectivity density and improve the integration density of neural network. In addition, the memristance of memristor varies with the voltage(or current) applying to its ends, therefore it has the programmable characteristics. Switching power converter is widely used in high efficiency power supply and DC motor drive because of its small size, light weight, high efficiency and high power density. The programmable memristor can be applied to control the power- switching converters, which can make the converters stabilize at different equilibrium points. And this control method has other better performance. Based on a deep study on memrisor’s nonlinear characteristics and synaptic properties, this paper discusses memristor’s applications in artificial neural networks and switching power converter. The main research contents of this paper include the following parts:Firstly, Gale memristor is used to construct a memristor-based pulse coupled neural network(M-PCNN), which internalizes the biological characteristics of memristor into PCNN to make it more intelligent. MATLAB simulation verifies that this network can process medical image including Computed Tomography(CT) and Magnetic Resonance Imaging(MRI) image fusion, CT image de-noising and CT image edge extraction.Then, a memristor – based BP neural network PID controller is proposed. As well as the weight updating rule of spintronic memristor using as electronic synaps is educed. Applying this scheme to control Buck converter system and MATLAB simulation verifies the correctness of the theoretical derivation.In addition, a programmable memristive bridge circuit is proposed using the programmable characteristic of spintronic memristor. This programmable memristor is applied to replace the feedback resistor of the pulse generator circuit, in which the pulse width can be adjustable. SPICE simulation verifies that the output voltage of Boost converter can be stabilized at different equilibrium points using this pulse generator to control the MOSFET in Boost circuit.Lastly, a sliding surface with double integral equation is introduced to design a PWM-based sliding-mode controller in this paper. The additional integral term can make the control system keep the same order as the original system and improve the steady-state performance of the controlled system. PSIM simulation certifies that this method has strong robustness on controlling a two-phase interleaved boost converter.