A New Approach to Identifying Operational Conditions in Multivariable Dynamic Processes Using Multidimensional Projection Techniques SPE 69523
Language: Inglés Series: Ranson, A ; Publication details: Dallas, Texas Society of Petroleum Engineers 2001Online resources: Summary: Industrial processes may have a large number of measured variables. Combining these variables is possible to identify different operational regions and detect anomalies or deviations within the process. However, it is hard to visualize operational conditions and identify problems on line. The difficulties arise because the field operators must observe and analyze dozens of variables concurrently in order to identify the operational state of the process. This is not a new problem and previous attempts to find a solution have limitations to overcome. One traditional approach was to assign alarms to each variable to alert operators about potential problems. However, far from solving the problem, this technique aggravated it, mainly because there were too many alarms and a number of them turned to be false. This prompted the operators to ignore the control system, which made things worse. A novel multivariable-process visualization technique is being proposed to help in the tasks of supervision, observation of operational states, and detection of problems in multivariable processes. The technique combines steady-state-detection filters, statistical filters, pattern matching, and spatial transformations for the gathered data. A spatial transformation is the key element for data processing, and consists of a space-reduction algorithm based on keeping a set of attributes unchanged between the original space and the projected space. This paper describes a general methodology to monitor multivariable industrial processes. The methodology was applied to a control test case as well as a real time debutanizer column process.| Current library | Status | Barcode | |
|---|---|---|---|
| Biblioteca virtual | Not for loan | 200003089 |
Industrial processes may have a large number of measured variables. Combining these variables is possible to identify different operational regions and detect anomalies or deviations within the process. However, it is hard to visualize operational conditions and identify problems on line. The difficulties arise because the field operators must observe and analyze dozens of variables concurrently in order to identify the operational state of the process. This is not a new problem and previous attempts to find a solution have limitations to overcome. One traditional approach was to assign alarms to each variable to alert operators about potential problems. However, far from solving the problem, this technique aggravated it, mainly because there were too many alarms and a number of them turned to be false. This prompted the operators to ignore the control system, which made things worse. A novel multivariable-process visualization technique is being proposed to help in the tasks of supervision, observation of operational states, and detection of problems in multivariable processes. The technique combines steady-state-detection filters, statistical filters, pattern matching, and spatial transformations for the gathered data. A spatial transformation is the key element for data processing, and consists of a space-reduction algorithm based on keeping a set of attributes unchanged between the original space and the projected space. This paper describes a general methodology to monitor multivariable industrial processes. The methodology was applied to a control test case as well as a real time debutanizer column process.
