Intelligent fault diagnosis of machinery using digital twin-assisted deep transfer learning

Digital twin (DT) is emerging as a key technology for smart manufacturing. The high fidelity DT model of the physical assets can produce system performance data that close to reality, which provides remarkable opportunities machine fault diagnosis when measured condition are insufficient. This paper presents an intelligent framework machinery based on and deep transfer learning. First, built by establishing simulation with further updating through continuously from asset. Second, all important conditions be simulated DT. Third, new-type structure novel sparse de-noising auto-encoder (NSDAE) developed pre-trained source domain, generated Then, achieve accurate possible variations in working characteristics, NSDAE fine-tuned using parameter only one sample target domain. presented method validated case study triplex pump diagnosis. experimental results demonstrate proposed achieves limited amount outperforms other state-of-the-art data-driven methods.