A hybrid Quantum Denoising Autoencoder-CNN classification algorithm for canonical fluid dynamics and heat transfer problems

Recently, quantum computing has made remarkable strides in machine learning, leading to a significant impact on the field of quantum machine learning. To overcome the limitations posed by noisy intermediate scale quantum systems, the hybrid approach, combining both quantum and classical computing, has emerged as the most successful application mode. In this research, we introduce an enhanced algorithm called the Quantum Denoising Autoencoder-CNN (QDAE-CNN). This novel approach offers fast training speed, a lightweight design, and high performance. The QDAE-CNN method involves denoising the snapshots acquired from low-fidelity simulations through an Autoencoder section. Subsequently, the convolution layers employ a parameterized quantum circuit to perform feature mapping, ultimately achieving the task of classifying various numerical modeling datasets associated with conventional single/multiphase systems. The results demonstrate that QDAE-CNN effectively classifies distinct multiphysics and multiscale phenomena with acceptable classification accuracy. Moreover, when compared to numerous well-known landmark models, QDAE-CNN exhibits evident advantages. Overall, the successful extension of quantum computing into machine learning and the development of the hybrid QDAE-CNN algorithm signify promising advancements in the field of quantum machine learning, paving the way for future breakthroughs in various applications.