Quantum Computing Achieves Precise Machine Failure Detection Using 133 Qubits

Scientists are addressing the critical industrial challenge of promptly detecting machine failures to improve efficiency and reduce costly downtime. Larry Bowden (Digital, Woodside Energy, Perth, Australia), Qi Chu, and Bernard Cena (Digital, Woodside Energy, Perth, Australia), alongside Kentaro Ohno, Bob Parney, and Deepak Sharma et al. (Quantum, IBM Research), present a novel failure detection algorithm that utilizes quantum computing in combination with a statistical change-point detection approach. Their research is significant because it leverages projected feature maps to enhance anomaly detection. Importantly, they have demonstrated the feasibility of integrating quantum computing into real-world industrial maintenance, successfully executing the algorithm on IBM’s 133-qubit Heron processor. This work validates a powerful new system for identifying anomalies in complex, noisy data and opens promising avenues for advanced predictive maintenance strategies.

Researchers consider multidimensional sensor readings from industrial machinery. Access to normal data refers to periods during which machines operate under steady-state conditions. Stationarity implies that the machines remain in a steady state in the wide sense. Let T denote the given target and normal time series, where xt represents d-dimensional vectors for each timestamp t. We assume a subset Xnorm = {xt}t1≤t≤t2 over an interval [t1, t2] corresponds to a normal dataset, i.e., one that does not contain anomalies. Given a window length L0 and a sliding width w0, time windows are formally defined as Xs = {xt}t=sw, …, sw+L, with s = 1, 2, … .

With this notation, the task is to measure the statistical divergence between distributions of Xnorm and Xs for each s. This divergence is treated as a score representing the likelihood that an anomaly or change has occurred. Despite considerable advances in change-point detection algorithms, achieving highly precise detection remains challenging, particularly for multidimensional data. Among existing methods, density ratio estimation is empirically the most powerful non-parametric approach to change-point detection. Therefore, we incorporate quantum modeling into this approach to enhance accuracy.

Computing Divergence via Density Ratio Estimation: Let p(x) and p′(x) denote density functions representing probability distributions over a domain in Rd. Density ratio estimation aims to estimate the ratio function r(x) := p(x)/p′(x) from sample sets. In projected quantum models, we utilize a projection of ρ(x) rather than ρ(x) itself to represent the quantum feature of x. In this study, we employ the one-particle reduced density matrix (1-RDM) ρk(x), with k = 1, …, nq, obtained by projecting ρ(x) onto each qubit: ρk(x) := Trj=k[ρ(x)] (14). Although the projection loses information on the entanglement of quantum states, it can improve prediction accuracy for machine learning models. Based on this observation, the main idea of this work is to build a detection model using 1-RDM features instead of the full density matrix.

Quantum change-point detection identifies machine anomalies

Experiments revealed the algorithm’s capability to accurately identify anomalies present within noisy time series data, marking a significant step towards proactive diagnostics. Researchers formulated machine failure detection as a change-point detection problem applied to multi-variate time series data, assuming access to normal operational data and a degree of stationarity. The study leveraged projected quantum models, a novel quantum machine learning technique, combined with statistical divergence estimation to measure anomaly scores. Measurements confirm that a large anomaly score indicates a likely machine fault, enabling timely intervention and preventative maintenance.

Tests prove the effectiveness of quantum feature transformation in identifying change-points within complex time series data. The team extracted projected quantum features, transforming original sensor data into a dimension dependent on the quantum circuit employed. Data shows that this process enhances the precision of anomaly detection in machine monitoring systems. Results underscore the potential of this quantum-based failure detection system to improve machine reliability and reduce maintenance costs. While acknowledging current computational overhead that may not yet guarantee speed advantages over classical methods, scientists highlight the transformative potential and broader implications of quantum computing in predictive maintenance. The research provides a framework for identifying subtle anomalies in machine operation data, facilitating timely interventions and preventative maintenance strategies, a crucial advancement for industrial efficiency.

Quantum anomaly detection for industrial machines promises increased

The researchers validated their algorithm using both benchmark multi-dimensional time series datasets and real-world data gathered from operational machines, confirming its practical applicability. This work highlights the potential of computing in industrial diagnostics and establishes a foundation for more advanced predictive maintenance algorithms. The authors acknowledge that the precise reasons for the performance improvements observed with projected quantum models, particularly in handling noisy data, require further investigation. Future research will focus on analysing the quantum feature extraction process, potentially through Fourier analysis, to better understand its impact on detection accuracy. Additionally, exploration of alternative quantum circuits tailored to this specific application is planned, aiming to optimise performance and broaden the scope of this promising technology.