Quantum Optimization of Bayesian Network Learning for Smart Maintenance Service Systems

Smart maintenance service systems increasingly rely on advanced diagnostics to support servitization models that deliver reliability and availability as managed outcomes. This paper introduces a Quantum Computing-as-a-Service (QCaaS) approach for diagnostic intelligence by integrating Bayesian Network (BN) learning with digital twin (DT) representations. BN structure learning, intervention analysis, and DT model fusion are formulated as Quadratic Unconstrained Binary Optimization (QUBO) problems and solved using the Quantum Approximate Optimization Algorithm (QAOA) implemented in Qiskit Aer. A hybrid workflow connects BN-to-QUBO encoding, DT data pipelines, and QCaaS-based optimization services. Benchmark experiments demonstrate that quantuminspired methods demonstrate improved structure accuracy and runtime scalability compared to classical solvers, positioning QCaaS as a scalable computational capability that enhances fault reasoning, uncertainty modeling, and decision support in digital-twin-enabled maintenance servitization ecosystems.