Within manufacturing processes, faults and failures may cause severe economic loss. With the vision of Industry 4.0, artificial intelligence techniques such as data mining play a crucial role in automatic fault and failure prediction. However, due to the heterogeneous nature of industrial data, data mining results normally lack both machine and human-understandable representation and interpretation of knowledge. This may cause the semantic gap issue, which stands for the incoherence between the knowledge extracted from industrial data and the interpretation of the knowledge from a user. To address this issue, ontology-based approaches have been used to bridge the semantic gap between data mining results and users. However, only a few existing ontology-based approaches provide satisfactory knowledge modeling and representation for all the essential concepts in predictive maintenance. Moreover, most of the existing research works merely focus on the classification of operating conditions of machines, while lacking the extraction of specific temporal information of failure occurrence. This brings obstacles for users to perform maintenance actions with the consideration of temporal constraints. To tackle these challenges, in this paper we introduce a novel hybrid approach to facilitate predictive maintenance tasks in manufacturing processes. The proposed approach is a combination of data mining and semantics, within which chronicle mining is used to predict the future failures of the monitored industrial machinery, and a Manufacturing Predictive Maintenance Ontology (MPMO) with its rule-based extension is used to predict temporal constraints of failures and to represent the predictive results formally. As a result, Semantic Web Rule Language (SWRL) rules are constructed for predicting the occurrence time of machinery failures in the future. The proposed rules provide explicit knowledge representation and semantic enrichment of failure prediction results, thus easing the understanding of the inferred knowledge. A case study on a semi-conductor manufacturing process is used to demonstrate our approach in detail. The evaluation of results shows that the MPMO ontology is free of bad practices in the structural, functional, and usability-profiling dimensions. The constructed SWRL rules posses more than 80% of True Positive Rate, Precision, and F-measure, which shows promising performance in failure prediction.

中文翻译：

---

将编年史挖掘和语义相结合以进行制造过程中的预测性维护

在制造过程中，故障和失败可能会导致严重的经济损失。在工业4.0的愿景下，诸如数据挖掘之类的人工智能技术在自动故障和故障预测中起着至关重要的作用。但是，由于工业数据的异构性质，数据挖掘结果通常缺乏机器和人类可理解的知识表示和解释。这可能会导致语义差距问题，这代表从工业数据中提取的知识与对用户的知识的解释之间的不连贯性。为了解决这个问题，基于本体的方法已被用来弥合数据挖掘结果和用户之间的语义鸿沟。然而，目前只有少数几种基于本体的方法可以为预测性维护中的所有基本概念提供令人满意的知识建模和表示。而且，大多数现有的研究工作仅集中在机器的运行状况的分类上，而缺少对故障发生的具体时间信息的提取。考虑到时间限制，这给用户执行维护动作带来了障碍。为了解决这些挑战，在本文中，我们介绍了一种新颖的混合方法，以促进制造过程中的预测性维护任务。提议的方法是数据挖掘和语义的结合，其中使用历史记录挖掘来预测受监视工业机械的未来故障，制造预测维护本体（MPMO）及其基于规则的扩展用于预测故障的时间约束并正式表示预测结果。结果，构建了语义Web规则语言（SWRL）规则，以预测未来机器故障的发生时间。所提出的规则提供了明确的知识表示和故障预测结果的语义丰富，从而简化了对推理知识的理解。通过对半导体制造过程的案例研究来详细说明我们的方法。结果评估表明，MPMO本体在结构，功能和可用性配置方面没有不良实践。构造的SWRL规则具有超过真正率，精度和F测度的80％，