Incremental Abstraction for Diagnosability Verification of Modular Systems
Paper i proceeding, 2019

In a diagnosability verifier with polynomial complexity, a non-diagnosable system generates uncertain loops. Such forbidden loops are in this paper transformed to forbidden states by simple detector automata. The forbidden state problem is trivially transformed to a nonblocking problem by considering all states except the forbidden ones as marked states. This transformation is combined with one of the most efficient abstractions for modular systems called conflict equivalence, where nonblocking properties are preserved. In the resulting abstraction, local events are hidden and more local events are achieved when subsystems are synchronized. This incremental abstraction is applied to a scalable production system, including parallel lines where buffers and machines in each line include some typical failures and feedback flows. For this modular system, the proposed diagnosability algorithm shows great results, where diagnosability of systems including millions of states is analyzed in less than a second.

diagnosability

abstraction

verification

nonblocking

Modular systems

Författare

Mona Noori-Hosseini

Chalmers, Elektroteknik, System- och reglerteknik, Automation

Bengt Lennartson

Chalmers, Elektroteknik, System- och reglerteknik

24th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA)
Zaragoza, Spain,

Ämneskategorier

Inbäddad systemteknik

Datorsystem

Annan elektroteknik och elektronik

Mer information

Skapat

2019-07-27