Negative Correlated Evolutionary Model For Analog Circuits

As the important interface between digital processing and the external world, analog circuits still remain irreplaceable in modern electronic systems. As the use of the modern electronic systems becoming increasingly wide and deep in recent years, its application environments are becoming more and more diversified and complicated. The potential occurrence of fault in circuit may be varied and unpredictable. In order to ensure its stability in the actual working environment, its fault-tolerant design of analog circuits becomes particularly important.When some component failures experienced, analog circuit can maintain its functioning, we take the circuit is fault-tolerant. For special circumstances, even in an unknown environment, if the traditional manual design methods can not complete a circuit fault-tolerant design, it can be through the evolution of a fault-tolerant approach to solve. The advantage of evolutionary design of fault-tolerant analog circuit design is that it does not rely on too much experience and stationary design method of designers, can explore more ideas, and be able to search more new combination of circuit structure and component parameters. Achieving fault tolerance at the same time, design time and cost can be reduced significantly. Existing research works about the fault-tolerant evolutionary design of analog circuit are mainly for the faults which are known a prior. The specialized circuits robust to the special failure are weak to the unknown problems. The fault-tolerant evolutionary design based on negative-correlation redundancy approach is a novel and promising way for fault-tolerant design of analog circuits under uncertain faults. However, it is Just a prototype. There has much more room for improvement. Conducting some more in-depth research on this method is necessary. This thesis shows a exploratory investigation on the fault-tolerant evolutionary design based on negative-correlation redundancy approach.In this thesis, the three main works are as follows:(1) A new negative-correlation redundancy evolutionary design model is proposed from the perspective of constrained optimization. For the problem of negative-correlation coefficient is difficult to determine and thus affect the existing evolutionary algorithm convergence efficiency, stochastic ranking is introduced to the model to avoid selecting the hard-to-set negative-correlation coefficient. In the model, circuit function and correlation coefficient are respectively taken as the objective function and constraints, and the behavior of the mechanism of negative-correlation evolutionary models is observed. Experimental results show that the convergence efficiency of the proposed negative-correlation model is improved.(2) For the problem of limited capacity of the scale of the existing populations-based negative-correlation evolutionary model, a single population scalable negative-correlation model is proposed. Compared to the populations-based negative-correlation evolutionary model, the proposed model can provide more combinations of negatively-correlated redundancies, and has a higher convergence efficiency and scale scalability.(3) From the frequency domain characteristic of circuit output signal, the weighted average method is proposed to combine the negatively correlated analog modules. Experimental results on the analog filter show that combining the negatively correlated modules by the weighted average method is feasible on some failures.In summary, this thesis carries out some exploratory research on the fault-tolerant evolutionary design based on negative-correlation redundancy approach. A negative-correlation evolutionary model with a higher convergence efficiency and scale scalability is proposed. The fault-tolerant evolutionary design based on negative-correlation redundancy approach is put to towards the practical application of a step forward.