Performance evaluation of elevators using a novel hierarchical softmax regression model

Abstract

Elevator performance evaluation plays a guiding role in improving passenger experience and directing the maintenance work. Traditional elevator performance evaluation studies mainly depend on defect data or pertinent norms while ignoring the operation performance. In order to solve these issues, a new method for performance evaluation using hierarchical softmax regression (HSR) based on the operating signal of the elevator measured by sensors is proposed in this paper. By analyzing the vibration in the time and frequency domains, the elevator status can be monitored. Firstly, signal characteristics are studied according to the principles of evaluation indexes. Then, an index system that can characterize elevator performance with practical significance from various aspects such as vibration signal and car operation acceleration is established. Furthermore, an improved weight assignment technique is applied to decrease the dimensionality of the features, which considers data information and knowledge. Finally, an evaluation model is developed and tested for the case using a substantial quantity of data on elevator operations. Unlike traditional methods of manually evaluating performance according to standards, the proposed method aims to automatically evaluate performance based on operating conditions. The effectiveness of the proposed method is verified using an elevator dataset. Through comparison experiment, it is demonstrated that the proposed method achieves better results in terms of rationality.