The deterioration in global energy markets that started in late 2014 forced the Oil and Gas industry to seek new ways to cut costs. During this period, the focus shifted to operational excellence. Industry executives recognized the economic promise of digitalization, and Maintenance 4.0 was incorporated into strategic plans.
At the same time, in practice, most Oil and Gas companies are laggards from the perspective of infrastructure and readiness for the widespread adoption of Machine Learning-based Predictive Maintenance.
The underlying challenges facing the industry – costly unscheduled downtime and an aging infrastructure – are long-term in nature and affect downstream refining activities. They are not affected by fluctuations in commodity pricing.
This article provides a framework for companies in the Oil and Gas sector to scale deployment across their organizations.
Higher Commodity Prices Do Not Hide Systemic Problems
It can be argued that relatively high commodity prices can hide the industry’s systemic problems. From the perspective of asset maintenance, research studies indicate some underlying trends that should concern the industry. First, it has been estimated that the aging infrastructure in Oil and Gas fields accounts for 50% of maintenance projects. Second, over 40% of respondents to a Honeywell survey admitted that they used equipment much harder than they should. The confluence of these factors hits the bottom line: According to GE, unscheduled downtime costs offshore O&M operators an average of almost $50 million a year based on 2016 commodity pricing. In some cases, the cost can exceed $88 million.
To what extent is the industry prepared to adopt Industry 4.0-based Predictive Maintenance? At best, the outlook is mixed. For instance, according to the IFS Digital Change Survey, only 19% of Oil and Gas companies consider themselves “advanced” in leveraging digital transformation. This is compared to 39% for Construction & Contracting and 29% for Manufacturing. Across all industries, 31% are considered advanced.
Framework to Jumpstart Machine Learning Based Predictive Maintenance
Based on our experience working with several Oil and Gas organizations, an industry-wide misconception exists that a dearth of talented Big Data scientists and engineers is preventing widescale deployment of Machine Learning based Predictive Maintenance solutions. This concern is also reflected in PwC’s 21st CEO Survey: Key findings from the oil and gas industry. Two-thirds of O&M CEOs surveyed indicated that they were concerned about the availability of digital skills in their workplaces. The reality is that the industry is facing an overall employee shortage. It is unrealistic to base plans for scaling Industry 4.0 on the recruiting of sought-after Big Data professionals, especially because the industry will be grappling with a shortage of between 10,000 and 40,000 petro-technical professionals by 2025.
The framework I outline below is based on successful best practices across multiple industries.
Organizational Alignment: The Core Enabler of Predictive Maintenance
In the last couple of years, Industrial IoT has been a catalyst for the convergence of traditional Information Technology and Operational Technology. To scale IIoT [didn’t we agree to replace IIoT ? ] Predictive Maintenance, Oil and Gas companies must formalize organizational changes.
Given the strategic importance of Industry 4.0 and Big Data, we expect more Oil and Gas operators to appoint Chief Digital Officers or CDOs. The scope of the CIO’s role is likely to expand, as security, data networks, etc. are basic elements of Industrial IoT. The COO will add new areas of responsibility, including the use of Machine Learning in maintenance activities.
As new organizational structures evolve, it will become necessary to formalize new and existing roles, reporting structures and KPIs. One idea is to base compensation goals for COOs and CIOs on their ability to align and achieve joint objectives.
A core pillar of Industry 4.0 Maintenance is that organization-wide standards and processes must be harmonized. Although change does not happen overnight, the starting point is organization-wide common practices.
Incorporate Auto-ML in Machine Learning Applications
The challenge involved in recruiting Big Data professionals is not unique to the Oil and Gas sector. Automated Machine Learning (AutoML) has become an increasingly important field within the Machine Learning community because it addresses the skillset shortage.
AutoML automates many time-consuming and repetitive Machine Learning tasks. It is a continuous process that incorporates feedback from the data scientist on the pipelines that have been used to date. It adapts the optimization process of the Machine Learning Pipeline in real time and provides recommendations for model, process and hyperparameter selections.
AutoML is applied primarily to three processes within the Machine Learning Pipeline:
Why is AutoML critical for scaling Industry 4.0 Predictive Maintenance? Industrial producers cannot implement Machine Learning predictive maintenance solutions across multiple facilities by hiring Big Data scientists or engineers to perform analytical functions. The supply of highly skilled Big Data professionals in the labor market simply cannot meet the demands of the Oil and Gas sector.
Data Governance: Prerequisite to Scale Predictive Maintenance
According to the Data Governance Institute, “Data Governance is a system of decision rights and accountabilities for information-related processes, executed according to agreed-upon models which describe who can take what actions with what information, and when, under what circumstances, using what methods.”
Process industries such as Oil and Gas companies generate sensor data that is analyzed by Machine Learning algorithms. Although it would be a stretch to define data as the new oil of the 21st century, the ability to implement Machine Learning Predictive Maintenance depends entirely on access to clean, real-time data.
McKinsey & Company estimates that 99% of data that Oil and Gas operators generate is not used. Some of the reasons that McKinsey provided include the fact that data integrity is not maintained, data is not streamed/stored, and data is not analyzed. Many turf wars have been waged over data ownership and access between Operational Technology and Information Technology groups.
The first step is that senior management must set a company-wide priority to establish a Data Governance framework. The processes within the framework should support the end goal of scaling a Predictive Maintenance solution across an entire Oil and Gas operator’s asset base.
Summary and Conclusion
We are past the point of convincing Oil and Gas operators to adopt Industry 4.0 Maintenance practices. Industry executives recognize that the future is digital. However, several building blocks spanning organization, technology and process are necessary to lay the foundation for Machine Learning Predictive Maintenance.