| 초록 |
At an industrial process, a lot of process variables are measured using sensors. These measurements inevitably have noise that cannot be predicted with certainty. It is caused by the combination of many sources such as signal conversion noise, ambient condition change and so on. The existence of noise breaks the mass balance around each unit. It is usually eliminated by filters. But there still exists imbalance to a certain degree. Thus, as applying DR technique that is based on process model such as mass balance and energy balance, we can find more accurate adjustments of measurement values and reach the true value as close as possible. But, in the real plant, there are random errors (noise) as well as gross errors that are relatively larger than random error due to sensor fault, miscalibration and so on. When we apply DR to the measurement with gross error, it leads to distortions at all the related measurements. Thus, we need to detect gross error in advance and then apply DR to the cleaned measurements. For the case study, we utilized DR and GED techniques in utility system case with gross errors. As a result, we can adjust measurement to the true value as close as possible. |
