Artificial Intelligence (AI) has transformed our world. It has made our day to day lives more efficient and propelled us to rethink how we integrate information, use and consume data, solve business problems and so on. With a strong track record of unlocking value across a range of sectors, AI presents significant opportunities for process industries such as oil & gas, mining and chemical manufacturing. However, these industries have been slow to embrace the technology and leverage its full potential. Studies show that by applying the lessons that have worked in other heavy industries, companies in the process industry can generate gains of 15% or more in efficiency, throughput, reduced waste, and other metrics1.
At this point the real question around AI in the process industry isn’t “if” or “when”, but “how” and “what first”. The key is to start small, one unit or workstream at a time and then apply the learnings and scale up enterprise-wide
In practice there isn’t one single path that gets a company, division, or site to a fully AI enabled future. You can’t even whittle it down to 2 or 3 different standard options, in fact there are many different paths that can be taken, but all of those paths in reality represent different combinations of the relatively standard set of options. Before we get into answering “how” and “what first”, let’s start by taking a look at some of the more common options for steps along that path.
Complex Troubleshooting: At its core, AI is a tool for complex analysis of data allowing for identification of correlations that are often not apparent to people. Utilizing this correlation capability, AI can be used to support complex troubleshooting activities providing an early entry point for AI into the organization.
Predictive Asset Care: The promise of predictive analytics; whether through AI or model-based approaches, unlocks the potential to shift the base approach to managing asset reliability from a reactive basis to a proactive one. Beyond maintenance efficiency, the ability to detect impending failures days, weeks, or even months in advance opens up a plethora of possibilities from supply-chain optimization, to predictive feedstock and product management, to more effective supply chain management, etc…
Early Event Detection: Without the constraints of typical model-based approaches to predictive analytics, AI more easily expands into process condition analytics. While the full extent of this capability will play itself out in optimization efforts, there is a more imminent application in providing early warning of process conditions that could potentially lead to abnormal operating conditions, which in turn could lead to quality issues or even process safety event. This early warning, coupled with embedded experiential-based guidance and corollary visualization allows for identification and mitigation of abnormal process conditions before they progress into process safety events.
TAR Cycle Optimization: Turnaround cycle management is key in all production facilities. The ability to maintain asset integrity while minimizing the impact of lost production and the inflated cost structure that often accompanies TARs can provide a significant competitive advantage. AI can support TAR Cycle optimization in a number of ways from simply allowing for major maintenance activities to be done in conjunction with needed repair work thanks to early warnings to allowing for better risk analysis when adjusting the timing of TARs.
Predictive Optimization: This represents the Holy Grail of AI for the process industry. Played out to its fullest potential, AI can guide on-line optimization efforts either in conjunction with Advanced Process Controllers or directly through the DCS. Utilized in this fashion, AI would proactively drive optimization across an entire facility balancing the various competing priorities.
Sustainability: The key benefit in utilizing AI as part of an overall sustainability strategy is the ability to effectively model a significantly larger set of independent variables and thereby monitor and optimize around much more complex view of the process and the associated environmental conditions.
In the right hands, AI can future-proof the process industry
While not all inclusive, the application options mentioned above provide a strong business case for the adoption of AI. Over the next decade the combined business case for AI will drive its adoption through-out every aspect of facility operations. So going back to the original “what first” question, the answer depends on what the organization is ready for and what problems it is trying to solve today. That said, when evaluating optional first steps, look for the easiest steps to start with. Take the ones that not only provide immediate value but help acclimate the organization to the power of AI.
A lot of organizations are pursuing advanced analytics for the promise of revolutionizing their future operations. While this is not necessarily a mistake, this approach makes the change management aspect much harder because it requires such a big step in changing how people work. Further, it front loads the technical infrastructure requirements meaning a lot of work needs to be done before the employees on the operational frontlines see any real benefit which further exacerbates the organizational change challenges. Given that the technical challenges with data lakes, data security, etc… are fairly well understood at this point, the real “how” question comes down to managing the organizational change elements of implementing advanced analytics as those often present the most significant barriers. As you further unpack the answer to “how”, you start focusing on pacing implementation to allow the organization to acclimatize to the changes to how work will get done.
For those of us working in the process industries, most of our work processes and practices are aligned to a reactive world. A shift to a proactive, predictive one means a significant change in approach and requires a shift in the core comfort zone of how to operate. So, when answering the question “what first”, look for small steps that deliver immediate value while helping employees build trust in the technology and its predictive capabilities. Remember that AI is a journey and not a single step, and sometimes the best way to predict the future is to start by better understanding the past.