Detecting and learning from patterns and trends in data, processes, and asset conditions allows organizations to improve their production processes and maintenance practices. The challenge plants face is how to process the massive volume of available data, seeing through the noise and uncovering actionable intelligence. The rapid growth in devices and systems that make up the Industrial Internet of Things (IIoT) is heightening the complexity of the task.
Process engineers have historically had a tough time identifying patterns in operations and production environments. Data pattern discovery can lead to both incremental and disruptive process optimization and improvement, says Richard Beeson, CTO at OSIsoft. Falkonry and OSIsoft have partnered to simplify and accelerate signal data pattern identification and condition recognition with Falkonry Service.
“With technologies like Falkonry Service that provide real-time pattern recognition across complex processes, we empower our process engineers to impact process operations through early awareness of known good or bad states and, more importantly, new or unknown operating states,” explains Beeson.
TrendMiner aims to simplify the mining of massive amounts of data generated in process manufacturing with its patent-pending pattern recognition and machine learning technologies. The plug-and-play solution delivers big data search and predictive analytics for industrial process measurement data. It connects easily to existing data sources and is designed for use by the average historian user – no data scientists or big data infrastructure is required.
Machine learning technology is revolutionizing how plant maintenance is performed. With Mtell Smart Machines, pattern recognition is used to “make machines smart.” In other words, it enables them to identify normal as well as abnormal behaviors and alert Maintenance to issues before they cause problems.
“Discovering behavioral and failure patterns is only one step in preventing process and mechanical equipment degradation,” remarks Michael Brooks, President and COO at Mtell. “The real value comes from detecting a root cause from the earliest symptoms automatically, in real-time, along with prescriptive action to stop the damage from happening in the first place.”
Monitoring DCS signal patterns allows for proactive maintenance. Scientech’s Predictive Pattern Recognition (PdP) application for the power industry uses existing plant instrumentation available through installed DCS, data historian, and other data systems for early detection of abnormal operating conditions.
“The algorithm embodied in PdP produces accurate estimates of the expected values of sensors by utilizing models that have ‘learned’ the normal operation of the equipment based on its historical performance. The estimated values are compared to the real-time values to evaluate overall equipment health,” says James Brower, manager of Scientech Plant Optimization, Curtiss-Wright Nuclear.
For power generation plant performance and condition monitoring, the EtaPRO System from GP Strategies uses sensor history and advanced pattern recognition (APR) to detect anomalies in a system or process. “All plants suffer from faulty instrumentation that can blind operators to real problems,” suggests Richard DesJardins, Vice President of GP Strategies’ Performance Engineering Division.
A real and immediate benefit of EtaPRO APR is its ability to point to specific sensors that have poor indication or have failed altogether, thus allowing maintenance personnel to focus on known problems, says DesJardins. In addition to fostering instrumentation reliability, EtaPRO combines APR with advanced vibration signal analysis to provide early warning of conditions and fault symptoms that can lead to equipment failure.
Performance improves when failure trends are spotted and fixed by reliability engineers. Inconsistent, overly general, or poorly defined work order problem codes such as “Other” obscure these trends. To make failure analysis faster and more accurate, Jeff Wahl, maintenance manager at Polar Field Services, developed TrendWords, a multilingual contextual search engine for computerized maintenance management systems. TrendWords uses contextual failure analysis (CFA) to highlight systemic failures in and across sites, facilities, systems, and equipment.
“Looking for a better way to trend work order failures led me to the idea of searching work order text in places like description fields for failure-related keywords,” says Wahl. TrendWords can query directly into the CMMS, making the data searchable to extrapolate failure trends. It works with any text, virtually without limits, including from multiple years and systems. The goal is to offer TrendWords as a service and/or a tool with training.