AIoT-Driven Intelligent Elevator Maintenance Management: Machine Learning-Based Prediction of MTBF

Abstract

Traditional elevator maintenance and operation practices commonly face industry-wide challenges, such as delayed fault early-warning systems, homogeneous maintenance and operation strategies, and low accuracy in Mean Time Between Failures (MTBF) prediction. To address these issues, this study proposes an intelligent elevator maintenance and management system empowered by artificial intelligence and the Internet of Things (IoT), establishing a MTBF prediction framework that integrates an improved noise reduction algorithm with a hybrid machine learning model. The research leverages a multi-source sensing network to achieve comprehensive collection of elevator operational status data; utilizes an edge-cloud collaborative architecture for hierarchical data processing and model inference; and employs digital twin technology to analyze fault evolution mechanisms and optimize maintenance strategies. This paper employs a cosine-based optimized variational modal decomposition algorithm for adaptive noise reduction of industrial time-series signals, constructs a CNN-BiLSTM prediction model incorporating an attention mechanism, uses the Particle Swarm Optimization (PSO) algorithm to optimize the model's hyperparameters, and applies the Weibull distribution to calibrate the reliability of the prediction results. To tackle the challenges associated with limited elevator datasets and cross-model adaptability, this study introduces transfer learning and an unsupervised health index pre-training method to enhance the model's generalization capability. Experimental results based on real-world measurement datasets from commercial complexes demonstrate that the proposed model achieves an average absolute percentage error of only 4.27%, representing a significant improvement in prediction accuracy compared to standalone time-series models and convolutional neural network models. Practical engineering applications confirm that this intelligent maintenance system can effectively reduce elevator fault frequency, shorten downtime, and substantially lower operational and maintenance costs, providing reliable technical support for the intelligent lifecycle maintenance of elevators.