The complexity of modern safety critical systems is becoming higher with technology level growth. Nowadays the most important and vital systems of automotive, aerospace, nuclear industries count millions of lines of software code and tens of thousands of hardware components and sensors. All of these constituents operate in integrated environment interacting with each other — this leads to enormous calculation task when testing and safety assessment are performed.
There are several formal methods that are used to assess reliability and safety of NPP I&C (Nuclear Power Plant Instrumentation and Control) systems. Most of them require significant involvement of experts and confidence in their experience which vastly affects trustworthiness of assessment results.
The goal of our research is to improve the quality of safety and reliability assessment as result of experts involvement mitigation by process automation. We propose usage of automated FMEDA (Failure Modes, Effects and Diagnostic Analysis) and FIT (Fault Insertion Testing) combination extended whith multiple faults approach as well as special methods for quantitative assessment of experts involvement level and their decisions uncertainty. These methods allow to perform safety and reliability assessment without specifying the degree of confidence in experts.
Traditional FMEDA approach has several bottlenecks like the need of manual processing of huge number of technical documents (system specification, datasheets etc.), manual assignment of failure modes and effects based on personal experience. Human factor is another source of uncertainty. Such things like tiredness, emotional disorders, distraction or lack of experience could be the reasons of under- and over-estimation.
Basing on our research in field of expert-related errors we propose expert involvement degree (EID) metric that indicates the level of technique automation and expert uncertainty degree (EUD) metric which is complex measure of experts decisions uncertainty within assessment. We propose usage of total expert trustworthiness degree (ETD) indicator as function of EID and EUD.
Expert uncertainty assessment and Multi-FIT as FMEDA verification are implemented in AXMEA (Automated X-Modes and Effects Analysis) software tool. Proposed Multi-FIT technique in combination with FMEDA was used during internal activities of SIL3 certification of FPGA-based (Field Programmable Gate Array) RadICS platform for NPP I&C systems. The proposed expert trustworthiness degree calculation is going to be used during production activities of RPC Radiy (Research and Production Corporation).
Our future work is related to research in expert uncertainty field and extension of AXMEA tool with new failure data sources as well as software optimization and further automation.
