The paper presents the diagnostic procedure for the ratchet mechanisms' fault detection based on the acoustic signal analysis. The diagnostic framework was proposed, consisting in three steps. First, the symptoms are extracted from audio recordings using the proposed measurement system. Next, training and testing data sets, representing various faults of the mechanism are created. Finally, the selected Artificial Intelligence-based classifier is used to extract knowledge about the relation between the internal mechanism’s state and symptoms observed in measurement data. The classifier is used to automatically evaluate state of the actual mechanism. Experiments were performed to verify the system’s efficiency (especially ability to detect and locate faults related with the pawl degradation) during the analysis of the selected gears in the Bicycle Motocross (BMX). The selected classifiers were proven to be suitable for the task. The system’s accuracy close to 100% makes it a perfect tool for the analysis of real-world objects.

本文提出了基于声学信号分析的棘轮机构故障检测诊断程序。提出了诊断框架，包括三个步骤。首先，使用建议的测量系统从录音中提取症状。接下来，创建训练和测试数据集，代表机制的各种故障。最后，选定的基于人工智能的分类器用于提取有关内部机制状态与测量数据中观察到的症状之间关系的知识。分类器用于自动评估实际机构的状态。在对自行车越野赛 (BMX) 中选定的齿轮进行分析期间，进行了实验以验证系统的效率（尤其是检测和定位与棘爪退化相关的故障的能力）。所选的分类器被证明适合该任务。该系统接近 100% 的准确度使其成为分析现实世界对象的完美工具。