Computer-aided acceptance planning: Generating quality acceptance parameters and stratified sampling plans through neural network learning ability and CAD modeling
Abstract
A quality acceptance sampling plan (QASP) is the key in designing Quality Assurance specifications. The QASP guides the decision between accepting or rejecting the quality of products by specifying the requirements of: (1) how many measurements are needed, (2) where to take these measurements, and (3) how to make an acceptance or rejection decision based on measured data. This research tackles the problems encountered in the development of a QASP. This research is the first work exploring the learning ability of an Artificial Neural Network (ANN) to solve the problem of designing an acceptance sampling plan. The ANN features the capability of "learning-from-example." A special approach has been designed to create a training database that generates "good examples" and filters out "bad examples." Therefore, the obtained training examples have the advantage of providing more efficient sampling plans. Namely, the trained ANN can produce an acceptance plan that has a smaller sample size, but still obtain the desired quality levels. The obtained ANN in this research can also be directly applied to other types of acceptance planning. Namely, without re-training, the ANN can be broadly used in other construction areas, such as highway pavement and concrete tests. Meanwhile, CAD modeling is studied to help generate a stratified sampling scheme and random inspection spots. The CAD system defines a basic surface element that is suitable for representing the surface of steel structures. This surface element with four vertexes is both simple and easy to operate. The defined surface element can be efficiently processed, stratified, and grouped to simulate sampling lots. A random sampling algorithm has also been defined to pick inspection spots. As a result, contractors could pay more attention to maintaining the equal quality over the entire construction projects. This should lead to more efficient use of taxpayers' money. The proposed CAD modeling provides the capability to manage the costs associated with the inspection risks. This work has resulted in the development of two packages. The software called Q-Design (abbreviation of Quality acceptance plan Designer) has been developed to fulfill the Artificial Neural Network Module, and the I-CAD (abbreviation of Inspection plan generator in CAD system) has been developed to support the CAD modeling module.
Degree
Ph.D.
Advisors
Chang, Purdue University.
Subject Area
Civil engineering|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.