Artificial Intelligence Drives Improvement in Maintenance (Advertorial)

By Mike Brooks, Global Director, Asset Performance Management, Aspen Technology

My passionate belief is that technology supported by new-found data science techniques reveals truth that facilitates fundamental improvements. In the field of maintenance for chemicals, oil and gas, and mining, the use of break-fix, calendar, usage, condition, and reliability-centered maintenance (RCM) techniques has been the basis of 40 years of improvements. We have seen a definitive progression of increased machine reliability with the enhancements in inspection methodologies, including improvements due to risk assessments in RCM but at excessive skills, time, and costs. The results assure machines are more reliable: but they still break down. There are a couple of reasons why.

Experimentation and word of mouth without science drove most improvements. Historically, improving reliability fixated on when to inspect and service; with progressively more complicated ways to decide. That progression is severely limited. In 2015, ARC Advisory Group highlighted that 82 percent of assets suffer from breakdown caused by errant process behavior in a ‘seemingly’ random failure pattern.

Typically, such failures will not be found by periodic inspections. They are effectively caused by chance events in the manufacturing process that cannot be found by inspections. Such events include pumps losing feed, or cavitation, and compressors suffering almost imperceptible liquid carry-over that over months can lead to catastrophic failure. Damaging events can also include an unintentional erroneous setpoint entry that takes a machine out of its safety and design performance envelope. More recently, however, we have witnessed a sea change in terms of asset performance management (APM). The use of data analytics and artificial intelligence (AI) is transforming APM. The intent herein is not to manage the maintenance execution process but to dispatch a notice when the service and inspection is required.

AspenTech has led the way in this process, especially since the acquisition of Mtelligence Corporation (known as Mtell) in 2016. The new Aspen Mtell team began with decades of experience in maintenance and operations. Having seen the effects of all maintenance methods, the team set out to stop machines breaking! We did not build a platform and then go looking for a problem to solve. Instead, we used our domain experience to craft the Mtell application to explicitly meet reliability and maintenance needs. We insisted whatever product methodologies we prepared they must not require new skills or intense experience in process, mechanical, thermodynamic engineering or emerging statistical and data science techniques. Instead we assured the product assembly and deployment must fit within the work processes and skills already in place at our customers.

Consequently, it must not be just about technology; it is about customers being able to work with the software to deploy and maintain it fast and easily, scaling to rapidly blanket a whole operation, to sustain and maintain it over its lifecycle. In our product the intense technology is hidden on the inside with a user experience to deploy it using what they know already. They do not need lifelines from expert engineering or data science. Similar to the iPhone, the technology is all there, but it’s hidden on the inside so that even someone without specific technology expertise can use it. This is a breakthrough in the industry where other products need experts and a lot of time to implement.

In our system, we brought real science in the way we use machine learning rather than anecdotal references. That truth shows the reality from which real improvements are made. We deploy Autonomous Agents that operate every few minutes doing the work so customers do not have to. Our methodology is accessible to employees across the organization so that within hours they are productive; building Mtell Agents to protect equipment. The Mtell application bypasses all the disadvantages of engineering/model/statistical solutions and provides the latest AI/machine learning technology. Mtell also has its own proprietary process to use explicit pattern recognition to develop two types of Autonomous Agents: Anomaly Agents, with all the smarts to recognize normal machine behavior and the deviations that indicate impending failures. Inline, in real-time with one click they adapt to drift and changes in process behavior.

Additionally, unlike alternative solutions, Mtell deploys Failure Agents that measure the exact timing and unique patterns that lead to an explicit failure, Such Agents are not based on anomaly detection. As a result, they provide earlier and more accurate alerts months and weeks in advance (not just days) with exact timing of degradation to failure; along with prescriptive advice for what service is required or the operational changes to avoid a problem. Often early detection and small operational changes can eliminate the maintenance altogether. The insight gained goes beyond wear and tear on the machine to address the fundamental root causes of damage. Lastly, Mtell Agents are not restricted to models or types of equipment. Agents detect normal behavior and precise failure patterns across diverse chemical and manufacturing industries for all kinds of assets and failure modes: rotating and static machines, mobile vehicles, and process equipment such as heat exchangers or furnaces.

The ability to deploy prescriptive analytics helps to improve the earning potential of the businesses by making assets more available. Typically manufacturing facilities can expect significant gains in plant throughput which far outweigh the maintenance associated with a failure. For example, a low density polyethylene (LDPE) manufacturer experienced 27 days advanced warning of mechanical failure avoiding production losses that were 5-10 times the repair costs. Also, superior maintenance service management of planned maintenance will occur as your organization becomes comfortable with the accuracy of Mtell alerts. Chemicals, oil and gas, and mining will also see fewer safety and environmental events when extremely early warnings allow the time to plan interventions.

Lastly, a central pillar of AspenTech’s mission is digitalization of the world’s most capital-intensive industries. To increase the capability of our products for our customers, AspenTech is using the combination of domain knowledge and AI analytics. This is the powerful and necessary combination that enables manufacturing companies to identify the sources of uptime losses and margin leakage to promote the best possible outcomes, in turn enabling them to achieve operational excellence and competitive edge.