<p>Until the early 1900s, furniture was built by hand, one piece at a time. The industrial revolution and modern manufacturing technology has changed all of that. Today, as the furniture industry moves firmly into the next century, computerized systems and automated manufacturing have become more common to the industry. This thesis represents an effort to analyze the current practices in computer-aided design of upholstered furniture, specifically plywood frame furniture, and to develop new procedures for reducing the lead-time in upholstery product development. Different 3-D modeling techniques for designing plywood furniture frames and their features have been developed and implemented. A plywood frame feature library has been created, and geometric relations needed to fully constrain each feature type have been developed. This reduces modeling time and also increases consistency in the solid models. A new reverse engineering procedure, using an articulating arm, has been proposed, implemented, and tested for 3-D digitization of plywood frames. The proposed methodology eliminates some of the traditional processes currently followed in the industry, thus making the product development faster and more streamlined. Further, an algorithm has been developed, implemented and tested for automatically mirroring plywood upholstery frame assemblies in a CAD system. The algorithm considerably reduces the modeling lead-time in the product development process. Finally, some future work that considers currently available 3-D CAD technologies has been recommended which would help close the gap between upholstery designers and manufacturers.<P>

<p>The use of non-linear display methodologies has been successfullydemonstrated in a number of text- based and simple, graphic networkapplications; however, most studies have focused on computationalefficiency with very little attention to human perceptual limitations.While the benefits of non-linear displays are purported to be due toadditional context with no loss of required detail, reports of end-userperformance gains have been limited at best. This research entailed thedesign of a non-linear tactical map display, where the distortionparameters were based upon empirical test results of navigation- relatedtasks combining various scales and distortion techniques. A total ofthree experiments were conducted: two to define the non-linear displayparameters, and the third to attempt to validate the display through aperformance test using military participants. Results indicate thatadditional scales can indeed be combined on the same map surface with nosignificant penalty in navigation accuracy. However, there may be aresponse time difference depending on the particular distortion techniqueapplied. A modified fisheye distortion proved to be better suited to thistask environment than a Cartesian bifocal distortion. Performance testingversus a traditional, pan and zoom display interface yielded nodifferences in navigation task accuracy or response times. However, theexpected gain in random event detection (through display of additionalcontext) was not evident. <P>

<p>WARDONO, BAGAS. Algorithms for the multi-stage parallel machine problem with buffer constraints. (Under the direction of Yahya Fathi).We study two problems in the context of scheduling N jobs on parallel machines in L successive stages with the primary objective of minimizing the makespan. In the first problem we assume that there are unlimited buffer capacities between successive stages. We refer to this problem as the FSPM problem. In the second problem, we assume that the buffer capacities are finite and given, and we refer to it as FSPM/b problem. Both problems are shown to be NP-hard.We focus our attention in developing heuristic search procedures for solving both problems. These procedures include the local improvement, the genetic algorithm, and the tabu search. In this context, we use two different solution representations, namely the matrix representation and the vector representation. To evaluate these procedures on an empirical basis, we carry out an extensive computational experiment consisting of two different types of data sets. The problem instances in one data set are constructed in such a manner that the corresponding optimal makespan can be easily determined. The optimal makespan of the instances in the second data set are not known. We also develop a set of lower bounds for the optimal makespan of the problem. The experimental results with the procedures that we propose show that the search procedures which are based on vector representation perform reasonably well. Among the various procedures based on this representation, our tabu search implementation was the most effective.In the context of FSPM/b problem, we also conduct an experiment with a third set of data which is available in the open literature. From the experimental results, we demonstrate the characteristic of the proposed algorithms with various buffer configurations.<P>

<p>This study was conducted to develop solution algorithms for the edge-disjoint path problem with proportional path-costs in an undirected network with single source-destination pair. Edge-disjoint path problem arises in connectivity analysis, network flow problems, VLSI-design and variety of applications. The complexity of all the variants of the disjoint path problemsis NP-hard. We proposed two genetic algorithms which differ in solution representation.The performance of both algorithms tested on small, medium and large size complete graphs using different cost structures. As a result, genetic algorithm with priority-based encoding is shown to beinefficient, whereas genetic algorithm with variable-length path encoding is shown to be satisfactoryfor finding the edge-disjoint paths with proportional path-costs.<P>

<p>High-level process planning plays an important role in determining candidate process domains at the configuration design stage. Changing the processes domainslater increases the product development cycle and the product development cost. Therefore, determining the most appropriate manufacturing processes at thebeginning stages of the design process becomes critical. However, high-level process planning systems have traditionally lacked integration of design synthesis anddesign evaluation. The main objective of this research has been to develop a FEature-Based Design And Process Planning (FEBDAPP) system that helps designers decide whether ornot the design is worth pursuing by providing manufacturing advice to designers during the design process. In order to achieve the main objective, the following taskshave been accomplished: (1) developed a hybrid system incorporating design by feature and feature recognition approaches capable of reducing the complexity offeature recognition algorithms without sacrificing flexibility in creating a part design, (2) developed a comprehensive set of feature mapping algorithms capable oftransforming a primary part representation into a secondary part representation that is required as input to downstream applications, and (3) developed aCAD-based interface capable of integrating a current CAD system with a high-level process planning system.This research contributes significantly to the availability of early design tools that enhance and at the same time shorten the design process cycle. The implementationof feature technology in this research will support the development of applications such as tooling cost estimation and manufacturability analysis. Finally, theFEBDAPP system is intended to be an effective concurrent engineering tool that bridges the gap between design and manufacturing.<P>

<p>Furniture manufacturers today have to deal with the conflicting objectives of increased product variety with reduced lead-time and inventory levels. The prevalentsentiment in the industry is a build to order production strategy, which tends to increase the production costs. Furniture makers are hence looking for technologicalsolutions to counter these issues. Fixture-less machining on CNC routers is an alternative gaining prominence among manufacturers to reduce setup times andincrease productivity. However, in the absence of locating pins for mounting the parts, the orientation and part placement becomes critical and small errors canincrease rejection rates. Updating the CNC part programs after determining the exact orientation and location of the part blanks is a likely solution. This thesisrepresents an effort to develop a computer vision based system to enable this. Some of the popular techniques in computer vision for object recognition were henceimplemented. These however did not prove worthwhile as the efficacy of the algorithms was undermined by the texture from wood grains. A template matchingprocedure was identified as being better suited for the purpose because of its inherent tolerance for noise. The generation of the template to account for hidden linesand actual projected lengths, both functions of the viewing direction and camera location, was however the biggest challenge. The fact that most furnituremanufacturers have a very large product range that needs machining on the router compounds the problem. A novel CAD based template generation algorithm washence proposed and implemented. The developed algorithms enabled orientation determination with a negligible error, and the average error in locating the part wasless than 0.2 percent. The approach was therefore found to have a potential in reducing the lead times while specifying tighter tolerances to reduce raw-materialconsumption. Finally, some future work to completely automate the process and to extend the applicability of the approach has been proposed.<P>

<p>KOC, BAHATTIN Computational geometric analysis and planning for 3D rapid prototyping processes. (Under the direction of Professor Yuan-Shin Lee)<br><br>In this research, computational geometric analysis and planning methods are developed for rapid prototyping (RP) processes. To increase the accuracy of the slicing contours from the STL file, biarc curve fitting of the linear contours has been developed. The mathematics formulation and a Max-Fit algorithm have been developed to find the best biarc curve fitting of the STL slicing data points. Max-Fit biarc fitting algorithm progresses through all the points on the slicing contour data to find an efficient biarc fitting. The results show that rough cross-sectional contours from the STL models can be smoothed with the developed method of Max-Fit biarc. Therefore, less strict requirements on the STL triangulation tolerance can be used when the STL model is generated for rapid prototyping. To increase the efficiency of the rapid prototyping process, non-uniform offsetting and hollowing of solid objects by using biarcs fitting have been developed. The developed non-uniform offsetting method can reduce the area that needs to be built so the total RP build time can be reduced significantly. Using the developed method can also reduce the expensive build material usage. The hollowed objects with constant wall thickness are constructed by offsetting the vertices in their corresponding normal directions. To reduce the staircase effect in the 2D layers of RP processes, a new approach of ruled layer approximation of STL models has been developed. A surface error analysis method has been developed for the proposed ruled layer approximation. Using the surface error assessment, the STL model can be sliced adaptively to reduce the total number of layers without sacrificing the required surface accuracy. To fabricate the ruled layers, the material removal process, specifically 5-axis machining, is integrated into the traditional 2D RP processes. The tool paths are generated for the ruled layers to machine the thick layers fabricated by a traditional RP process. Computer implementation and illustrative examples of the proposed methods are also presented in this dissertation. The results show the developed methods can significantly increase the accuracy and the efficiency of RP processes. <P>

<p>The problem of concern to us in this thesis is the scheduling ofprecedence-related jobs non-preemptively on two identical parallelprocessors to minimize the sum of the weighted completion times. The problemis known to be NP-hard.We develop, in chapter 2, a binary integer program which iscapable of solving only small size problems (no larger than 12jobs) to optimality at the present time. We also present a linearprogramming(LP) model adopted from the literature todetermine the lower bound on the optimum. This LP stands us ingood stead when we perform the optimization via the GeneticAlgorithm approach (which is the subject matter of chapter 3). Wealso present a dynamic programming formulation based on theapproach used for solving the "weighted earliness-tardiness"problem. Although DP expands somewhat the size of the problemsthat can be solved to optimality, its computing burden becomesonerous for more than 25 jobs.In an attempt to solve larger, and more realistic problems, a GeneticAlgorithm(GA) is presented in chapter 3. The salient feature of the GAmodel is that the "initial population" of trial solutions are not allrandomly generated but are constituted from a set of priority rules whichare known to be "good" relaxation (in the sense of being "close" to theoptimum) of the original problem. Also, generation of infeasible solutionsis avoided by the use of post-processing procedures after crossover andmutation operations. Computational results show that the GA approach arrivesto within 20% of the lower bound (and hence of the optimum) in very fewiterations.<P>

<p>In this research, a novel population-based global optimization method has been studied. The method is called Electromagnetism-like Mechanism or in short EM. The proposed method mimicks the behavior of electrically charged particles. In other words, a set of points is sampled from the feasible region and these points imitate the role of the charged particles in basic electromagnetism. The underlying idea of the method is directing sample points toward local optimizers, which point out attractive regions of the feasible space.The proposed method has been applied to different test problems from the literature. Moreover, the viability of the method has been tested by comparing its results with other reported results from the literature. Without using the higher order information, EM has converged rapidly (in terms of the number of function evaluations) to the global optimum and produced highly efficient results for problems of varying degree of difficulty.After a systematic study of the underlying stochastic process, the proof of convergence to the global optimum has been given for the proposed method. The thrust of the proof has been to show that in the limit, at least one of the points in the population moves to the neighborhood of the global optimum with probability one.The structure of the proposed method is very flexible permitting the easy development of variations. Capitalizing on this, several variants of the proposed method has been developed and compared with the other methods from the literature. These variants of EM have been able to provide accurate answers to selected problems and in many cases have been able to outperform other well-known methods.<P>

This dissertation studies scheduling for service stability and for supply chain coordination as well. The scheduling problems for service stability are studied from the single perspective of a firm itself, while the scheduling problems for supply chain coordination are investigated from the perspective of a supply chain. Both the studies have broad applications in real life. In the first study, several job scheduling problems are addressed, with the measure of performance being job completion time variance (CTV). CTV minimization is used to represent service stability, since it means that jobs are completed in a relative concentrated period of time. CTV minimization also conforms to the Just-in-time philosophy. Two scheduling problems are studied on multiple identical parallel machines. The one problem does not restrict the idle times of machines before their job processing, while the other does. For these two scheduling problems, desirable properties are explored and heuristic algorithms are proposed. Computational results show the excellent performances of the proposed algorithms. The third scheduling problem in the first study is considered on a single machine and from the users’ perspective rather than the system’s perspective. The performance measure is thus class-based completion time variance (CB-CTV). This problem is shown to be able to be transformed into multiple CTV problems. Therefore, the well-developed desirable properties of the CTV problem can be applied to solve the CB-CTV problem. The tradeoff between the CB-CTV problem and the CTV problem is also investigated. The second study deals with scheduling coordination in a supply chain, since supply chain coordination is increasingly critical in recent years. Usually, different standpoints prevent decision makers in a supply chain from having agreement on a certain scheduling decision. Therefore conflicts arise. In pursuit of excellent performance of the whole supply chain, coordination among decision makers is needed. In this study, the scheduling conflicts are measured and analyzed from different perspectives of decision makers, and cooperation mechanisms are proposed based on different scenarios of the relative bargaining power among decision makers. The cooperation savings are examined as well.