Digital Transformation with Predictive Maintenance Drives Cost Savings

技术

分析与建模 - 机器学习
分析与建模 - 预测分析

适用行业

石油和天然气

适用功能

维护

用例

预测性维护

服务

数据科学服务

挑战

The customer, a diversified energy company with operations in refining, marketing, midstream, chemicals and specialties, had experienced three previous failures of a hydrogen compressor resulting in millions in production losses and additional maintenance costs. The company had begun its own digital transformation initiative that uses big data, machine learning and artificial intelligence (AI) to drive cultural change in the organization. As part of the initiative, they were investigating predictive maintenance. The customer decided to organize a competitive bakeoff, trimming an initial list of ten predictive analytics vendors to a handful of finalists. Ultimately, AspenTech was chosen as the sole vendor to execute an online pilot project.

关于客户

The customer is a diversified energy company with operations in refining, marketing, midstream, chemicals and specialties. They operate more than a dozen refineries in the U.S. and Europe with a total capacity of over 2 million barrels of crude oil per day. The company had begun its own digital transformation initiative that uses big data, machine learning and artificial intelligence (AI) to drive cultural change in the organization. As part of the initiative, they were investigating predictive maintenance.

解决方案

Aspen Mtell was implemented as the predictive maintenance solution. A hydrogen compressor in one of the refineries had multiple historical ring and piston failures costing over $250 million USD across just 3 events. Aspen Mtell was able to provide notification of pending failures over 35 days in advance. With that amount of warning, the plant could have scheduled the shutdown at a more opportune time within the 35-day window, reducing downtime by as much as 8 days. They would also have saved over 30 percent on the repair costs by planning work in advance. The combined production and maintenance savings from these three events alone would have been more than $75 million. Seven anomaly agents and four failure agents were created for the compressor. These agents watched for a range of failure types, including piston and piston ring failures (38 days lead time), valve failures (24 days lead time) and lubricator failures (32 days lead time).

运营影响

Aspen Mtell predicted a compressor failure 35 days in advance, allowing the company to avoid an emergency shutdown and meet production goals.
Having early failure predictions provided time to plan repairs and adjust scheduling and production.
The customer recognized the importance of Aspen Mtell’s ability to combine the mechanical view with the process view to find the earliest signs of failure.

数量效益

Reduced maintenance costs: planned maintenance has a savings potential of 30 percent over emergency maintenance
Minimized production losses by planning the shutdown: $30M USD potential savings