Closing the performance gap with machine learning

The performance and uptime of heavy rotating equipment play an important role in business profitability in industries such as oil & gas, refining, mining, chemicals, and power generation. One of the most important metrics in these capital-intensive industries is return on capital employed (ROCE), which depends in large part on the availability of rotating equipment. In a recent survey conducted by ARC, over 80 percent of the respondents indicated that they target 90 to 95 percent availability of rotating equipment. Machine builders and manufacturers employ a wide variety of methods to achieve this level.

Our survey goal had two primary objectives: 1) identify current industry practices to reach the target availability; and 2) identify the layers of protection incorporated in machinery for predictive maintenance, diagnostics, and failure analysis purposes.

Owner Operators Seek Productivity Improvements

Based on a combination of interviews and survey results, there appears to be a measurable gap between the capabilities of machine builders and what some leading industrial organizations are seeking. While rotating equipment suppliers are incorporating condition monitoring, predictive maintenance, and safeguarding technologies into the systems, these tend to be relatively conservative implementations. For the most part, failure analysis and predictive maintenance solutions embedded in machinery are generally limited in capability. The solutions rarely correlate information from multiple sensors. In many cases, designs that realize a machine's full potential, and optimal individual machine performance, may only be achieved through analysis of in-service data. Most machine builders/system integrators employ off-the-shelf PLCs and AC drives (some with embedded PLCs) to control and monitor the machinery. These systems simply do not have the computational capacity needed to implement advanced predictive maintenance and failure analysis algorithms. A productivity and performance gap is a challenge facing the industry today. This is becoming more accentuated as the chemical, oil & gas and mining industries are under pressure to improve ROCE with declining commodity prices.

Closing the Productivity Gap with "Machine Learning"

One of the underlying concepts of the IIoT is the use of big data and analytics to identify patterns and trends not readily visible using traditional statistical tools. This concept has been touted as "machine learning" in the industry. Ideally, a machine performance monitoring system has "machine learning" capabilities, but this a much larger concept than just embedding algorithms in the machine. It is about leveraging the computational capabilities available in the cloud and deriving new intelligence about failure modes in production equipment that correlates multidimensional, time stamped machine process variables.

Respondents were asked a question reading: "Imagine a hardware and software system capable of automatically measuring and analyzing machine performance continuously, allowing the system to learn and adapt to individual motors and predict future maintenance requirements well in advance of a motor or component failure. How compelling does this seem versus existing available solutions?"

Overwhelmingly, respondents responded positively to the possibility of machine learning systems that can adapt to system variations. Over 80 percent of the respondents found this as a "compelling" proposition. In the end, manufacturers want to use intelligent systems that can remove the human element from the decision making process. The notion of "machine learning" is a critical element of the value proposition of IIoT. It is clear that manufacturing organizations are ready to leverage analytics in the cloud capabilities that are available today.

For more information see the full report.