Latif Salum

The cellular manufacturing layout problem

Most of the techniques for the cell design (CD) problem in cellular manufacturing (CM) consider only the machine-part incidence matrix as the input. They then convert this matrix into a block diagonal form to create a dedicated and independent cell for each part family, without laying out machines. Yet due to the lack of information on the layout, hence the lack of information on the performance of the shop floor, the benefits of CM cannot be validated. Furthermore, operational issues, e.g. scheduling, have an effect on the CD. However, current techniques are not comprehensive enough for a complete CD which should consider the aforementioned issues as a whole. This paper proposes a two-phase method based on total manufacturing lead time (MLT) reduction to overcome these drawbacks. In the first phase, the system is simulated by considering all of the operational issues under the assumption of zero material handling times to minimize total MLT. Besides minimized total MLT, the first phase also yields the waiting times of parts and the volume of the parts-flow between machines, which are used to find similarity measures between machines in the second phase. The second phase then exploits an algorithm which creates and uses these similarity measures to construct a layout by locating machines with higher similarity next to each other to justify the assumption of minimal total MLT and to minimize total material handling time.

A multi-criteria adaptive control scheme based on neural networks and fuzzy inference for DRC manufacturing systems

Manufacturing systems are uncertain and dynamic systems, hence, they require real-time scheduling to adapt to changing manufacturing conditions. Current real-time scheduling approaches have been devised mainly for machine-only constrained systems, in which the shop capacity is constrained only by machine capacity, rather than for dual resource constrained (DRC) systems, in which the shop capacity is constrained by machine and worker capacity. In particular, there is no study on DRC system scheduling in which the ‘where’ and ‘when’ worker assignment rules, basic features of DRC systems, are altered in real-time (dynamically selected) to respond to new manufacturing conditions. Besides, multi-criteria DRC system scheduling has not yet been addressed extensively. Also, there has been little research on the interactions of dynamically selected job dispatching, worker assignment and job routing rules, which have a significant impact on DRC system performance. This paper proposes a multi-criteria real-time scheduling methodology for DRC systems to address these issues, and investigates these interactions. The methodology employs artificial neural networks as meta-models to reduce computational complexity and a fuzzy inference system to cope with multiple performance criteria. Various simulation experiments demonstrate that the methodology provides satisfactory results for real-time DRC systems scheduling.

Using the when/where rules in dual resource constrained systems for a hybrid push-pull control

This paper proposes a production control system, DRC-HPP, which uses the when/where rules in dual resource constrained (DRC) systems for a hybrid push-pull (HPP) control, to overcome some difficulties in modelling/implementing DRC/Kanban systems. These rules and the novel ‘process-or-transport’ and ‘whereto’ rules are embedded in some policies workers use to decide when to process (transport) parts, and where (whereto). Unlike most control systems, in which a group of workers is always responsible for transporting and another group is always responsible for processing parts, workers in DRC-HPP are responsible for both transporting and processing parts, as in the Toyota Sewn Products Management System (TSS). Yet, unlike TSS, DRC-HPP can be applied in any layout type. Workers transport parts when they are idle in part processing to enhance their utilisations and synchronise transportation. Since the transportation does not require special worker skills, the cost of training workers is not incurred. DRC-HPP is compared with different benchmarks through simulation experiments to evaluate its performance. It performs well under relatively short transportation times with respect to processing times. If they are relatively longer, the issue becomes to determine the number of workers to achieve a performance level. DRC-HPP also facilitates bottleneck management.

Rule-based modeling and constraint programming based solution of the assembly line balancing problem

Highlights? Instead of using precedence graphs, a rule-based assembly model is proposed to represent all possible assembly sequences of a product. ? The proposed rule base simultaneously chooses the assembly plan and balances the assembly line. ? A rule-based model is more effective in balancing the assembly line while addressing alternative assembly plans. ? It is easier to map a rule-based model to a constraint programming model than to an integer programming model. ? Constraint programming is more successful compared to solution quality and time than integer programming. The assembly line balancing problem employs traditional precedence graphs to model precedence relations among assembly tasks. Yet they cannot address alternative ways of assembling a product. That is, they only model conjunctions, not disjunctions. Moreover, some additional constraints need also to be considered, but these constraints cannot be modeled effectively through precedence graphs, e.g., constraints indicating certain tasks cannot be assigned into the same station. To address these issues, this paper proposes to model assembly constraints through the well known If-then rules, and to solve the rule-based model through constraint programming (CP), as CP naturally models logical assertions. The paper also shows how to map a rule-based model to a CP or an integer programming (IP) model. Finally, a computational experiment is carried out to analyze the performances of CP and IP models with respect to modeling capability, solution quality and time. The results reveal that CP is more effective and efficient than IP.

Modern QFD-based requirements analysis for enterprise modelling: enterprise-QFD

Competitiveness and globalisation force enterprises quickly to adapt to changing conditions of markets. Enterprises employ modelling methodologies to organise their strategic knowledge to cope with this change, which results in an enterprise (data) model. Requirements discovery and analysis is the most important phase in creating the enterprise model because any mistake in the requirements discovery deteriorates the validity of the model, resulting in user dissatisfaction. Quality function deployment (QFD) is a well-known and integrated approach used in converting the requirements of users into final product specifications. This paper modifies QFD for enterprise modelling, and proposes Enterprise-QFD, which provides a common platform to be integrated with any methodology for discovering and analysing customer requirements. The modified QFD tables involve some required columns added and unnecessary ones deleted based on enterprise modelling. A novel matrix content and sequence is also proposed.

Avoiding state explosion in a class of Petri nets

A novel tool, superposition chain, is proposed to avoid state explosion.This can be used in some classes of Petri nets.The future research direction is to incorporate the idea into the acyclic safe PNs.As the reachability problem of an acyclic safe PN is NP-complete, the effect would be immense.A long way has been passed in this direction. This paper introduces leveled Petri nets (PNs), and proposes a novel PN analysis tool, the superposition chain (SC), to avoid state explosion. It also introduces underlying tools-superposition and the leveled token game-to tackle the P vs NP problem, a well known problem in CS/AI community. The leveled token game, defined over a leveled PN, generates the SC of the PN. The leveling is based on the transitions such that a transition and all its input places are in the same level, and that there is no causality among transitions in a level, while transitions across levels indicate causality. The enabling rule is extended by superposition and firing history. Superposition of markings is defined by a set ź M of places p marked in superposition, and denotes that each p in ź M is marked individually, yet it is uncertain if all p in ź M are marked together. In other words, superposition loses which p in ź M is marked by conflicting transitions, which are revealed by the transition firing history. The firing history of p ź ź M is also defined by a set, h ( p ) , and denotes transition firings participated in p ź ź M , yet does not enumerate their firing sequences to avoid the state explosion. Then, the compound firing history defined over ź M , h ( ź M ) , is used to reveal all conflicting transitions participated in ź M . Hence, ź M is not coverable as a whole, if there are conflicting firings in h ( ź M ) , which is used for the transition enabling. Consequently, the SC, generated by the leveled token game, specifies the PN behavior, as a reachability tree, generated by the (conventional) token game, also specifies a PN behavior.

Rule-Based Modeling of Assembly Constraints for Line Balancing

The assembly line balancing problem employs traditional precedence diagrams to model precedence relations among assembly tasks. Yet they cannot address alternative ways of assembling a product. That is, they only model conjunctions (AND relations), not disjunctions (OR relations). However, there may be some alternative assembly plans for a product. Moreover, some complicated constraints need also to be considered, but these constraints cannot be modeled effectively through traditional precedence diagrams. This paper proposes a rule-based representation to address these issues. The rule-base alone can model any constraint effectively. The paper also shows that the rule-base can easily be employed by heuristics used for line balancing problems.

A new class of a high-level Petri net for modelling logical OR efficiently: Coloured AND/OR Petri nets (CARPN)

In Petri nets (PN), because all the input places of a transition are connected to each other with Logical AND to enable the transition, Logical OR connections amongst its input places cannot be modelled. In the case there are k alternatives (Logical ORs) to perform a task, e.g. assembly of a product, k transitions are required for modelling these k alternatives in a PN model, each transition representing each alternative. This inflates the number of transitions in the PN when the number of alternatives increases. This paper introduces CARPN to model the same k alternatives with fewer transitions and places. The similarities between PN and CARPN are discussed, and based on these similarities formal analysis methods are introduced, as well as a new type of graphical representation. Because the elements of the incidence matrix of a CARPN are integers, the current techniques to obtain the invariants of PN can also be applied to CARPN. In the case of systems having similar structures, the reduction of the incidence matrix, keeping its elements integers, is also discussed.

Automation of Kanban modelling through object-oriented visual-interactive simulation

Kanban is a control system performing routine material control transactions on a shop floor. Object Orientation (OO) is a modelling approach with properties, e.g. inheritance, encapsulation and reusability. These properties of Kanban and OO have led to the idea of creating object classes performing the Kanban process automatically. Through the application of this idea, potential users do not have to model Kanban separately for their own models. They can model Kanban-controlled manufacturing systems by exploiting the object classes created previously. This enables them to focus on only their own models, without considering Kanban itself, and, consequently, the modelling time is reduced significantly. During a simulation run, Kanban is simulated automatically. An OO and visual-interactive simulation language, SIMPLE++, has been used for modelling. The advantages of this approach over conventional simulation have also been discussed and demonstrated.