Magazine article AI Magazine

Ontology Reengineering: A Case Study from the Automotive Industry

Magazine article AI Magazine

Ontology Reengineering: A Case Study from the Automotive Industry

Article excerpt

The Direct Labor Management System (DLMS) (Rychtyckyj 1999) was initially developed and deployed in Ford Motor Company's North American assembly plants back in the early 1990s. It was recognized that an ontology and a reasoner were required to represent the complex knowledge in the manufacturing process. This was done by creating an implementation of the KL-ONE language using the LISP programming language and developing a classifier that could reason with the ontology. This implementation turned out to be extremely successful and became the production version as the system was expanded to assembly plants first in Europe and then the rest of the world. Throughout this, the KL-ONE architecture remained in place as the ontology was expanded and maintained through thousands of updates.

As the semantic web architecture and standards were developed, it became obvious that the Global Study Process Allocation System (GSPAS) KL-ONE ontology would be much more usable and of better value to Ford if it could be rewritten into OWL/RDF. An ontology based on modern semantic web standards would be much easier to maintain and could be extended and utilized for other applications in the company. The main issue was in terms of time and resources: GSPAS was a production system with high value to the business customers and it was impossible to spare the people to redo the ontology and keep the existing system in production. An alternative solution was needed and Ford found it by partnering with the Indian Institute of Technology Madras (IITM) in Chennai, India. Ford elected to partner with IITM because the university has an excellent reputation with a strong background in artificial intelligence (Khemani 2013), and moreover, Ford wanted to develop a strong relationship with the university.

The results of this project were very successful. The IITM team delivered a reengineered OWL/RDF ontology that contained the knowledge in the existing KL-ONE ontology. The Ford team validated and updated the ontology to meet Ford's requirements and has deployed the lexical ontology into the GSPAS application. In the rest of the article we will describe the structure and usage of the existing KL-ONE ontology, and then describe the conversion approach and the conversion process.

In this article, we refer to the GSPAS KL-ONE ontology as GSPAS KB or as GSPAS ontology or as KL-ONE ontology, and refer to the reengineered GSPAS OWL ontology as new ontology or as OWL ontology.

GSPAS and the KL-ONE Ontology

Ford's DLMS was developed to standardize vehicle assembly, improve efficiency, and reduce cost throughout the entire manufacturing process planning system. DLMS was then integrated into Ford's Global Study Process Allocation System, which is currently used across all of Ford's global vehicle assembly and powertrain plants.

Artificial intelligence in GSPAS is used for several different purposes: (1) Validate the correctness of process sheets that describe assembly operations. (2) Develop a list of operator work instructions and associated MODAPTS (modular arrangement of predetermined time standards) codes (Sullivan, Carey, and Farrell 2001) for each assembly operation in the process sheet. (3) Check the process sheet for ergonomic concerns. (4) Translate the process sheets into the language used at a particular assembly plant.

Figure 1 shows the architecture of the GSPAS AI application. Figure 2 shows a sample process sheet with five build steps and two tool specifications; at such granularity, thousands of process sheets are used to document the build steps for a whole vehicle. The core of the GSPAS AI application is an ontology that contains relevant knowledge about Ford's manufacturing processes including the labor requirements for the assembly operations, part and tooling information, workplace ergonomic concerns, linguistic representation of Standard Language (Rychtyckyj 2006) and other concepts. Figure 3 shows how this ontology is used to generate operator work instructions and MODAPTS codes. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.