Ahmed Azab

Semi-Generative Macro-Process Planning For Reconfigurable Manufacturing

Global competition and frequent market changes are challenges facing manufacturing enterprises at present. Manufacturers are faced with new unpredicted modifications at the part design level, which require increased functionality at the system design level. Reconfigurable Manufacturing Systems (RMS) addresses this situation by providing the exact capacity needed when needed. Process planning concepts and methods should be developed to support this new manufacturing environment. Variant process planning systems with their rigid definition of the boundaries of part families do not satisfactorily support Reconfigurable Manufacturing Systems. A semigenerative macro process planning system has been developed and is reported in this paper. Precedence graphs, which depict the precedence relationships between features/operations, are reconfigured by adding and removing nodes. The problem of generating optimal macro-level process plans is combinatorial in nature and proven NP-hard. Hence, a random-based heuristic based on Simulated Annealing is tailored for this problem. Finally, a realistic case study is presented to illustrate the proposed methodology. A family of single-cylinder front covers is used. The proposed method produced good quality optimal solutions and is proven efficient in terms of computation time as demonstrated by the obtained results.

Mathematical modelling and a meta-heuristic for flexible job shop scheduling

This study develops new solution methodologies for the flexible job shop scheduling problem (F-JSSP). As a first step towards dealing with this complex problem, mathematical modellings have been used; two novel effective position- and sequence-based mixed integer linear programming (MILP) models have been developed to fully characterise operations of the shop floor. The developed MILP models are capable of solving both partially and totally F-JSSPs. Size complexities, solution effectiveness and computational efficiencies of the developed MILPs are numerically explored and comprehensively compared vis-à-vis the makespan optimisation criterion. The acquired results demonstrate that the proposed MILPs, by virtue of its structural efficiencies, outperform the state-of-the-art MILPs in literature. The F-JSSP is strongly NP-hard; hence, it renders even the developed enhanced MILPs inefficient in generating schedules with the desired quality for industrial scale problems. Thus, a meta-heuristic that is a hybrid of Artificial Immune and Simulated Annealing (AISA) Algorithms has been proposed and developed for larger instances of the F-JSSP. Optimality gap is measured through comparison of AISA’s suboptimal solutions with its MILP exact optimal counterparts obtained for small- to medium-size benchmarks of F-JSSP. The AISA’s results were examined further by comparing them with seven of the best-performing meta-heuristics applied to the same benchmark. The performed comparative analysis demonstrated the superiority of the developed AISA algorithm. An industrial problem in a mould- and die-making shop was used for verification.

Change in Manufacturing - Research and Industrial Challenges

Change, in products and systems, has become a constant in manufacturing. This sector continues to witness major market shifts, introduction of new materials and processing technologies as well as great changes in consumer preferences and productsvariety. This presents significant challenges to industrialists, researchers and educators alike. Changes can most often be anticipated but some go beyond the design range. This requires providing innovative change enablers and adaptation mechanisms to mitigate the effects of, and capitalize on, changes in manufacturing. Experiential research and learning is essential to enhance responsiveness and prepare production leaders of the future. The training and experimentation could be significantly enhanced if manufacturing systems could be brought into the laboratories of academic and research institutions. This paper describes the latest state-of-the art fully reconfigurable “plug & play” changeable and flexible “Factory-in-the-Lab” infrastructure and supporting design innovation and advanced research environment. It discusses the use of the iFactory and iDesign system to address these challenges and develop many key technologies and strategies for success. It presents severa novel approaches and methods for achieving the desired balance in the wide spectrum of variation in markets, products, processes, systems and manufacturing enterprises.

Feature-based hybrid inspection planning: A mathematical programming approach

Intelligent planning for inspection of parts with complex geometric surfaces using contact or non-contact devices is still a major challenge. Contact measurement is widely used in manufacturing owing to its superiority in point accuracy. However, the volumetric accuracy of non-contact measurement techniques is better owing to the large number of points that can be measured in a short time. Consequently, contact measurement is usually used for mechanical parts with prismatic shapes while non-contact measurement methods are mostly used with free-form shapes. Complex parts that include both prismatic and free form shapes may require inspection using both techniques. It may not be possible to fully digitise a part using a single type of sensor owing to occlusion or accessibility issues. This paper proposes a hybrid (contact/non-contact) inspection planning approach that capitalises on the advantages of both inspection techniques. In the beginning, a knowledge-based system has been developed for selecting the most suitable sensor for the inspected features using a proposed inspection-specific features taxonomy. Additionally, a new travel salesperson problem (TSP) formulation has been developed for sequencing of hybrid inspection tasks, where a novel sub-tour elimination constraint has been formulated. The proposed 0–1 integer mathematical model minimises the non-digitisation related time between successive inspection operations. The developed hybrid inspection planning system not only overcomes the incompleteness of information when each sensor type is used separately, but also improves the accuracy of the point cloud obtained by using both sensors. The developed hybrid inspection planner was applied to the inspection of a water pump housing of an automotive engine. The applicability of the developed inspection planning framework and methodology for inspecting complex parts and associated dies and moulds has been demonstrated. The developed system makes it easier to utilise the powerful combination of measurement sensors while automating the development of an optimal sensor-task assignment as well as inspection tasks sequence. It has the potential benefits of reducing the inspection time and cost and increasing the efficiency of the whole inspection process, which have positive effects on the early product development stages and quality assurance activities.

Modeling and heuristics for scheduling of distributed job shops

This paper deals with the problem of distributed job shop scheduling in which the classical single-facility job shop is extended to the multi-facility one. The mathematical formulation of the problem is comprehensively discussed. Two different mixed integer linear programming models in form of sequence and position based variables are proposed. Using commercial software of CPLEX, the small sized problems are optimally solved. To solve large sized problems, besides adapting three well-known heuristics, three greedy heuristics are developed. The basic idea behind the developed heuristics is to iteratively insert operations (one at each iteration) into a sequence to build up a complete permutation of operations. The permutation scheme, although having several advantages, suffers from redundancy which is having many different permutations representing the same schedule. The issue is analyzed to recognize the redundant permutation. That improves efficiency of heuristics. Comprehensive experiments are conducted to evaluate the performance of the two models and the six heuristics. The results show sequence based model and greedy heuristics equipped with redundancy exclusion are effective for the problem.

Reconfigurable Control Structure for Robots in Assembly

The increased use of changeable characteristics in modern manufacturing and robotic systems and applications call for improved system control design that offers some degree of reconfigurability. The need for control reconfiguration of robotic systems arises due to some inherent characteristics of the robotic system, variations of robot parameters due to environmental changes, major task changes typical in production changeover or manufacturing system reconfiguration, or geometry changes due to the reconfiguration of modular manipulators. In this paper, a reconfigurable controller, the Supervisory Control Switching System (SCSS), is proposed to meet the new on-line demands for changeability in robotic systems. The SCSS is capable of selecting the most suitable controller for a particular task or situation, from separate controllers designed a priori. The applicability and effectiveness of the developed switching control scheme have been illustrated through computer simulations of an AdeptOne SCARA manipulators carrying out assembly tasks.

Mathematical modeling and hybridized evolutionary LP local search method for lot-sizing with supplier selection, inventory shortage, and quantity discounts

Abstract This paper addresses the multi-period inventory lot-sizing problem with supplier selection and inventory shortage, and it considers both all-units and incremental quantity discounts. A unique preprocessing approach is introduced that transforms discount quantity intervals into newer ones, revealing the supplier that has the minimum total ordering, purchasing, and transportation costs. This transformation changes the lot-sizing problem with multiple quantity discount models into a problem of a single quantity discount schedule. The problem is formulated as a Mixed Integer Non-Linear Programming (MINLP) model. Since the problem is intractable, a hybridized search method is developed, where both an Evolutionary Algorithm (EA) and a Linear Programming (LP) driven local search are combined. For initialization, Wagner-Whitin (WW), back-shifting and relaxed LP approaches are used. Finally, for validation and justification purposes, test cases from the industry and literature are used.

Reconfiguring Process Plans: A New Approach to Minimize Change

In a customer driven market, the increasing number of product variants is a challenge most engineering companies face. Unpredictable changes in product design and associated engineering specifications trigger frequent changes in process plans, which often dictate costly and time consuming changes to jigs, fixtures and machinery. Process Planning should be further developed to cope with evolving parts and product families, increased mass customization and reduced-time-tomarket. Agility and responsiveness to change is important in process planning. The current methods do not satisfactorily support this changeable manufacturing environment. They involve re-planning or pre-planning, where new process plans are generated from scratch every time change takes place, which results in production delays and high costs due to consequential changes and disruptions on the shop floor. The obvious cost, limitations and computational burden associated with the re-planning/pre-planning efforts are avoided by the developed methods. A novel process planning concept and a new mathematical programming model have been developed to genuinely reconfigure process plans to optimize the scope, extent and cost of reconfiguration and to overcome the complexity and flaws of existing models. Hence, process planning has been fundamentally changed from an act of sequencing to that of insertion. For the first time, the developed methods reconfigure process plans to account for changes in parts’ features beyond the scope of original product families. A new criterion in process planning has been introduced to quantify the extent of resulting plan changes and their downstream implications. The presented method was shown to be cost effective, time saving, and conceptually and computationally superior. This was illustrated using two case studies in different engineering domains. The developed hypothesis and model have potential applications in other disciplines of engineering and sciences.

Optimal Sequencing of Machining Operations for Changeable Manufacturing

Process planning is a key enabler of the various production planning activities, that lies at the core of current state-of-the-art product lifecycle management key components. Operations sequencing is the first and main step to take place at the macro planning level, and is defined as the problem of sequencing a global set of machining sub-operations required to machine a certain part in order to minimize changeover time/cost while satisfying a number of precedence constraints. The problem is well known to be NP-complete. An integer based model has been suggested to minimize the changeover time between successive sub-operations, and was used as the basis for the developed evolutionary metaheuristic. A variant of the canonical Genetic Algorithms is developed, where tailored genetic operators specific to the problem at hand were used. A greedy algorithm is developed for initialization of the genetic population. The developed approach was applied to a benchmark problem of varying geometry that requires different machining configurations; results proved to be superior in terms of quality of solutions.

Mathematical modeling and multi-start search simulated annealing for unequal-area facility layout problem

Abstract In this paper, a mixed integer nonlinear programming model (MINLP) is formulated to allocate the position of a number of unequal-area rectangular facilities within the continuum of a planar plant site with a predetermined fixed area. Facilities have predetermined dimensions and are not orientation-free. A continuous approach to the problem is taken. Constraints are developed to eliminate the possible overlap between the different facilities. The model accommodates for aisles, whether vertical or horizontal, as well as blocks and preference locations, where no facilities are allowed to be placed. The problem seeks to minimize total material handling the cost. Four test cases including one from the local industry is used to justify the developed model. The problem at hand is computationally intractable; hence, a novel Simulated Annealing (SA) algorithm is developed to solve large instances of the problem. A unique heuristic algorithm is used for initialization. A multi-start search mechanism is implemented to increase the diversity and mitigate the chances of getting entrapped in local optima. For validation, a group of benchmark problems is being used.