The data is contributed by industrial control practitioners as well as by academic researchers who work closely with industry. The aim of this repository is to provide a test environment for control loop performance monitoring methods. 

You will find the data repository here.

The data sets are arranged according to the cause of the poor performance of the control loops. These categories are described in Bauer et al. (2016) and are ordered according to how often they occur in industrial processes.

The repository is described in the following paper. For valve stiction, a comparative study has been conducted using industrial data that is also available on this repository, Bacci di Capaci et al., 2016. This data repository wants to encourage the development of similar studies.

 

An initiative by SACAC – the South African Council of Automation and Control

SACAC is the national member organization of IFAC and a promoter of control and automation developments and technologies. For any queries or if you would like to contribute industrial process data please contact the team using the following email address: piddata@sacac.org.za

Why the need for an industrial data repository?

The majority of controllers in the process industries are single-input-single-output control loops using the PID algorithm. Although successful and widespread in its applications, studies consistently find that a large proportion of control loops are not fulfilling their task to their best potential, that is, they are performing poorly’. The repository makes industrial data available to researchers who develop methods for the detection of poorly performing loops. Many algorithms and methods have been developed to detect, diagnose control loops that are performing poorly for various reasons. Some of these methods are implemented in industrial software tools. All commercially successful methods have been tested on industrial data. However, these are individual datasets and are generally not shared with the research community. It becomes therefore difficult to compare and evaluate methods. The repository is made available to promote the successful development of control loop performance monitoring methods.