Why Drilling Equipment Condition-based Maintenance Needs Operational Context?
Condition-based Maintenance (CBM) is becoming ubiquitous today for monitoring equipment health in the drilling industry. One increasingly popular CBM solution is to use external sensors such as supply current and vibration sensors to predict abnormal equipment operations. External sensors are easy to install and they can collect much higher frequency data than internal equipment data to offer advanced detection capabilities.
While sensor-based CBM can work very well for equipment with constant load, such as those used in manufacturing plants, they fall short when it comes to drilling equipment. The reason for this is that drilling equipment is variable load - the workload ramps up and down depending on drilling conditions.
Example 1 | Condition-based Maintenance (CBM) for Drilling Equipment
Mud Pump Vibration and Pressure Data
Plot showing vibration and pressure data for a mud pump.
The plot above shows vibration and pressure data for a mud pump. The vibration was increasing from about 5mm/s to peaking over 15mm/s over a span of 4 days. Many predictive algorithms would flag this as a potential issue.
However, when we examine the pressure data in the same time range, we see that the pressure was also increasing, from about 2000 PSI to 4000 PSI. So the mud pump was simply working harder, which caused the vibration to increase.
Example 2 | Condition-based Maintenance (CBM) for Drilling Equipment
Top Drive Current and Torque Data
The plot below compares supply current to torque for a top drive drive motor. Again, we see that the supply current can vary greatly because it needs to generate the torque required. What’s interesting here is that the torque data is more complex - those spikes represent making connections. If we zoom in, we could also see drilling and reaming states. These variable conditions would confuse most predictive algorithms.
The load range for drilling equipment such as mud pump and top drive could span by a factor of 10 or more. This presents a challenge to any sensor-only predictive algorithms, which tend to under-predict during light load conditions and over-predict during heavy load conditions. This is why there are more false negatives during surface drilling and more false positives during intermediate drilling.
The answer is operational context
How do we solve this problem? The answer is to provide operational context. By adding real-time operational data such as pressure and torque and adjusting the predictive algorithms accordingly, predictions can become accurate again across all operating conditions. What’s more, new algorithms can be developed by leveraging the available operational context. For example, by segmenting operational conditions to specifically match to the signatures of swab failures, swab failures can be accurately predicted.
Prescient provides operationally context-aware CBM solutions that integrate operational data such as EDR and rig state to provide accurate failure predictions across complex rig operating conditions and states. For more details please contact us.