Unlocking Control Performance Insights: Evaluating Large Data Sets for Industrial Optimization
Publication: On Evaluating Control Performance on Large Data Sets
Authors: Alexander Horch, Friedrun Heiber
Publication Summary
The publication discusses the evaluation of control performance using performance indices from large amounts of measurement data. It explores the usefulness of simple performance measures and evaluates established methods and new ideas on industrial data sets. The publication emphasizes the importance of considering multiple data sets for performance assessment and suggests the use of simple statistics and more advanced indices. It also discusses the potential for building a nonlinearity map of the process and the benefits of combining different indices for a comprehensive analysis.
What is the focus of the publication?
The focus of the publication is twofold. Firstly, it aims to assess the usefulness of simple control performance indices when evaluated on multiple data sets. Secondly, it investigates the continuous collection of information from multiple data sets to build a nonlinearity map of the process, which can be utilized for tuning procedures. The publication evaluates various performance indices, both simple and more advanced ones, on industrial data sets and discusses the implications and requirements for industrial control performance monitoring tools.
What are the key conclusions of the publication?
The key conclusions of the publication are as follows:
- Simple control performance indices can provide useful information when evaluated on multiple data sets.
- The availability of many data sets can be used to build a nonlinearity map of the process, which is useful for tuning procedures.
- Simple statistics are useful for quick scans of large amounts of data, while more complex indices are valuable when averaging over multiple data sets.
- Combining different indices can provide a comprehensive analysis of control performance.
- Correlation of indices for different control loops can reveal common oscillatory behavior.
- Static input-output maps can be created to analyze experimental data and identify suitable data sets for model identification.
- Storing and combining assessment information from single evaluations can provide valuable insights for future use.
- The publication emphasizes the importance of considering multiple data sets for performance assessment and the benefits of using a combination of simple and advanced performance indices.
What are the advantages of using simple performance measures?
The advantages of using simple performance measures are as follows:
- Fast and overview-like scans: Simple statistics provide a quick way to scan large amounts of data and get a general understanding of control performance without extensive computations.
- Easy computation: Simple indices can be evaluated with a modest amount of computations, making them computationally efficient.
- Regular assessment: Since performance monitoring typically involves assessing many control loops regularly, simple indices are appealing as they can be easily computed and evaluated for multiple data sets.
- Low computational effort: Some simple indices, like the minimum-variance control performance index, combine low computational effort with important information about the current loop performance.
- Popular and widely used: Simple indices, like the control error mean and variance, have gained popularity over the years due to their simplicity and ability to provide valuable information about control performance.
Overall, simple performance measures offer a quick and efficient way to assess control performance across multiple data sets, making them useful in industrial settings where regular monitoring is required.
How can the availability of many data sets be utilized for control performance assessment?
The availability of many data sets can be utilized for control performance assessment in several ways. One approach is to evaluate simple control performance indices on multiple data sets. This allows for a quick and overview-like scan of large amounts of data, providing valuable information about control performance. Additionally, the continuous collection of information from multiple data sets can be used to build a nonlinearity map of the process, which is useful for tuning procedures. By analyzing and comparing performance indices across different data sets, patterns and trends can be identified, leading to a better understanding of control loop behavior. Furthermore, the combination of different indices can provide a comprehensive analysis of control performance, offering deeper insights into the performance of the control loops. Overall, the availability of many data sets enables a more thorough and accurate assessment of control performance.
What are the implications for industrial tools discussed in the publication?
The publication discusses several implications for industrial tools in the context of control performance assessment. These implications include:
- Analysis of Performance Indices: Industrial tools should enable the analysis of performance indices such as plot combinations, correlation, and trend plots. This allows for a comprehensive understanding of control performance.
- Application-Dependent Selection: Tools should provide the capability to selectively discard or prioritize specific performance indices based on the application requirements. This flexibility ensures that the most relevant indices are considered for each control loop.
- Index Database and Data Retrieval: Industrial tools should offer an index database that allows for search queries, making it easier to retrieve specific performance indices. Additionally, tools should facilitate the retrieval of data collection dates and specific data sets for further analysis.
- Support for Model Identification: The availability of many data sets can be utilized to build a nonlinearity map of the process, which aids in tuning procedures. Industrial tools should support the identification of suitable data sets for model identification and provide mechanisms to store and retrieve this information.
Overall, the implications for industrial tools revolve around enabling comprehensive analysis, flexible selection of performance indices, efficient data retrieval, and support for model identification and tuning procedures. These features can enhance control performance assessment and optimization in industrial settings.
