Case study: Centralized data platforms slash maintenance hours at Cosmo Oil refineries

Key takeaways

• Centralized data platforms cut engineer data-collection time to minutes, boosting maintenance productivity.
• Digital twins enable remote monitoring, VR access, and predictive maintenance across multiple refinery sites.
• Collaborative reliability hubs streamline decision-making, knowledge transfer, and cross-site troubleshooting.

Mounting external forces inspired Cosmo Oil, a division of Cosmo Energy Group, to rethink its data management and maintenance strategies. The Japanese petrochemical company’s biggest impediment to efficiency was its highly dispersed and siloed maintenance data. Solving that would open the door to modern maintenance practices.

By implementing a new foundational data platform, refinery digital twin, centralized hub for maintenance operations, and AI, Cosmo Oil not only slashed the time required to find and analyze data, but also enabled real-time condition monitoring and predictive and collaborative maintenance across its three refineries.

Maintenance workforce shortages and declining fuel demand drive need for smarter refinery operations

Demographic and demand shifts posed a growing risk to Cosmo Oil’s competitiveness and growth. Recruiting new engineers was more difficult with Japan’s declining birth rate and fewer petroleum engineering graduates in the talent pool. Meanwhile, domestic demand for fuel oil was declining. Accordingly, Cosmo Oil sought to operate its refineries more productively and autonomously with smaller, dedicated teams per site.

A prime opportunity was improving access to vital maintenance information because its engineers were spending 70-80% of their time collecting data. For example, troubleshooting critical equipment involved collecting information such as piping and instrumentation diagrams (P&ID), facility layouts, equipment drawings, inspection records, and maintenance plans from disparate, siloed data sources – all before organizing and adding context to the data to make maintenance planning decisions.

Cosmo Oil wanted to empower its engineers and maintenance experts to quickly extract operational, maintenance, and inspection insights and make data-driven decisions. This would require rapidly consolidating information from Excel files, PDFs, historical documents, paper maintenance records, asset performance management (APM) data, and more.

Digital twin capabilities were needed to replicate the three real-world refineries as a single plant in a virtual space, with virtual reality (VR) enabling remote access to the digital twin. The envisioned process was for an engineer wearing VR goggles to enter a virtual space, touch heat exchangers or pumps, and sequentially check the associated data.

The ideal platform would:

• Centralize data management across the Sakai, Chiba, and Yokkaichi refineries
• Enable remote condition monitoring
• Support maintenance collaboration among the refineries
• Move the refineries toward predictive maintenance
• Facilitate knowledge transfer from retiring domain experts to the new generation
• Drive high operational efficiency and productivity

Building a refinery digital twin and AI-driven maintenance platform with Cognite Data Fusion

The search led to Cognite Data Fusion, an industrial data and AI platform from Cognite. Cognite Data Fusion would allow Cosmo Oil to collect and integrate millions of maintenance data points, automatically contextualize the data, and maintain a dynamic digital twin that makes maintenance information and insights immediately available to plant workers.

Additionally, using AI and machine learning for predictive maintenance and condition sensors to capture real-time wireless monitoring data, such as vibration patterns, it would be possible to predict when a defined threshold value will be breached, allowing maintenance to be planned and scheduled before failure.

The successful proof of concept (PoC) ended three months early because the data collection and automatic contextualization proved exceptionally fast. “The speed was such that if we input data in the evening, the data construction task would be completed by the next morning,” observed Mr. Kiyohide Yoshii, the Head of Maintenance Strategy and AMP Group at Cosmo Oil’s Engineering Department.

After choosing Cognite Data Fusion for its refineries, it was used in August 2023 to build a digital twin of the refinery. The company is now using AI to perform predictive maintenance on rotating machinery abnormalities, and its efforts are improving day by day. The AI identifies issues, generates alerts, and proposes options for resolution.

Extending the value of the platform is Cosmo Oil’s Reliability Center of Excellence (RCoE), a centralized hub for maintenance operations. The integrated monitoring room enables collaborative maintenance by skilled engineers at the plants. Interactive P&IDs and a dashboard dedicated to monitoring rotating machines further elevate productivity.

“By consolidating the three refineries into one virtual space, we can obtain data from Sakai, Chiba, and Yokkaichi, all from the Tokyo headquarters. This allows us to support maintenance operations for all refineries while in Tokyo or enable engineers from Chiba and Yokkaichi to collaborate on troubleshooting in Sakai,” explained Mr. Yoshii.

Operational gains from predictive maintenance, digital twins, and centralized reliability management

With Cognite Data Fusion, Cosmo Oil realizes savings at multiple levels:

• It slashed the number of work hours spent by engineers on data collection, which previously consumed 70-80% of their time.
• It reduced the effort to format data construction prompts from 1-2 hours per day in the field to 10 minutes from a desk.
• It reduced costly unplanned downtime with remote monitoring and predictive maintenance, allowing repairs to be carried out systematically at the optimal time.
• It introduced immediate data access and reduced from over an hour to about five minutes the time it took to compare and examine the data.
• It reduced reliance on the judgement of experienced engineers to detect degradation and impending failures by enabling data-based decisions.
• It provided immediate insight into field situations from a laptop, increasing the speed and precision of planning asset installations and modifications.
• It minimized the risks associated with hazardous environments by enabling VR capabilities.

“By integrating the vast volumes of data generated daily, we have constructed a comprehensive digital plant for collaborative maintenance,” explained Noriko Rzonca, Chief Digital Officer of Cosmo Energy Holdings. “This digital plant empowers us to apply AI and machine learning to predict the condition of equipment, enabling predictive maintenance. As a result, we are optimizing production processes, enhancing safety, maximizing productivity, and enabling work remotely.”

Looking forward, Cosmo Oil is considering introducing Cognite InRobot to utilize robots in on-site inspection work, furthering the company’s evolution into a more flexible and resilient organization.