Online feature-based multisensor object detection system for bucket-wheel excavators

Abstract

This paper proposes the development of a monitoring system to be applied for a production system (bucket-wheel excavator). The system is based on a multisensor-based object detection system. The main objective of the detection system is to obtain—in real time—reliable decisions on the presence of large stones in the transported overburden. The highest possible detection rate and the lowest possible false alarm rate should be achieved to avoid disturbances or failures of the production process. Due to the complexity of the considered production system, different physical effects are taken into account. The detection system consists of two detection modules (acceleration and laser scanner module), a plausibility module (weightometer module), and a fusion module. The acceleration module consists of five acceleration sensors. The acceleration signals are individually undergone preprocessing, feature extraction, and classification processes. The preliminary decisions of different sensor channels are fused to obtain statements about the presence of large stones. In the laser scanner module, the signal is prefiltered, filtered, validated, and classified in order to detect excavated stones. The weightometer module is based on two load cells signals. It is developed to approve the plausibility of the positive statements of the acceleration module. The fusion module is developed in order to synchronize and combine the output statements of different modules to obtain the production system state with respect to the presence of an object. The detection system is developed based on the acquired knowledge from the analysis of the production process and the analysis of the acquired data during the production process. The designed system has been implemented using standard industrial hardware. The testing results to be reported show that the system requirements can be fulfilled.