MAY-JUN 2017

Issue link:

Contents of this Issue


Page 18 of 53

INTECH MAY/JUNE 2017 19 PROCESS AUTOMATION long as the data is correlated in time, a pattern-recognition engine can ex- tract useful characterizations of state. This type of data-driven model does not replace the need for theory-based models that offer a more precise char - acterization of behavior, but they of- fer a powerful additional tool to the operations team. How can classification be predictive? If classification is simply a way to char- acterize the state of some entity at a particular time, how can it ever be pre- dictive? It is true that an individual clas- sification of a condition state is not a prediction. Some conditions, however, are precursors of other states that are yet to come. The classic example is a down- time condition in a machine. In almost every case, a machine will start exhib- iting some changes in behavior before it erodes into a condition that requires downtime. Identifying these early states is how classification can be predictive. Material processing example A global leader in mineral production faced a situation like the one described in this article. Investments in instru- mentation and data collection were producing large volumes of operation- al data, but efforts to turn this data into meaningful improvements in opera- tional efficiency were falling short. The production line experienced fre- quent, unexpected downtime due to variations in raw material that adversely affected a critical process line machine. These downtime events lasted anywhere from two to 24 hours per occurrence, and they negatively impacted revenue and increased the cost of production. Data, in the form of motor currents, temperatures, valve settings, and stoi- chiometric measurements, was collected from the process line, stored in a process historian, and made available to the op- erations team through dashboards and other means. The thresholds, rules, and engineering-based models in use were, however, unable to reliably identify con- ditions leading to the downtime events. To solve this problem, a pattern-rec- ognition engine was installed and inte- grated with the plant's process histori- an. Members of the process operations team completed the following tasks in approximately three weeks: l Configured a data stream for pattern recognition. Here there was a single entity corresponding to the material processing line and seven signals corresponding to select motor currents, temperatures, and valve settings along the line. The team identified a window of historic data to learn from. This window was chosen to include example periods of known condi- tions, such as the downtime event. l Created a model from raw patterns in the seven selected signals from a seg- ment of the provided history window. l Provided a few examples in the his- tory window of known downtime events and periods of normality. l Identified possible patterns that could indicate a bad raw material condition leading to the downtime event. l Created a new model using the provided labels. l Tested the updated model on other parts of the history window data, and confirmed that the model detected bad raw material conditions up to 12 hours in advance of downtime events. l Set up notifications in the historian to alert operators of a bad raw material condition. l Turned on live monitoring of the data stream. The condition stream produced by the pattern-recognition engine could provide very early warnings of a previ - ously hidden bad raw material condi- tion. This awareness let the operations team take corrective actions and avoid many of the costly downtime events that had plagued them previously. Process data analysis Industrial and manufacturing operations data analysis represents a different type of "big data" challenge than those faced in e- commerce, social media, search, or other domains. Process data analysis is a long- tail situation: Data volumes are extremely large, but there are many focused, "small" problems that need to be solved as op- posed to a short list of "big" problems. Process data analysis requires a highly scalable approach that puts capabilities in the hands of subject-matter experts and that facilitates quick wins and incremen- tal growth. Pattern recognition proves to be a reliable method of analyzing big data by leveraging existing assets (i.e., tribal knowledge, operational data stores), pro - viding context to events, and uncovering paths for process optimization. n ABOUT THE AUTHOR Greg Olsen, PhD (greg.olsen@falkonry. com), is the senior vice president of prod - ucts at Falkonry, a Silicon Valley company that helps the Global 2000 accelerate continuous improvement of production operations through advanced pattern rec - ognition AI. He has more than 20 years of experience bringing innovative software products to market, including cofounding two software companies. View the online version at By Joe Alford, InTech Editorial Advisory Board As noted in the article, accurate data analysis de- pends on valid data sets, and this should be an important consideration in preparing for analy - sis. Many excellent data analysis algorithms and programs are available (e.g., neural nets, PCA, partial least squares, statistical regression, and feature extraction). Obtaining value from using these techniques depends on proper data set collection and preparation, which involves iden - tifying and eliminating data outliers, estimating missing data, aligning data, timewise, with other data (e.g., lab data), combining discrete and continuous data, normalizing data, and ensur - ing consistent context of data collected. Many industrial data mining efforts have been nega - tively affected by failure to properly choose and prepare data sets, and resource-intensive data preparation requirements can limit the time spent data mining. Observation

Articles in this issue

Links on this page

Archives of this issue

view archives of InTech - MAY-JUN 2017