Design and Implementation of Machine Learning and Rule-Based System for Verifying Automation System Designs
Yinkfu Chuye, Kenedy (2013)
Yinkfu Chuye, Kenedy
Metropolia Ammattikorkeakoulu
2013
All rights reserved
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2013120520269
https://urn.fi/URN:NBN:fi:amk-2013120520269
Tiivistelmä
A machine learning and rule-based system has long been applied in major areas such as bioinformatics, natural language processing and structural health monitoring. Machine learning uses algorithms to discover patterns in data and construct predictive models, whereas a rule-based system encapsulates knowledge of the domain expert from data thereby making decisions using rules. This thesis demonstrated a hybrid approach com-bining machine learning and a rule-based system for verifying automation system designs.
Companies producing automation plant design are faced with problems such as delivering quality products to meet the customer’s requirements within a time frame. There is, there-fore, the need for the companies to improvise their automation process in order to save design time. Data produced during the automation design process of a plant can be used to create a knowledge base, which is a collection of rules captured from the data pattern. An intelligent system can be created using this knowledge base to verify automation design quality assurance.
The goal of this thesis was to design and implement a system that can be used by less experienced automation engineers at Pöyry. The method applied in this thesis uses an inductive learning algorithm to generate production rules from training data. The production rules generated were used for building a knowledge base for a rule-based system. In order to evaluate the performance of the knowledge base, three different learning classifiers were used. Effective score of the learning classifiers prove a decision-tree learner to be the best classifier with an average performance score of 92%. The outcome of the thesis was a desktop application developed using Java GUI (Graphical User Interface) widget toolkit. This application can be used to perform task such as verification of system design.
This thesis was carried out for the company Pöyry to solve automation design carried out by less experienced engineers. The results of the thesis illustrate that the proposed thesis can be implemented in a real situation.
Companies producing automation plant design are faced with problems such as delivering quality products to meet the customer’s requirements within a time frame. There is, there-fore, the need for the companies to improvise their automation process in order to save design time. Data produced during the automation design process of a plant can be used to create a knowledge base, which is a collection of rules captured from the data pattern. An intelligent system can be created using this knowledge base to verify automation design quality assurance.
The goal of this thesis was to design and implement a system that can be used by less experienced automation engineers at Pöyry. The method applied in this thesis uses an inductive learning algorithm to generate production rules from training data. The production rules generated were used for building a knowledge base for a rule-based system. In order to evaluate the performance of the knowledge base, three different learning classifiers were used. Effective score of the learning classifiers prove a decision-tree learner to be the best classifier with an average performance score of 92%. The outcome of the thesis was a desktop application developed using Java GUI (Graphical User Interface) widget toolkit. This application can be used to perform task such as verification of system design.
This thesis was carried out for the company Pöyry to solve automation design carried out by less experienced engineers. The results of the thesis illustrate that the proposed thesis can be implemented in a real situation.