In today’s volatile markets, industrial organizations face ever-growing competition and economic risks. There are very few levers to drive differentiation and profitability.
These trends are a big reason why industrial firms are going digital. By ingesting and analyzing operational asset data in real time, manufacturers can gain a significant advantage by assessing and acting on identified risks, which can reduce cost, improve quality, and increase throughput.
Polinter, a petrochemical company that produces polyethylene, has achieved remarkable results in doing so. By deploying an asset performance management system including condition-based maintenance across its four largest sites, it saw:
- Preventive maintenance activities increase by 25 percent, optimizing maintenance spend
- Reactive maintenance activities decrease by 14 percent, reducing downtime
- Mechanical availability improvement for a gain of $3.7 million in additional profits over a three-year period
As shown by Polinter, implementation of predictive analytics enables real-time asset maintenance, providing significant efficiencies and benefits over traditional time- or schedule-based activities. And while asset management is nothing new, a significant influx of industrial data requires new approaches to harness and leverage rapidly growing availability of that data.
Just consider the sheer volume of industrial equipment in use today: large assets like oil rigs, pipelines, and windmills, as well as smaller, less visible assets like pumps, valves, and filter units. Not only are thousands of these assets in operation, they are complex and deployed across a huge variety of geographies and environments with variable operating parameters.
So leveraging predictive analytics requires a thoughtful asset strategy to get the most bang for the buck and ensure benefits are received.
This calls for risk assessments that determine which assets are most critical to deliver on production goals and present the greatest risk in terms of potential financial, safety, and environmental impacts. Effective deployments require organizations to first determine which equipment to focus predictive analytics on, and what’s at risk if they fail.
Intersection of High Value and High Risk
“Broadly speaking, we apply predictive techniques to provide early warning of a potential failure leading to downtime—be it in two days, in two months, or two years, the data in context will tell us you're going to have a problem,” explained Joe Nichols, VP of Product Strategy at GE Digital.
“By using risk assessment techniques, we can determine what data to collect and what analytics to apply to determine if there is a potential failure risk. Once a risk is identified, the system automatically recommends the best corrective actions to prevent failure—be it through further inspection, operational adjustments, or maintenance.”
GE Digital makes use of “digital twins” to accurately model assets using the right data, apply analytics and rules, and surface recommendations, all of which help drive highly effective asset management practices and deliver on business objectives.
The concept of digital twins has been in use for more than 30 years. NASA, for example, has run complex simulations of spacecraft for decades. Now it’s back in the limelight with IoT applications. In fact, Gartner recently listed the technology as one of the top-10 strategic trends of 2019.
“We determine which assets are most critical and define the most appropriate analytical technique from the data we are collecting,” said Nichols. “This data can be analyzed and actioned quickly for more basic challenges, or bundled into a digital twin, which is a highly structured, virtual representation of the asset that can continuously monitor and provide early warning on potential failures.”
The APM Health solution ingests and analyzes the data as it's coming in, and compares it against the risk models. In real time, the system will run analytical routines—from simple to multivariate—to identify and recommend actions.
Elements of Success for Predictive Maintenance
The APM Health solution comprises three elements—Predix Essentials, Health Manager, and Rounds.
Predix Essentials is the underlying collection of platform capabilities for managing models and asset health. It’s where asset data is collected, contextualized, and standardized—enabling a common way to see, analyze, and recommend action on this information.
The system gathers asset data in three ways: historians, edge, and direct operator input. Historians pull and store equipment information, while Predix Edge technology directly collects data via machine-connected sensors. And operators will input data as they observe or interact with the assets. One or more methods can be used to acquire and transmit the data to the application.
Health Manager leverages the aggregated data to determine and report in real time the machine status in its operating context. It detects anomalies by calculating metrics and driving rule-based exceptions. And it supports continuous surveillance as well as discrete or periodic surveillance via rounds and calibration management as shown in Figure 1.
“Based upon all the data coming together, and upon our predictive analytics, Health Manager provides one place where you can see which assets are operating correctly and if not, why not,” said Nichols. “The system makes recommendations about the actions to be taken to return the asset to health in an end-to-end process, giving the data context and making it useful.”
Rounds is a mobile application that puts Predix APM in the hands of field personnel. It enables the remote integration of equipment strategies into day-to-day operations and surveillance programs—facilitating efficient evaluation of asset performance.
While many field assets have sensors to automatically transmit data, a big population does not. This equipment requires humans to interact with the machines directly to observe, inspect, and evaluate their current condition. Rounds provides a structured process to guide field personal through risk and condition assessments, and collect that data locally and accurately while at the machine.
Data collected by the application can be transmitted in real time to the back-end system. Offline, the data is stored on the mobile device, and once connected is transmitted to Predix Essentials and Health Management tools (Figure 2).
Reducing Downtime on the Factory Floor
The Intel® and GE Digital partnership goes beyond technology integration. The two companies worked together, deploying the APM solution to fan filter units (FFU) in an Intel semiconductor factory.
In this processing environment, the FFUs are highly critical assets with the most operational impact from system failure. A contamination problem can result in a line shutdown, damaged products, and more.
Intel was able to improve factory floor operations for energy efficiency, forecasting, yield, and worker productivity. It increased FFU uptime by more than 97 percent by ordering replacement parts upon potential failure detection and proactively replacing them before the failure occurs.
Ultimately, Intel reduced unscheduled downtime by 300 percent compared to traditional manual inspection processes. The deployment demonstrated the ROI potential of an edge-computing and cloud-based IIoT predictive-maintenance solution in semiconductor manufacturing.
The APM Health solution is optimized for Intel technology, which provides an architecture that enables GE Digital to deliver an end-to-end solution from edge to cloud. As part of a broader portfolio of Predix products, APM Health can be deployed alone, or customers can add additional products as their needs evolve.
Nichols concluded, “Our collaboration with Intel is a great example of how we can partner to combine great technologies, asset management best practices, and semiconductor domain expertise to deliver highly beneficial predictive analytics and solve a broad range of customer challenges.”