Data-driven Predictive Maintenance: a paper making case Davide Raffaele Mondi Group Marxergasse 4a 1030, Vienna, Austria davide.raffaele@mondigroup.com Guenter Roehrich Mondi Group Marxergasse 4a 1030, Vienna, Austria Guenter.Roehrich @mondigroup.com
ABSTRACT Condition monitoring together with predictive maintenance of bearings and other equipment used by the industry avoids severe economic losses resulting from unexpected failures, greatly improves the system reliability and allows a more efficient usage of hu man experts’ time. This paper describes a predictive maintenance system, based on a data science approach. The system was developed and tested on a single real paper machine, and then verified with multiple external validations. Results show a proper behaviour of the approach on predicting different machine states with high accuracy. Keywords Predictive maintenance, Data Science, Industry 4.0, Vibration measurements, Data Analysis, Bearings, Condition Monitoring. 1. INTRODUCTION Predictive maintenance, also known as "on-line asset monitoring", or "smart condition-based maintenance", involves the intelligent monitoring of equipment to prevent future failures and has received growing attention in research. In the past decade, predictive maintenance has advanced from relying on visual inspection methods to automated ones that leverage advanced signal processing techniques such as pattern recognition, machine learning, neural networks, and fuzzy logic. These automated methods offer viable solutions for various industries by detecting and collecting sensitive information from equipment, where human observation may fall short [1] [2]. Integrated sensors and predictive maintenance can work together to prevent unnecessary equipment replacement, reduce machine downtime, pinpoint the root cause of faults, and ultimately save costs and improve efficiency. Predictive maintenance shares similarities with preventive maintenance, as both involve scheduling maintenance activities in advance to avoid machine failures. However, unlike traditional preventive maintenance, predictive maintenance schedules activities based on data collected from sensors and analysed by algorithms. [3], [4] [1]. Traditionally, predictive maintenance aims to schedule interventions on a machine based on health condition predictions derived from high frequency data collected by sensors. However, in the paper industry, paper machines require to stop production at regular intervals for technical reasons (e.g. felts changes), which creates a great opportunity for preventive maintenance actions. In this context, predictive maintenance is no longer a regression problem: when to schedule maintenance? It becomes a classification problem: what parts of a machine should be exchanged at a given point in time? We focus on one type of machine part: rolling-element bearings. Bearing products are important components of paper machine as
their break can cause significant production losses and even damages to the machine. The problem is how to predict the state of a bearing using vibration data. In predictive maintenance three kinds of approaches can be distinguished [1]: 1) Data-driven approach, 2) Model-based approach, and 3) Hybrid approach. While the data-driven approach uses historical data to learn patterns and system behaviour. The model-based approach is expert-driven and has the ability to incorporate physical understanding of the target product, relying on the analytical model to represent the behaviour of the system. A hybrid approach data combines data- driven as week as model-based approach is also found in the literature [5] . With the exponential increase of data being collected and available in production environment, the use of data-driven approach is increasing [6], [7] and will be the focus of this paper. The authors intend to build upon the aforementioned ideas and present a method for early-stage detection of degradation in bearings. To increase efficiency and prevent downtimes it is crucial to decide which bearings to replace at a given point in time. The contributions in this paper primarily address the application of data science approaches to real-world data related to paper machines on the field, the high level of accuracy on predictive state of the bearings and, moreover, the specificity of the application for the papermaking industry. This paper is organized as follows: Section 2 gives an overview of predictive maintenance approach used in literature. Section 3 provides an introduction to domain knowledge of bearings and their faults modes. Section 4 provides a sound foundation for the methodology followed in this paper. Section 5 demonstrates the experimental results based on the real use case under which predictive maintenance has been implemented and tested. Finally, Section 6 provides a reflection on the results that have been achieved and compares them to the current state of the literature. 2. RELATED WORK Paper industry has an extensive literature on data-driven methodologies, but it rather focusses on virtual sensors [8] [9] [10] and not on predictive maintenance. For this reason, the authors intend to hold the rest of the section industry-agnostic. Bearing degradation modelling methods can be categorized into two groups: continuous degradation, focused on building a single model to capture the degradation, and discrete degradation stage models, typically classification models that predict degradation changes over time.
Made with FlippingBook flipbook maker