The oil industry involves costly assembly of equipment like oil rig engines. These systems are equipped with oil temperature, viscosity, differential pressure, and cable tension sensors to determine equipment health status.
The oil and gas sector incurs high operational costs for oil well engine maintenance. Current methods like manual analysis and inspection are time-consuming and restricted by a limited ability to detect faults at scale. This leads to a rise in maintenance overhead, causes unplanned downtime, and reduces lifespan of oil well engines.
For oil industry assets, expensive and inefficient emergency repairs cause massive cost bleeds and destabilize the system. The lack of an integrated cloud platform results in limited abilities to drive insights on assembly breakdown scenarios.
Our client from the oil industry faced operational expenditure and frequent unplanned downtime which hampered the overall productivity of the unit. Periodic inspection and maintenance of assembly are expensive and not foolproof to notice any indications of unplanned breakdowns.
We noted the following prominent issues with our clients’ assets.
- Data collection and analysis is challenging
- Reduced safety concerns for staff
- Getting insights from real-time and historical oil well data is challenging
- Lack of an integrated cloud platform
- Limited ability to notice failure scenarios at scale
- Periodic maintenance is not effective
With clarity on above issues, our client helped us to set the following objectives.
- Help technicians to note hazardous scenario for oil well assembly and take proactive actions
- Reduce the cost of unscheduled maintenance and improve assembly lifespan
- Enhance accuracy in detecting machinery failure and degrade patterns
AI-Driven Predictive Maintenance Solution To Optimize Oil Industry Assets
Cutting-edge technologies help to optimize assets and overcome the drawbacks of traditional methods used to track asset health in ways never imagined before.
After analyzing the client’s requirements, we proposed a Google Cloud Platform–based cloud Machine Learning solution. We went with Google Cloud Platform due to its secure and responsive data pipeline, edge computing and fully manageable, optimized and scalable solutions.
We created a hypothesis with the precise quality data and metrics to determine the exact breakdown scenario for oil well engines. For this, we considered different parameters such as equipment age, noise, efficiency, load, heat, vibration, etc. We combined numerous data points such as sensor data, legacy data, technician feedback, and weather data to build our model.
After transforming the raw data into features, we developed a LSTM model to predict the remaining useful life of oil well engines. This RNN model uses real-time sensor data and historical data to compute failure risk scores for oil engines. The reason to choose the LSTM model was its remarkable performance in computing RUL using three gates--input, forget, output gate and tanh function.
Our Google Cloud ML experts:
- Designed the LSTM model for predictive maintenance of oil well engines.
- Trained the hierarchical model with precise data pertaining to different breakdown scenarios and evaluated its performance on the holdout set.
The model is trained based on the following data points:
With this digital transformation, the trained model offers predictions on the RUL of oil rig engines backed by explainable decisions. These predictions help plan optimized maintenance schedules and engine/part replacement strategies with the primary goal of reducing unplanned downtime
During the project delivery phase, we empowered technicians with a powerful dashboard to know the health status of an assembly and alerts to take proactive actions. We relied upon UX workshop, knowledge transfer, and comprehensive documentation for a better user experience.