This is an extended abstract of the manuscript 'Solving Industrial Fault Diagnosis Problems with Quantum approaches to perform fault diagnosis: (i) using Grover's algorithm, and (ii) using the Quantum Approximate We found the Grover algorithm generates more solutions and needs some postprocessing to obtain minimal diagnoses, but it is much faster than the more accurate QAOA approach. Modern industrial cyber-physical systems are characterized by high interconnectivity, modularity, throughput, and number of system parameters. But what happens if such systems fail? Especially in environments with little human oversight (such as in a ship's engine room) reliable and automated diagnosis increases system resilience. However, fault diagnosis is a challenging task that demands in many cases exponential runtime2[]. To mitigate the runtime issues, eficient quantum computation for fault diagnosis problems has become an important area of applied research. Nowadays, quantum fault diagnosis faces two problems: i) Eficiently solving the diagnosis problem: Many diagnostic procedures are based on reduction to the NP-complete satisfiability problem3][. ii) Obtaining available models from heterogeneous cyber-physical systems: In practice, many companies still rely on manual and expert-driven diagnosis and repair procedures. These, however, are often expensive, slow, and error prone.
1. Extended Abstract NP-complete. CEUR Workshop
ISSN1613-0073 Preprocessing / Model Generation
C0ANDC1 IMPLIES O1 C0ANDC2 IMPLIES O2 ...
In our article, we present a data-driven and explainable fault diagnosis approach and demonstrate its usage on an IBM Falcon quantum processor. For this, we make three contributions: (i) A novel algorithm qDiagCPS_Grover that trades speed for accuracy. (ii) The novel algoritqhmDiagCPS_QAOA which is slower in absolute terms, but more accurate. (iii) The algoritGhTmSD that creates propositional logic system descriptions (system models) from data.
The motivation behind our contributions is the following: Since diagnosis is an NP-complete problem, it is reasonable to look for more eficient solutions using quantum computers. To realise this, we use two approaches: a) We convert a propositional logic model into a SAT problem and use Grover’s algorithm 5[] to find the minimum satisfying assignment. b) We transform the propositional logic model into a quadratic binary optimization problem (QUBO), which is solved using the Quantum Approximate Optimization Algorithm (QAOA). We conclude that both solutions find the minimum cardinality diagnosis ′ ⊂ , with ∈ being the set of faulty components within the set of all components .
The evaluation was performed on several systems and benchmarks modelling use-cases from the process industry. While we received good results, especially with the QAOA-based algorithm, too few qubits were available to perform relevant evaluations that scale for cyber-physical systems. Declaration on Generative AI The author(s) have not employed any Generative AI tools.