Abstract
The purpose of this ISA SafeSec event is to share some perspectives on the use of AI for safety-critical systems and invite for a discussion in a Q&A session.
The presentations covered during the first 30 minutes focuses on the following: Sr. Researcher Thor Myklebust will first give an overview of key frameworks and standards on AI supported by some examples ongoing projects where AI is applied to safety-critical systems. Professor Lundteigen will continue with some perspectives on the potential use of AI for safety-instrumented systems for process industry, with basis in the recently published ISO/ IEC TR 5469. Despite the opportunities from using AI, there are many challenges that must be overcome to ensure performance within the constraints of safety. PhD candidate Niclas Flehmig will share some examples for why the performance of a trained AI system may degrade after being put into operation. A proposal for an extension to the reference architecture of ISO/IEC TR 5469 is presented and discussed.
The presentations covered during the first 30 minutes focuses on the following: Sr. Researcher Thor Myklebust will first give an overview of key frameworks and standards on AI supported by some examples ongoing projects where AI is applied to safety-critical systems. Professor Lundteigen will continue with some perspectives on the potential use of AI for safety-instrumented systems for process industry, with basis in the recently published ISO/ IEC TR 5469. Despite the opportunities from using AI, there are many challenges that must be overcome to ensure performance within the constraints of safety. PhD candidate Niclas Flehmig will share some examples for why the performance of a trained AI system may degrade after being put into operation. A proposal for an extension to the reference architecture of ISO/IEC TR 5469 is presented and discussed.
