by Mario Montag

Asset-intensive organizations of all sizes and levels of manufacturing maturity are unleashing the power of predictive analytics to gain significant improvements in asset reliability.

As the wealth of information generated by sensors and machines in asset-intensive organizations increases, the opportunity for data-driven optimization drastically increases as well. Utilizing predictive analytics within asset-intensive industries presents itself as a nonintrusive method of predictive maintenance. Until now, data overload had served as a challenge and a threat to decision clarity and the ability to provide accurate, actionable information. According to past studies, an effective predictive maintenance program can result in a savings of eight percent to 12 percent over a program simply utilizing preventive maintenance strategies1. Organizations should begin planning and executing predictive analytics initiatives because the benefits are significant, they are easy to get started and they enable forward-thinking organizations to gain a competitive edge.

BENEFITS OF PREDICTIVE ANALYTICS

Although the phrase predictive analytics represents a wide variety of attributes from statistics, mathematical modeling and simulations, data mining and machine learning, it can be summarized as a method to predict the likelihood of future events or the attributes associated with such asset failures. Such a strategy can capitalize on an organization’s “big data” by analyzing any available historical data to forecast future events at up to real-time speed in an effort to enhance future operating procedures. Whether predicting asset remaining useful life (RUL), probability of failure, conditional trending, or any other arbitrarily defined feature, all will result in elevated efficiency and effectiveness throughout daily processes, which translates into reduced maintenance costs and increased revenue.

Some organizations see the value of being able to predict future asset failures as the Holy Grail, but opt to start with enhanced root cause analysis using the same statistical methods to predict future events. The ability to analyze many different factors, like machine sensor data, historical failures, weather, employee training, preventive maintenance records, OEM manufacturers and any additional factors influencing asset reliability, provide a unique perspective in gaining additional visibility into why specific failures took place in the past. This root cause analytical exercise builds a strong foundation for the deployment of predictive analytics.

Predictive analytics serves as another tool in the predictive maintenance toolbox, in addition to vibration, infrared and oil analysis, as a few other examples. Depending on a facility’s reliance on a reactive maintenance approach, implementation of a fully functional predictive maintenance program can result in:

1. Savings of 30 to 40 percent;
2. 10-fold return on investment (ROI);
3. Maintenance cost reduction of 25 to 30 percent;
4. Breakdown elimination of up to 75 percent;
5. Reduction in downtime from 35 to 45 percent;
6. Increase in production of 20 to 25 percent.

Utilizing an array of statistical models with varying combinations and contexts of predictors (e.g., indicator readings), and the results delivered in a simple, yet actionable format, a manufacturer has the ability to:

1. Dispatch a technician with a preventive maintenance work order;
2. Ensure repair parts are in inventory prior to downtime;
3. Shift from reactive to preventive maintenance on an as needed basis;
4. Increase operational efficiencies and finished goods output;
5. Increase safety;
6. Save time and money.

Data supporting the use of predictive analytics in the workplace finds that predictive analytics is the future of asset reliability optimization. Organizations have been deploying traditional predictive maintenance processes and tools for decades. The historical lack of predictive analytics in asset maintenance departments is due to the lack of knowing how to get started and simple deployment solutions. New technologies and the growing trend in predictive analytics are making it much easier for organizations to start predictive analytics initiatives.

HOW TO GET STARTED WITH PREDICTIVE ANALYTICS

One of the main reasons for implementing predictive analytics in the plant is to positively affect reliability centered maintenance (RCM) initiatives. Understanding the intricacies and capabilities of predictive analytics is vital to the movement towards RCM. Predictive analytics is one of the main steps within RCM, but organizations don’t know how to deploy it or get started.

Close to 70 percent of asset reliability initiatives do not achieve the expected results. Similar figures are found in large-scale information technology initiatives, like enterprise resource planning (ERP) implementations. Predictive analytics initiatives also can be another statistic in projects that never achieve the desired or promised results. Therefore, it is strongly recommended that organizations start small with a few key assets in one or two facilities to understand the deployment process and achieve actual results using their own data and equipment. Once success is achieved in early proof of concept (POC) initiatives, it is easier to build the required buy-in from senior stakeholders with the support of a clear business case and ROI.

An organization can take a variety of paths to achieve predictive analytics results. The first possible solution can be made internally by hiring a team of data scientists and purchasing the required statistical software and IT big data infrastructure. Another option is for organizations to purchase enterprise-wide software licenses with services to monitor and manually create actionable reports using predictive analytics. This is often times the most expensive option. The third and simplest deployment approach takes the form of acquiring analytics as a service (AaaS) from niche players in the field of predictive analytics for asset maintenance.

Regardless of which choice is selected, it is important that data is captured into databases and maintenance processes are documented accurately and consistently. The better the data, the better the results produced by the statistical (predictive) models. Not all data scientists are the same, so domain expertise is another important factor for achieving desired results with predictive analytics.

Predicative analytics is providing great results to reduce asset failures, improve reliability and reduce maintenance costs. Companies want to tap into the extensive amounts of maintenance and asset health data being captured on a regular basis that are not being fully leveraged to improve reliability and operational performance.

DON’T GET LEFT BEHIND

As machine to machine (M2M) deployments and the Internet of Things (IoT) phenomena continue to grow, it is safe to assume predictive analytics will follow in close pursuit. These two terms refer to the network of physical objects accessed through the Internet, specifically communicating with devices of the same type. Predictive analytics has been successfully applied in financial markets and consumer purchasing (marketing) trends. This technology and capability has started to expand the deployment of techniques to improve asset reliability in heavy machinery and manufacturing equipment. Industry-leading companies are seeing results as high as 90 percent in accuracy to predict asset failures days in advance, which help move them from preventative to predictive maintenance.

Predictive analytics can provide manufacturers with daily predictions or probabilities of failure, or downtime, for specific assets. The system must be able to take complex data as inputs and produce solutions that are concise and easy to interpret for maintenance crews who are actionable on the ground. The ability to provide an actionable plan independent of analytical training is unique and tangible.

References:

1. DoD. “RCM.” Reliability Centered Maintenance. N.p., 10 Nov. 2012. Web. 06 Nov. 2013.