Early Detection of Valve Failures in Gas Lift Reciprocating Compressors
Early Detection of Valve Failures in Gas Lift Reciprocating Compressors
Problem
An upstream oil and gas company struggled with recurring downtime on gas engine and induction motor driven reciprocating compressor packages used in gas lift applications. These frequent machine trips forced the remote site to be manned around the clock, ensuring immediate restart to avoid prolonged downtime. While this practice resulted in acceptable equipment uptime, it came at a significant manpower cost, as continuous on-site monitoring was unsustainable for the company in the long term.
Solution
The company had previously tried two artificial intelligence (AI) implementations, both of which failed to deliver meaningful results due to the vendors’ lack of domain expertise in rotating equipment. The company then turned to Mechademy and deployed its fleet of compressor packages on the Turbomechanica platform. Within days of deployment, Turbomechanica had baselined the equipment performance and started detecting ongoing valve leakage and damage across the compressor fleet with over 95% accuracy.
Leveraging physics-based models that use real-gas equations of state, the Turbomechanica platform augments field data with synthetic sensors, including critical variables like cylinder volume ratio, adiabatic efficiency, polytropic exponent, rod loads, brake power, and mass flows. Its orchestration framework ensures seamless data flow across equipment and sub-components. As an example for reciprocating compressors, the brake power from each cylinder is summed up and compared to the output power from a driver gas engine or induction motor. Similarly, the calculated mass flow from one compression stage is sent to the next stage for evaluation.
The Turbomechanica diagnostic engine provides enriched diagnostics by combining knowledge from physics, machine learning, and statistical algorithms. With the ability to differentiate between valve leakage, valve damage, and transmitter drift, the platform provided the customer with an average of 7-10 days of advanced notification before valve failure. This enabled the company to take proactive measures to prevent equipment breakdowns.
Implementation
Before the equipment was fully integrated with Turbomechanica, historical data was analyzed to baseline the performance of the compressor packages. This initial assessment revealed several data issues, such as incorrect scaling and units, faulty sensor data, and mislabeled sensors. Once these issues were corrected, the platform’s early fault detection capabilities were fully operational from day one of 'go live'. Mechademy’s data integration team collaborated closely with the customer to ensure smooth data transfer and proper system setup.
Results
With advanced warning of valve leakage and impending valve damage, the company was able to schedule proactive repairs, eliminating the need for round-the-clock site manning. Over several months of monitoring, patterns in the data revealed that specific compressor setups in the field were prone to the highest number of failures. Armed with this insight, the customer made data-driven modifications to these units, significantly reducing the frequency of valve failures and improving overall equipment reliability.