Quantum Machine Learning for Health State Diagnosis and Prognostics Gabriel San Martín Silva Department of Civil and Environmental Engineering, University of California Los Angeles, USA. E-mail: gsanmartin@g.ucla.edu Enrique López Droguett, Ph.D. Department of Civil and Environmental Engineering, and Garrick Institute for the Risk Sciences, University of California Los Angeles, USA. E-mail: eald@ucla.edu Key words: Risk, reliability, quantum computing, machine learning, qubits, quantum machine learning, prognosis and health management, fault diagnosis ABSTRACT Quantum computing is a new field that has recently attracted researchers from a broad range of fields due to its representation power, flexibility and promising results in both speed and scalability. Since 2020, laboratories around the globe have started to experiment with models that lie in the juxtaposition between machine learning and quantum computing. The availability of quantum processing units (QPUs) to the general scientific community through open API’s (e.g., IBM’s Qiskit) have kindled the interest in developing and testing new approaches to old problems. In this paper, we present a hybrid quantum machine learning framework for health state diagnostics and prognostics. The framework is exemplified using a problem involving ball bearings dataset. To the best of our knowledge, this is the first attempt to harvest and leverage quantum computing to develop and apply a hybrid quantum-classical machine learning approach to a prognostics and health management (PHM) problem. We hope that this paper initiates the exploration and application of quantum machine learning algorithms in areas of risk and reliability. 1 INTRODUCTION Deep learning has been used for almost a decade now by the prognosis and health management community to address questions such as the identification of health states of rotation machinery [1], the prediction of remaining useful life [2] or Bayesian analysis within complex and connected systems [3]. Since 2015, commercial applications have also appeared, with a clear intention of leveraging the hidden value contained within the immense amount of operational data that companies have collected through the years. Examples of this can be found through the specialized literature [4], [5]. While both research and commercial applications have produced useful results within the context of big data and complex, multidimensional systems (e.g., the assessment of health states through the analysis of images or videos) there are challenges that need to be addressed before these applications can be safely implemented in safety critical operations. One of them is speed and scalability: even with deep learning, for extremely dense or complex models, their training and inference can be quite challenging to execute, even more so to deploy on site. Over the last two years, general media and the research community have been starting to pay attention to advances in the quantum computing field, motivated mainly by remarkable feats such as the quantum supremacy experiment [6] and the first hardware ready Noisy Intermediate-Scale Quantum (NISQ) computers becoming available to the general public through cloud services, with hopes of identifying possible ways to optimize and speed up existing algorithms, modifying them or developing novel ones that harvest quantum mechanics phenomena such as entanglement, superposition and interference [7]. In this context, we present what it is, to the best of our knowledge, the first attempt to develop and apply a hybrid quantum-classical machine learning algorithm to address PHM problems, more specifically, diagnosis of health states in rotary machinery. We also present and discuss the encoding schema used to convert classical health monitoring data into quantum data and injected into quantum circuits, a required step to process information within the quantum computing realm. We will also release the code in a public GitHub repository required to reproduce the experiments and results portrayed here. The rest of the paper is organized as follows. Section 2 presents a brief introduction to fundamental quantum computing concepts, such as qubits, quantum gates, encoding schemas and a high-level view on quantum machine learning. Section 3 follows with the description of the hybrid approach for tackling health diagnosis and prognosis problems. Section 4 describes the case study: both the dataset used, the model specification and the results obtained. We conclude with a brief overview of the main takes of this work in Section 5.