Artificial intelligence for refinery flow transmitters Preventing unplanned shutdowns and boosting refinery safety and efficiency through early anomaly detection
NC Chakrabarti, Priyang Shukla, and Jesse Mallhi Reliance Industries Ltd
I n an oil refinery, flow transmitters are crucial for monitoring processes, ensuring efficient control, optimal production, and strict safety. These vital instruments connect to process lines via impulse lines – small tubes transmitting fluid pressure to the sensor (see Figure 1 ). However, these lines are prone to choking from particulate matter or corrosion. Impulse lines are integral components of differential pressure (DP) flow measurement. They provide a conduit for the process fluid’s pressure to reach the DP transmitter. A flow element (like an orifice) creates a pressure drop across it, resulting in a high-pressure (HP) and a low-pressure (LP) tapping. These tapings are connected to the HP and LP sides of the DP transmitter via impulse lines. The DP transmitter measures the difference between these two pressures (ΔP=HP− LP). The flow rate is directly proportional to the square root of differential pressure (Flow ∝ √ΔP ). Any obstruction or choking within either the HP or LP impulse line will distort the true pressure readings, leading to inaccurate ΔP and, consequently, erroneous flow measurement. Such plugging can cause the DP signal to fluctuate erratically or become excessively high or low, effectively providing wrong flow readings to the control system. This fundamental understanding highlights why monitoring the health of these lines is critical for accurate process control and avoiding operational upsets, specifically where flow measurement is used for critical control, such as compressor surge control. Traditionally, refineries use static high and low thresholds for anomaly detection. This
Transmitter
P1
P2
Tx
High pressure impulse line
Low pressure impulse line
Direction of ow
Flow = K * sqrt (
)
P1 – p2 density
method struggles with impulse line choking, as the blockage causes a slow, gradual drift in flow readings, often staying within thresholds despite significant deviations. Furthermore, the dynamic nature of refinery processes results in frequent flow fluctuations, making it difficult to set thresholds that are sensitive enough to detect subtle anomalies while avoiding false alarms. These limitations necessitate a shift. Artificial intelligence (AI) offers a powerful alternative. By learning normal flow patterns, AI can identify subtle deviations indicating impending problems. AI-driven anomaly detection promises early warnings, enabling proactive measures to prevent unplanned shutdowns and ultimately boosting refinery safety and efficiency. Figure 1 Representation of transmitter with impulse line
Refining India
66
Powered by FlippingBook