Maintenance departments are seeking ways to transform data into information and methods to discover patterns of degradation in assets. These patterns are mostly invisible, until a failure occurs. The knowledge of such underlying patterns avoids the costly failures and unplanned downtime. Maintenance practices based on pattern recognition promise greater asset sustainability and eventually near-zero breakdown. Moreover, making the invisible visible can help adjust and tune the maintenance process.
Degradation patterns are complex to retrieve because of the type of assets itself, the occurrence of failure events without any recognizable symptoms at component level; variation of cycle time due to inconsistent operation. These uncertainties have adverse effects if there are no predictive analytics and control strategies implemented. New, smarter technologies are needed to make degradation patterns more transparent. One of the key measures is an extensive network of condition detectors (temperature, strain, vision, infrared, weight, impact, etc.) that monitor the asset. All these sensors generate a lot of data.
The application of big data is a topic of big interest in several organizations. There are roughly four categories of data: configuration data which is mostly regarded as static data, that change only in response to maintenance or modification, scheduling (i.e. ERP) data which is used to describe the use of the asset, status data (i.e. OEE) which is provided through interfaces to reporting systems and operational data. In essence, many asset managers benefit from today’s capability of IT in collecting, storing, processing, analyzing and visualizing large amounts of data. To collect knowledge from that date, maintenance managers still lacks methods as machine learning, artificial intelligence, and computational intelligence and patterns recognition.
Using huge volumes of data, in combination with failure data, maintenance data, inspection data, there are several possible approaches and methods to analyze these data including, correlation analysis, causal analysis, time series analysis and machine learning techniques to automatically learn rules and build failure prediction models. These models could be applied against both historical and real-time data to predict conditions leading to failure, thus avoiding service interruptions and increasing efficiency. Additionally, analytics and models can also be used for detecting root cause failure modes.