Headquartered in the UK, ManufacturingCo empowers leaders in the renewable energy industry. It helps them reduce asset maintenance costs, improve yield, increase uptime and ensure process quality. It provides data insights used in asset predictive maintenance. The energy companies leverage these insights to plan, maintain and sustain their operational efficiency.
But, if ManufacturingCo were to provide inaccurate insights, it could impair decisions about asset maintenance. Here’s how certain challenges affected its operations-
Manual and time-consuming fault detection
ManufacturingCo depended on a manual and lengthy process of fault detection. It used the FMECA (Failure Mode their Effect and Criticality Analysis) and FMSA (Failure Mode and Symptoms Analysis) to provide insights on the sensor vibration data. The notification alarms (installed in assets) rang without fulfilling the prerequisite of genuine fault detection. This resulted in an expensive spare parts inventory, time-consuming fault remedy process and operational inefficiency. An automated fault detection system was required to avoid undeclared asset failures and emergency shutdowns.
Unpredictable Lead time of Repair
ManufacturingCo suffered high latency periods in remedying asset faults. This happened due to a limited information generated about the nature of a fault. The fault location (Wind Turbine Generator component), cause of the fault (component looseness, misalignment or malfunction), its severity (on a scale of 1-5 where 5 suggested the highest level of severity) etc. remained unclear. They did not receive information about fault progression i.e. speed at which the fault was advancing. It would have helped site engineers prioritize faults and decide their remedy plans. There was also no information on the likelihood of a fault’s reoccurrence. Knowing the fault certainty (on a probability scale of 0-100%) would enable engineers at ManufacturingCo to predict the lead time of repair.
ManufacturingCo’s customers were unable to get a unified view of their assets at once. Every wind farm used an on-premise local system to store their operational data. It lacked to generate an umbrella view of all wind farms at once. A centralized solution would improve fault detection, speed up decision making and improve asset maintenance cycles.
Roadblocks to Machine Learning
ManufacturingCo’s decentralized software solution acted as the roadblock to implementing machine learning models for data generated on assets’ health. But the machine learning models required accurate and a huge amount of data. This data could be then used to generate actionable insights about future decisions related to predictive maintenance.
Trouble to scale up system
ManufacturingCo’s software solution had opportunities to automate predictive maintenance. But, it was challenged with scaling up its system. The main trouble for not being able to scale up was the decentralized nature of the system. A centralized system had the power to gather, store and analyze data from many sites, covering thousands of assets and offer a combined solution.
The Solution: MVP implementation of the IIoT Proof of Concept
The journey of coming up with a solution commenced with addressing the challenges. Machine Learning formed the foundation of solution- an automated fault diagnostic system. This meant replacing the age-old ‘if-then rules’ method of fault diagnosis with machine learning models. The historical and real-time data could be leveraged to increase the accuracy of fault diagnosis.
This new IIoT solution was implemented as an MVP (Minimum Viable Product). It displayed the potential to support predictive maintenance by decreasing the cost and risk of decision making associated with asset health. Today, it holds the power to connect millions of sensors globally and provide a big picture of condition monitoring.
Want to reduce equipment failure and improve performance of your assets?
Talk to our Solutions Specialist