Classification of Systematic Measurement Errors within the Framework of Robust Data Reconciliation
A robust data reconciliation strategy provides unbiased variable estimates in the presence of a moderate quantity of atypical measurements. However, estimates get worse if systematic measurement errors that persist in time (e.g., biases and drifts) are undetected and the breakdown point of the robus...
| Authors: | , , |
|---|---|
| Format: | article |
| Status: | Published version |
| Publication Date: | 2017 |
| Country: | Argentina |
| Institution: | Consejo Nacional de Investigaciones Científicas y Técnicas |
| Repository: | CONICET Digital (CONICET) |
| Language: | English |
| OAI Identifier: | oai:ri.conicet.gov.ar:11336/43006 |
| Online Access: | http://hdl.handle.net/11336/43006 |
| Access Level: | Open access |
| Keyword: | SYSTEMATIC MEASUREMENT ERRORS DATA RECONCILIATION ROBUST STATISTICS https://purl.org/becyt/ford/2.4 https://purl.org/becyt/ford/2 |
| Summary: | A robust data reconciliation strategy provides unbiased variable estimates in the presence of a moderate quantity of atypical measurements. However, estimates get worse if systematic measurement errors that persist in time (e.g., biases and drifts) are undetected and the breakdown point of the robust strategy is surpassed. The detection and classification of those errors allow taking corrective actions on the inputs of the robust data reconciliation that preserve the instrumentation system redundancy while the faulty sensor is repaired. In this work, a new methodology for variable estimation and systematic error classification, which is based on the concepts of robust statistics, is presented. It has been devised to be part of the real-time optimization loop of an industrial plant; therefore, it runs for processes operating under steady-state conditions. The robust measurement test is proposed in this article and used to detect the presence of sporadic and continuous systematic errors. Also, the robust linear regression of the data contained in a moving window is applied to classify the continuous errors as biases or drifts. Results highlight the performance of the proposed methodology to detect and classify outliers, biases, and drifts for linear and nonlinear benchmarks. |
|---|