Data analytics for data driven condition monitoring

Data analytics should be at the center of strategic maintenance decision making. The diversity and quality of data collected provides key intuition that drives effective decisions on complicated topics. Online condition monitoring is used to reduce time based preventive maintenance and to enable predictive maintenance. Effective interpretation of data leads to information that plant operators can turn into decisions and actions that improve operations and maintenance activities. Data analytics is the primary technique used to facilitate effective data interpretation that will generate revolutionary results. The starting point is the data. Patterns in the data are noted and observed. The patterns observed while the plant is operating under preset conditions define process states. These patterns are mathematically manipulated to highlight changes when process changes are detected. The methods that detect state changes usually rely on correlation algorithms. Statistics are used to determine if the changes in the patterns are real or caused by plant noise and uncertainty levels. Integrated tools are used to implement algorithms that form the data analytics process and automate the decision making. Operations research is necessary to understand the operational context of the data. Machine learning algorithms provide dynamic mathematical means that can understand the present state and predict the next state with a degree of certainty. It is this prediction and the associated prediction certainty that allows plant operators to make effective decisions. This paper will discuss the approach to build a roadmap that will migrate data analytic techniques into production facilities.