Safety Lifecycle for Developing Safety Critical Artificial Neural Networks

Abstract

Artificial neural networks are employed in many areas of industry such as medicine and defence. There are many techniques that aim to improve the performance of neural networks for safety-critical systems. However, there is a complete absence of analytical certification methods for neural network paradigms. Consequently, their role in safety-critical applications, if any, is typically restricted to advisory systems. It is therefore desirable to enable neural networks for highly-dependable roles. Existing safety lifecycles for neural networks do not focus on suitable safety processes for analytical arguments. This paper presents a safety lifecycle for artificial neural networks. The lifecycle focuses on managing behaviour represented by neural networks and to provide acceptable forms of safety assurance. A suitable neural network model is outlined and is based upon representing knowledge in symbolic form. Requirements for safety processes similar to those used for conventional systems are also established. Although developed specifically for decision theory applications, the safety lifecycle could apply to a wide range of application domains.