忆阻突触是神经形态计算系统的基本单元，对于构建新型非冯诺依曼架构的智能信息处理系统，实现信息存储和处理的有效融合具有重要意义。然而，目前忆阻突触研究存在机理尚不透彻，认知功能模拟效果不佳，故障诊断体系尚未完善等问题。基于此，本项目开展认知功能和基本结构与生物突触相匹配，且适用于神经形态计算系统的突触电路及其故障诊断研究。首先，基于实物忆阻器件，研究忆阻响应曲线的精细化拟合方法，构建具有平滑渐变多功能响应曲线的忆阻器模型；其次，阐述生物神经突触与忆阻特性之间的显/隐性关系，明确忆阻器复合电路的电学行为，建立具有不同电路结构的忆阻突触电路，完成突触行为测试；从器件和电路两个层面实现忆阻突触的故障定义和分类，根据不同故障对忆阻突触本身的影响，构建忆阻突触的故障字典，设计突触电路故障检测与诊断方法。本项目研究成果有望推动忆阻突触的优化发展，为构建基于忆阻突触的神经形态计算系统提供理论基础。

Memristive synaptic circuit is the basic unit of neuromorphic computing systems and key development for: 1) novel non-von Neumann architecture based intelligent information processing system and 2) realizing the effective fusion of information storage and processing. However, existing memristive synaptic circuits suffer from many limitations including the operating mechanism is still obscure, the cognition function modeling is not good, and the fault diagnosis system is still in progress. Therefore, the project will focus on the study of memristive synaptic circuit and fault diagnosis methods. The novel memristive synaptic circuit will possess biological synaptic features, both in cognition function and basic structure, and meanwhile, is suitable for neuromorphic computing system. The project first examines real memristive devices by 1) proposing the accurate fitting methods of memristive response curves and 2) designing the new memristor models with smooth, gradual, and multifunctional response curves. The project then expound the explicit and invisible relationship between biological synapses and memristive properties, investigate the complex electrical behaviors of the multiple memristor combination circuits, design and verify synaptic behaviors of many memristive synaptic circuits with different circuit configurations. Furthermore, the fault definition and classification of memristive synapses will be conducted from the device and circuit levels. The impact of different faults on the memristive synapses will be discussed, the corresponding fault dictionary will be built up, and the memristive synaptic circuit fault detection and diagnosis methods will be proposed. The successful completion of this project will promote the development of memristive synapses and provide theoretical foundation for memristive synapses-based neuromorphic computing systems.