Mechanical Acoustic Signal Assisted Translational Model for Industrial Human-Machine Interaction

Human-machine interaction (HMI) is a popular issue in industrial intelligent manufacturing research, and it is of great significance to realize mechanical anthropomorphic operation. This paper proposes a novel and universal translational model based on mechanical acoustic signals with visual information for expert operation imitation in mechanical operation environment. Through reasonable state characterization and feature extraction, the translational model is approximated as a hidden Markov model (HMM) and it is also designed for three challenges of industrial HMI: (i) Expectation-Maximization (EM) algorithm is first adopted to achieve signal fusion to relieve the interference of external noise in the propagation of mechanical acoustic signal. (ii) Adaptive clustering method is then designed to realize online reorganization of state points to obtain the same modelling effect as offline to overcome the bottleneck of modelling accuracy and online rapid application. (iii) Output prediction based on h-neighborhood is finally proposed to deal with situations of less data amount or data distortion. For above three points, this paper gives a complete design framework with key formulas and processing methods. Through the verification of typical industrial HMI application of rivet plugging, the proposed translational model can achieve RMSE of 0.1231 and at least 14.3% performance improvement compared with traditional methods.