As financial pressures in the chemical industry rise, more and more companies are looking for new ways to increase profits and mitigate financial risk through improved operational efficiency. With slim margins and volatile markets, they increasingly are turning to technology and, more specifically, data to help them optimize existing maintenance processes and operational resources.
Truth be told, chemical manufacturers have been using data from sensors embedded in high value assets for many years. However, a tremendous reduction in the cost of sensors has spawned the digitization of nearly every piece of equipment. In fact, Morgan Stanley predicts there will be nearly 75 billion connections between devices by 2020. Of course, data by itself is relatively useless. It is only when companies have the intelligent systems and infrastructure in place to take action based on the information that the true value is unleashed.
Benefiting from Predictive Maintenance
Operating in a capital intensive industry, chemical companies face big financial risks if maintenance issues arise. An effective predictive maintenance program requires the ability to monitor and present available equipment condition data such as diagnostic and performance data, and maintenance histories. Quickly compiling and analyzing this equipment data allows managers to predict when repairs will be needed and proactively meet the maintenance needs of major or critical equipment.
Historically, many chemical companies set maintenance repairs on a regular schedule (every week, two weeks, month, year, etc.). However, often repairs are not needed and maintenance is done “just in case.” With the power of in-memory computing, companies will no longer need to do these “just in case” repairs and will be able to use real time data to pinpoint the exact time maintenance is required.
Today companies take Gigabytes of data taken from their data historians, aggregate and analyze it with state of the art in-memory technology. Then they apply special algorithms typically sitting on top of such in-memory data bases to identify failure patterns and corresponding root-causes as foundation for predictive maintenance. At the most basic level, companies look at a small number of variables and using basic statistical techniques (e.g. trend analysis), try to predict and schedule the next round of maintenance.
Of course, maintenance issues are the result of many variables acting together. Variables such as temperature, running speed, pressure, and even the type of material being processed can all be factors in predicting when maintenance is needed. But how can we determine which variables act together? And at what levels do they affect the performance of the machinery? Add to the mix that not all variables are known and you can see why predictive analytics is more complicated than just predicting a time of maintenance.
Moving Toward Predictive Modeling
As a company becomes more experienced with basic predictive maintenance, the next step is to create models for even more accurate predictions. Predictive modeling involves using advanced statistical techniques to analyze multiple conditions that can affect equipment or processes. Once the model is created, actual data is plugged in and action can be taken on the resulting maintenance recommendations. Predictive models must continuously be refined to improve the accuracy of the recommended result, which requires a set of skills not normally found within traditional maintenance teams.
Getting Started with Predictive Analytics
As illustrated above, the business benefits associated with converting data into action are significant. However, one must be willing to embrace recent technology advancements in order to benefit from this data-intensive business environment. More specifically, to begin using predictive analytics for decision making, organizations need:
- the ability to store and process large amounts of data;
- in-memory processing capabilities powerful enough to analyze the information quickly and apply advanced algorithms for predictive insights;
- an integrated architecture capable of combining big data, analytics, mobile, cloud and social media on a single platform;
- a user-friendly, graphical interface to make sense of the data results and support augmented reality; and
- applications that provide real-time access to the right data at the right time.
With the infrastructure in place, operational and maintenance employees are empowered to make insightful, data based decisions that save considerable amounts of time and money. Even so, chemical companies are just beginning to understand how to operate in this new, data-intensive environment. Making operational decisions based on real-time information delivers wider margins, improves financial performance and helps meet success goals. In an environment where every dollar counts, chemical companies that can take information and transform it into innovative action will be the undisputed industry leaders.