Success Story
About the Client
Mechademy is an engineering consulting firm bringing together top turbomachinery experts from around the world. They provide consulting, commissioning, start-up, remote monitoring, and troubleshooting services for the oil and gas industry. Combining decades of industry experience with deep knowledge in thermodynamics, rotordynamics, aerodynamics, and fluid mechanics, they deliver solutions that are both practical and technically sound. Their expertise has helped clients worldwide achieve millions of dollars in value, offering reliable best practices and innovative solutions that address complex engineering challenges effectively.
Business Situation and Requirements
Production assets in oil and gas fields operate in harsh, remote environments, both offshore and onshore. These conditions often cause equipment failures, frequent downtime, and reduced asset performance. Disconnected systems and limited real-time visibility make it harder to monitor operations and respond quickly, impacting efficiency and production targets.
Mechademy aimed to build a smart monitoring and anomaly detection system for oil and gas refineries. The goal was to use AI and machine learning to detect faults early and predict failures. They wanted a reliable, scalable solution to improve asset performance and reduce downtime. To achieve this, they partnered with Unthinkable for its expertise in AI and industrial IoT.
The client chose Unthinkable for its experience in building advanced monitoring systems.
Their key requirements included:
Develop an AI-based software solution that works seamlessly with existing data historians or condition monitoring systems.
Detect equipment issues and underperformance early, often before traditional protection systems can respond.
Use machine learning and performance modeling to gain deeper insights into turbomachinery health.
Analyze data from multiple sensors, such as acoustic sensors, accelerometers, and infrared thermography, to spot sub-optimal operations.
Track trends and compare performance with models to monitor equipment more effectively.
Display key insights through an easy-to-use, customizable dashboard with visualizations like bar charts, pie charts, and line graphs.
Provide both a high-level overview and detailed drill-down views of plant performance to support better operational decisions.
Solution
The project began with our business analysts and solution architects defining the overall system architecture. They identified key requirements and planned how sensor data would be ingested, processed, and transformed. The approach for ETL pipelines, AI/ML integration, and real-time analytics was also outlined, helping shape a clear product vision and development roadmap.
With the foundation in place, the team selected the right technologies to support the solution. Tools like Apache Airflow, AWS EMR, Kubernetes, Node.js, React.js, MongoDB, Data Lake, and PySpark were chosen to ensure the platform remains scalable, efficient, and easy to manage.
Some of the key features were:
IoT Data Integration & Ingestion
- Integrated 100+ plant sensors with the platform through a connector agent.
- Connector agent retrieves data from plant historians and transfers it to the Mechademy data lake.
- Enables continuous and scalable ingestion of machine and equipment data.
Data Processing & ETL Pipelines
- Built several Extract, Transform, Load (ETL) pipelines to clean up raw plant data and organize it so it’s ready for analysis.
- Implemented missing data handling, data type validation, and outlier removal.
- Converted raw data into a standardized format suitable for analytics and modeling.
AI-Based Anomaly Detection
- Applied Machine Learning (ML) and Deep Learning (DL) models trained on historical plant data.
- Algorithms identify abnormal behavior and performance deviations.
- Helps detect potential equipment issues early.
Prescriptive Alerts & Decision Support
- Generated prescriptive alerts based on anomaly detection results.
- Alerts provide actionable insights to improve plant health and efficiency.
- Supports proactive maintenance and operational decision-making.
Data Visualization & Custom Dashboards
- Enabled visualization of raw, processed, and output data through graphs.
- Built customizable dashboards tailored to each client’s requirements.
- Allows both high-level plant overview and detailed drill-down analysis.
Security & Access Management
- Implemented authentication with role-based user groups and permissions.
- Ensured secure access and controlled system operations.
Dynamic ML/DL Model Deployment
- Enabled deployment of customized ML and DL models from the admin interface.
- Applied machine-specific algorithms based on configurable parameters.
- Supported dynamic model updates without disrupting system operations.
Scalable Analytics Platform
- Built a flexible architecture for IoT data ingestion and complex ETL pipelines.
- Enabled dynamic dashboard creation and analytics customization for different clients.
Impact
The client was highly satisfied with how Unthinkable brought their vision to life and successfully delivered the platform. Encouraged by the results, they are now planning further updates and enhancements to expand the system’s capabilities.
The solution also helped Mechademy evolve from an engineering consulting firm into a product and service provider in advanced analytics and Industrial IoT.
Their clients have shown strong interest in the platform and reported up to 80% reduction in unplanned downtime, significantly improving overall plant efficiency.
