The role of novel data in maintenance planning: Breakdown predictions for material handling equipment. (July 2022)

• Record Type: Journal Article
• Authors: Falkenberg, Sven F.; Spinler, Stefan
• Is Part Of: Computers & industrial engineering. Volume 169(2022)
• Publication Date: 2022-07
• Subjects: Maintenance -- Machine learning -- Reliability -- Data science; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2022.108230
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: Novel data sources are used to predictive failures of material handling equipment. A generalizable framework structures and extracts relevant predictor variables. A comprehensive study evaluates statistical learning methods for failure detection. Time and condition-based variables are combined in a robust approach. For practitioners, sensor and data collection recommendations are provided. Abstract: We build a predictive maintenance model for material handling equipment that incorporates novel data sources to forecast breakdowns. To this end, we develop a framework to structure and extract relevant predictor variables. Subsequently, we conduct a comprehensive study of statistical learning methods for failure detection. We show that the standard sensors in material handling equipment provide sufficient data to predict the majority of breakdowns ( > 85 % ). The findings are confirmed in two independent datasets and are thus transferable. Further, we provide a cost-based evaluation of those statistical learning methods and find that K-Nearest-Neighbors and Random Forest Classifier are cost-optimal. While most extant literature focuses on either time or condition-based maintenance, we suggest a more robust approach. We demonstrate that both time and condition are almost equally important. As a result, we present a prediction model that incorporates both variable types. From a managerial perspective we provide recommendations on data collection and highlight the importance to understand the cost ratio between breakdowns and preventive maintenance services.