Gideon Cohen

Neural networks implementations to control real-time manufacturing systems

Abstract The main objective of advanced manufacturing control techniques is to provide efficient and accurate tools in order to control machines and manufacturing systems in real-time operations. Recent developments and implementations of expert systems and neural networks support this objective. This research explores the use of neural networks to control several manufacturing systems in real-time operations: robot manipulators, tool changes, conveyor systems and machine faults diagnosis. The main barrier to wide implementation of neural networks is the huge computation resources (times and capacities) required to train a network. This research represents the use of a multi-layer architecture of networks (input layer, several hidden layers and an output layer) to define single-valued inter-relationships between system participants and to avoid the need for long training processes. The use of neural networks to control the above-mentioned systems was evaluated from the following parameters: the architectures, network training methods, efficiencies and accuracies of networks to perform the task of control. Several conclusions related to neural network implementations to manufacturing systems were produced: (1) the multi-layer architecture fits the complexity of manufacturing systems; (2) neural networks are efficient to control real-time operations of machines; (3) machines which were controlled by neural networks performed accurate results; and (4) the use of several hidden layers can replace the need for long training processes and saves on computation resources.

Concurrent system to resolve real-time conflicts in multi-robot systems

Abstract Multi-robot systems consist of several robots, machines and processes with strong interrelationships. A multi-robot system is characterized by the dynamics of its transfers from one state to another. Because of limited budgets for investment, multi-robot systems must be based on shared robots and machines. Conflicts occur when a shared robot or a shared machine must serve two or more processes or tasks concurrently. When a conflict occurs it must be resolved rapidly, to minimize its effects on manufacturing floor efficiencies. This paper describes a system based on an expert system, concurrent programming and Petri nets, which resolves conflicts generated by shared robots and machines. This system is referred to as a “concurrent” system. Many computer runs were performed. The results lead to the conclusion that the system under consideration provides significant benefits in real-time operation. The system is defined in generic terms, and can be implemented for a large variety of multi-robot systems.

Expert system to control and to design closed loop conveyor systems

Abstract A conveyor system is an important part of a manufacturing system. As such, the conveyor system must comply with all the requirements of a modern manufacturing system: high flexibility, high efficiency, and high speed—smart reasoning processes to generate future positions based on a given current status. Because a huge number of figures and numerical manipulatuions are associated with the conveyor systems operations, the traditional numerical control techniques cannot satisfy the requirements to initiate and to control operations. New techniques based on very efficient reasoning processes are required. This article discusses an expert system that consists of a knowledge base and an inference engine that was developed to control a converyor system real-time operation. The computer runs that were performed during this research lead to the conclusion that the developed expert system can be employed to control the conveyor system real time operations very effectively. The developed expert system is considered as a reliable simulator of a conveyor system which can be implemented to explore parameters interrelationships at the phase of system design. Computer runs were performed to analyze the interrelationships between operational parameters which characterize the explored conveyor system. The expert system was programmed in a way that provides a generic simulator, which can be employed in a large variety of conveyor systems.

Simulated intelligent flexible cell for the assembly of multi-component systems

Abstract An intelligent flexible cell for assembly is an autonomous, self-contained, highly productive assembly system with the ability to reason about actions and their results. Assembly sequence planning, as considered in this research, is part of a hierarchy of steps in an assembly system for manual, fixed automation, or programmable automation systems. Multi-component systems are represented in this paper by four different pneumatic motors, varying in power capability, size, weight and pattern, and in the method of assembly. The knowledge base of the expert system described contains facts and rules to define the flexible assembly cell and the structures of the pneumatic motors. The inference engine integrates the parameters which define the current situation of the assembly cell, with the rules and the facts in the knowledge base, to determine the next operations to be implemented. The numerical runs performed on a simulated cell address aspects associated with the design and methods of operation of flexible assembly cells: (a) the use of gripper adaptor adjustments; (b) the cost savings resulting from a vision system and (c) various methods of cell operation relating to batching and loading of parts, and release and dispatching operations.

Intelligent vision system to control flexible assembly cell feeding processes

Abstract A vision system which is installed at an assembly cell to detect its feeding system must analyze objects for inspection and assembly operations. This means that a description of the objects must be generated. A particular analytic technique may be characterized by the nature of the formal description generated, by the computation techniques employed and by the degree to which the formal description accurately reflects pertinent information about the object. To execute assembly tasks, the part identification, orientation, spatial position and geometric features were calculated. For inspection tasks, a list of deviations between the observed object and a reference object was generated to determine whether a part could proceed into the assembly cell. Since the vision system is associated with real-time assembly cell operations, its vision data processing time must be minimized. Therefore, intelligent processing techniques must be developed. Such techniques are provided by the developed knowledge-base and inference engine. The described vision system which is incorporated with the knowledge base and the inference engine is referred to as an “intelligent vision system”. The developed intelligent vision system was implemented in a simulation program to inspect the feeding systems of pneumatic motor assembly cells. The results support known interrelationships between parameters associated with detected assembly processes, and verify the reliability of the developed inference engine reasoning algorithms. Since the simulation was programmed in generic terms, it is not limited to a specific product but can be applied to a large variety of products.

Using AI techniques to optimize manufacturing shop-floor operations

Abstract The modern manufacturing shop-floor is characterized by its dynamic nature, and by the need to achieve high efficiency. The application of artificial intelligence techniques to aid data-processing and decision-making must address the dynamics and the requirements of high efficiency. This paper describes the knowledge-base and inference-engine which were developed to determine various shop-floor operational strategies to achieve maximum efficiency of machines. The Shortest Processing Time (SPT) strategy was found to optimize the operation of a shop-floor. The second purpose of the knowledge-base and inference-engine was to consider the effects on shop-floor efficiency of unpredicted obstacles. Unpredicted obstacles which may reduce shop-floor efficiency result from poor shop-floor organization. The purpose of this paper is to improve shop-floor organization so as to reduce the number of unpredicted obstacles, and to prevent further decreases in efficiency. The knowledge-based and inference-engine which were developed refer to a ‘shop-floor’ defined in generic terms. Modular function blocks were implemented to describe the generic shop-floor. Since generic terms were implemented, the knowledge-base and inference-engine are not limited to a specific shop-floor, but may be implemented in a large variety of shop-floors.

Intelligent Jig System to Automate Flexible Manufacturing System

Jig pallet systems are intended for the automatic, complete machining or assembly of parts families in the area of medium and large scale manufacturing. Their distinctive feature is that several machine tools, or assembly machines, are linked together to generate an overall system by means of common tool and workpiece supply with integrated computer control. A jig pallet system is considered to be intelligent, if its central processor is equipped with a knowledge-base and an inference-engine. A jig pallet system was structured. It consists of a central processor, tools supply system, workpiece supply system, manufacturing cell which includes four work stations and a local area network. A knowledge-base and inference-engine were developed to reason the next position of jig pallet systems. The jig pallet system, with its incorporated knowledge-base and inference-engine, was tested for a large variety of operational parameters to explore the ability of the central processor to control participants operations. The conclusion which is derived from these tests is that the central processor can control and optimize participants operation in real time with minor effects on the system efficiency.

Optimal architecture configuration of manufacturing floor control systems

A manufacturing control system configuration is defined by the number of parallel assembly cells slaved to a specific superior microprocessor, their superior microprocessor, and the associated microprocessors performance capacities. The above‐mentioned items of configuration were explored in this research to derive their effects on the efficiencies of manufacturing cells. The manufacturing cells are represented in this research by pneumatic motors assembly cells. A knowledge base and inference algorithms were developed and programmed to explore the effects of a large variety of architecture configurations on cells efficiencies. The achieved numerical results lead to the conclusion that exponential correlations characterize the interrelationships between assembly cell unutilization and architecture configurations parameters. Since exponential correlations characterize the interrelationships between assembly cells unutilization and architecture configurations parameters, an optimal architecture configuration is defined by three conditions, which match the beginning of the flattened parts of the exponential curves: (1) If the number of slaved assembly cells is considered, the optimal configuration is achieved when an additional slaved assembly cell results in a most significant reduction in assembly cells efficiency. (2) If the number of superior microprocessors is considered, the optimal configuration is achieved when an additional superior microprocessor does not result in a significant reduction in assembly cells unutilization. (3) If microprocessors capacity performance (bits/sec) is considered, the optimal configuration is achieved when additional capacity performance cannot reduce the assembly cells unutilization. Computer runs were performed to evaluate the optimal configurations. The developed knowledge base and inference algorithms were defined and programmed in generic terms and can be implemented in a large variety of systems.

Management of development projects: how to meet corporate business goals?

Two contradictory elements affect development projects in the modern era. New advanced development projects require huge resources and long schedules. On the other hand, the resources available for development projects are decreasing, and intensive competition between companies results in increased required performance and new products which must be developed with short schedules. Development projects have a significant effect on a companys business goals, and therefore attract the attention of corporate managers. A significant part of managers’ time is devoted to the control of development projects. Support tools are therefore developed and used to assist managers in controlling their projects. This paper describes a control tool to help managers make decisions on development projects. The tool described is based on Microsofts ‘Project’ software package, and was developed in-house. The package performs real-time calculations and simultaneously presents several control parameters. These parameters provide a comprehensive picture of project status. The major benefits of the control tools described are: (1) The ‘Project’ package provides the ability to work with small and combined work packages. It allows attention to be focused on small programs as well as on entire programs. (2) It complies fully with modern decentralised computer systems consisting of a central computer, local computerised networks, parallel PCs and work stations. It provides an interactive quick response ability to follow changes in project status, which result from design changes or progress updates. The control system was implemented by RAFAELs Missiles Division two years ago, and impressive results in terms of project efficiency were achieved. These achievements play a major role in meeting the Divisions business goals. The control tools described are general and can fit a large variety of projects and companies.

Expert system to match robots and to synchronize their operations to pick and place large parts

The operation of pick and place large parts is a very common operation on the manufacturing floor. The efficiency of the pick and place operation affects the efficiency of the entire manufacturing floor. Large parts in terms of this research are parts which must be transported by two robots.This paper concludes the development of an expert system and a numerical simulation to assign two available robots to perform the pick and place operations of large parts.The developed expert system consists of a knowledge-base structured in interrelated frames to define large manufacturing systems and an inference-engine to perform the robots matching and their performances monitoring. Matching robots considerations can not be performed by numerical algorithms in real time systems, because they are major computation time consumers. Therefore, unmatched robots can be assigned by the numerical algorithms to perform inefficient pick and place operations. The expert system and the numerical simulation were programmed, and many computer runs were performed to explore the efficiency of the two systems in a large variety of manufacturing systems.The achieved results show that the expert system has no advantage in assigning matched robots to perform the pick and place operations in systems characterized by very short machine process times (0.5 moment to 1 moment). If such short machine process times are relevant, the knowledge-base is updated very frequently, and the inference-engine reasoning processes can not be initiated. For longer machine process times, the frequency of knowledge-base changes is reduced, and the inference-engine reasoning processes can be initiated without delays. Therefore, the advantages of the expert system become significant if machine process times are extended.