Prognostic Health Management (PHM) is a predictive maintenance strategy, which is based on Condition Monitoring (CM) data and aims to predict the future states of machinery. The existing literature reports the PHM at two levels: methodological and applicative. From the methodological point of view, there are many publications and standards of a PHM system design. From the applicative point of view, many papers address the improvement of techniques adopted for realizing PHM tasks without covering the whole process. In these cases, most applications rely on a large amount of historical data to train models for diagnostic and prognostic purposes. Industries, very often, are not able to obtain these data. Thus, the most adopted approaches, based on batch and off-line analysis, cannot be adopted. In this paper, we present a novel framework and architecture that support the initial application of PHM from the machinery producers’ perspective. The proposed framework is based on an edge-cloud infrastructure that allows performing streaming analysis at the edge to reduce the quantity of the data to store in permanent memory, to know the health status of the machinery at any point in time, and to discover novel and anomalous behaviors. The collection of the data from multiple machines into a cloud server allows training more accurate diagnostic and prognostic models using a higher amount of data, whose results will serve to predict the health status in real-time at the edge. The so-built PHM system would allow industries to monitor and supervise a machinery network placed in different locations and can thus bring several benefits to both machinery producers and users. After a brief literature review of signal processing, feature extraction, diagnostics, and prognostics, including incremental and semi-supervised approaches for anomaly and novelty detection applied to data streams, a case study is presented. It was conducted on data collected from a test rig and shows the potential of the proposed framework in terms of the ability to detect changes in the operating conditions and abrupt faults and storage memory saving. The outcomes of our work, as well as its major novel aspect, is the design of a framework for a PHM system based on specific requirements that directly originate from the industrial field, together with indications on which techniques can be adopted to achieve such goals.

The pervasive digital innovation of the last decades has led to a remarkable transformation of maintenance strategies. The data collected from machinery and the extraction of valuable information through machine learning (ML) have assumed a crucial role. As a result, data-driven predictive maintenance (PdM) has received significant attention from academics and industries. However, practical issues are limiting the implementation of PdM in manufacturing plants. These issues are related to the availability, quantity, and completeness of the collected data, which do not contain all machinery health conditions, are often unprovided with the contextual information needed by ML models, and are huge in terms of gigabytes per minute. As an extension of previous work by the authors, this paper aims to validate the methodology for streaming fault and novelty detection that reduces the quantity of data to transfer and store, allows the automatic collection of contextual information, and recognizes novel system behaviors. Five distinct datasets are collected from the field, and results show that streaming and incremental clustering-based approaches are effective tools for obtaining labeled datasets and real-time feedback on the machinery’s health condition.

Predictive maintenance allows industries to keep their production systems available as much as possible. Reducing unforeseen shutdowns to a level that is close to zero has numerous advantages, including production cost savings, a high quality level of both products and processes, and a high safety level. Studies in this field have focused on a novel approach, prognostic health management (PHM), which relies on condition monitoring (CM) for predicting the remaining useful life (RUL) of a system. However, several issues remain in its application to real industrial contexts, e.g., the difficulties in conducting tests simulating each fault condition, the dynamic nature of industrial environments, and the need to handle large amounts of data collected from machinery. In this paper, a data-driven methodology for PHM implementation is proposed, which has the following characteristics: it is unsupervised, i.e., it does not require any prior knowledge regarding fault behaviors and it does not rely on pre-trained classification models, i.e., it can be applied “from scratch”; it can be applied online due to its low computational effort, which makes it suitable for edge computing; and, it includes all of the steps that are involved in a prognostic program, i.e., feature extraction, health indicator (HI) construction, health stage (HS) division, degradation modelling, and RUL prediction. Finally, the proposed methodology is applied in this study to a rotating component. The study results, in terms of the ability of the proposed approach to make a timely prediction of component fault conditions, are promising.

The development of predictive approaches to estimate supplier delivery risks has become vital for companies that rely heavily on outsourcing practices and lean management strategies in the era of the shortage economy. However, the literature that presents studies proposing the development of such approaches is still in its infancy, and several gaps have been found. In particular, most of the current studies present approaches that can only estimate whether suppliers will be late or not. Moreover, even if autocorrelation in data has been widely considered in demand forecasting, it has been neglected in supplier delivery risk predictions. Finally, current approaches struggle to consider macroeconomic data as input and rely mostly on machine learning models, while deep learning ones have rarely been investigated. The main contribution of this study is thus to propose a new approach that for the first time simultaneously adopts a deep learning model able to capture autocorrelation in data and integrates several macroeconomic indicators as input. Furthermore, as a second contribution, the performance of the proposed approach has been investigated in a real automotive case study and compared with those studies resulting from approaches that adopt traditional statistical models and models that do not consider macroeconomic indicators as additional inputs. The results highlight the capabilities of the proposed approach to provide good forecasts and outperform benchmarks for most of the considered predictions. Furthermore, the results provide evidence of the importance of considering macroeconomic indicators as additional input.

Predictive Maintenance (PdM) is the newest strategy for maintenance management in industrial contexts. It aims to predict the occurrence of a failure to minimize unexpected downtimes and maximize the useful life of components. In data-driven approaches, PdM makes use of Machine Learning (ML) algorithms to extract relevant features from signals, identify and classify possible faults (diagnostics), and predict the components’ remaining useful life (prognostics). The major challenge lies in the high complexity of industrial plants, where both operational conditions change over time and a large number of unknown modes occur. A solution to this problem is offered by novelty detection, where a representation of the machinery normal operating state is learned and compared with online measurements to identify new operating conditions. In this paper, a systematic study of autoencoder-based methods for novelty detection is conducted. We introduce an architecture template, which includes a classification layer to detect and separate the operative conditions, and a localizer for identifying the most influencing signals. Four implementations, with different deep learning models, are described and used to evaluate the approach on data collected from a test rig. The evaluation shows the effectiveness of the architecture and that the autoencoders outperform the current baselines.