A Primer on Predictive Maintenance
Chief Technology Officer
Predictive maintenance promises significant business and financial benefits, helping recoup the estimated $50 billion annual losses industrial manufacturers face from unplanned downtime. McKinsey Global Institute’s report on the Internet of Things expects a potential annual cost savings of between 10% to 40% resulting from the implementation of predictive maintenance practices in manufacturing operations by 2025 – an annual economic impact of up to $600 billion.
Traditional maintenance programs such as reactive, preventative or condition-based maintenance are not only costly but can also be inefficient. Some studies have shown that up to 40% of maintenance work is considered unnecessary and ties up surplus capital for extensive spare parts inventories. A small chemical plant, for example, may have over $4 million in parts on its shelves at any one time.
With this much potential cost savings, why aren’t more chemical and process manufacturers embracing predictive maintenance? One of the most significant barriers many organizations face is around one of the most critical assets companies own – their data.
Novity has a new approach to predictive maintenance, helping plant managers and operators dramatically reduce unexpected downtime with a precision of 90% or more accuracy. Let’s explore the data dilemma in predictive maintenance and how Novity can help you overcome this hurdle.
The Fatal Flaw of Predictive Maintenance
Predictive maintenance, the real-time monitoring of an active asset that uses its status and performance to predict when to perform maintenance, helps prioritize and schedule maintenance more effectively and efficiently. It can significantly reduce unplanned downtime and increase overall equipment effectiveness (OEE) – two goals on at the top of any process industry leader’s mind.
But why are more organizations not adopting predictive maintenance? If not done correctly, implementing a predictive maintenance program can quickly become expensive and unmanageable due to the large amount of data produced, analyzed and stored.
Although digital transformation is front-and-center with business leaders, most struggle to reap predictive maintenance rewards despite seeing returns from leveraging data across other business areas. However, data from process industries is unusually different from other business data we see in customer purchasing models, supply chains or fleet logistics.
Manufacturing plants are made up of a myriad of physical assets – pipes, valves, compressors, pumps, motors, bearings – all behaving differently under various conditions. However, most sensors and control systems were not designed to collect the type of data to be effective for a successful predictive maintenance program – let alone the frequency and volume of data.
We found three significant gaps in predictive maintenance deployments because of the data:
- Poor algorithm performance. Predictive algorithms in industrial applications fail to predict as many as half of the relevant failure events, and create a flood of false-positive alerts, or alerts which often arrive too late to be operational.
- High quality data is not available. To be effective, data-driven models require years of high-quality operational data with well-represented failures to inform the models.
- Predictive maintenance projects often turn into large-scale data acquisition and management problems, diverting attention from core operations.
Taken together, predictive maintenance deployments today are often expensive to set up, hard to manage, and give poor prediction results. We need to find a better way.
Predictive Maintenance the Right Way
Novity is turning the predictive maintenance paradigm upside down. We understand that process industries are unique and we help these organizations better understand and gather the correct data to generate value from day one. Our platform can reduce unplanned downtime with 90% or better prediction accuracy and with as much as three months lead time before failure, where current market leaders boast only 50 to 75%.
The Novity TruPrognostics™ engine is an always-on, real-time decision support platform showing you when critical equipment will fail, allowing you to perform the maintenance on your terms. The Novity TruPrognostics engine provides outstanding precision with a combination of machine learning and physics-based equipment models. Our best-in-class anomaly detection and remaining useful life (RUL) estimation algorithms allow maintenance staff to optimize planning and avoid lost dollars due to unexpected or excessive down time.
Novity makes predictive maintenance accessible to those who lack the massive amount of historical data required by traditional solutions using our library of pre-built physics-based models. These models capture the most common failure modes and operational attributes of critical production asset classes, enabling accurate prognosis soon after deployment in mere hours rather than the months a standard deployment would take.
We have consulted extensively with plant operators and managers to ensure that our recommendations enable operating teams to focus on achieving production goals, not large-scale data management. Plant-level dashboards provide real-time visibility to manage production threats. Additionally, an intuitive, list-based event management dashboard enables service management and staff to prioritize time easily and automatically record responses and actions.
Novity brings predictability and bottom-line impact to improve system uptime, reduce working capital tied up in spare parts inventory and avoiding the added cost of unnecessary repairs. Want to know how Novity can help you reach zero unplanned downtime? Sign up for a demo of the Novity TruPrognostics platform and see how you can start your predictive maintenance journey the right way.