A new whitepaper by ExperTune’s George Buckbee suggests that causes of many process problems go undetected for years. Buckbee advocates taking a closer look at upsets to identify their true root cause. But drilling down through the large dataset produced by a complex process can be extremely difficult. A single real problem can create multiple secondary effects in a plant. Trying to tune all the affected control loops is simply ‘chasing after symptoms, rather than getting to the heart of the issue.’
Buckbee divides upsets into two categories, cyclical and non-cyclical. Cyclical upsets or oscillations come from sticking control valves, poor controller tuning and other sources. The key to solving such issues is determining the period of oscillation. Process upsets that share a common cyclical root cause will be oscillating at the same period. Fourier analysis and ‘a little bit of logic’ determines whether the oscillation came from poor tuning, a faulty control valve, or some other cause. ExperTune’s PlantTriage control loop monitoring package provides such analysis and lets the investigator sort a list of observations by oscillation period to identify loops that share the same period. Next, look for the control loop that is furthest up-stream in the process—this is likely to be the root cause of the problem.
Non cyclical upsets are often associated with a process change such as start-up or shut-down of a unit operation. Buckbee suggests that a similar approach can be used—but with a longer correlation window. Again, PlantTriage provides the tools, identifying ‘interaction hot spots’ of the strongest interactions, shown on a ‘process interaction map.’ Buckbee notes that engineers report ‘a degree of surprise’ when finding that they have been working on symptoms, not the root cause, for many years. Solving a single problem can result in millions of dollars in annual revenue. More from expertune.com.
© Oil IT Journal - all rights reserved.