Hans Holm

Product manufacturability control for concurrent engineering

Abstract Design for manufacturability (DFM) is to ensure that a design is inherently economic to manufacture. This economy of manufacture is realized by controlling the manufacturability (M) of the design. Therefore, the measurement and control of M become the essential prerequisites for implementation of DFM. This paper proposes an architecture of an M control system comprising of five functional modules, which can realize multi-feedback control of M by using two types of M measurement methods: process-driven M calculation and design-driven M estimation. For the later, the artificial neural network (ANN) technique has been applied to inferentially estimate M under the condition that no complete product information is available in the early stages of design. A hierarchy of M criteria has been proposed systematically and a product model proposed for recording product information in the process of design.

A reference architecture for the design of an industrial temperature feedback welding control system

Abstract This paper describes a provisional not fully developed architecture of a thermal feedback system for closed-loop control of robotic arc welding. The architecture is based on an off-line planning system for welding and an on-line welding control system, where the off-line planning system builds the reference to be followed by the on-line closed-loop system. The main aim of this architecture is to develop a method by which thermal feedback welding control systems can be designed. This control system is meant for control of welding of thick-walled sheet metal geometries and the system feeds back the temperatures on the surface of the geometry. The temperature is fed back by an infrared CCD-array camera.

Real-time control system for automatic one-of-a-kind production: specification and implementation of a test bed

A control system design method for one-of-a-kind production is proposed, and a control system based on this method is specified. Using this method, the sequencing of prioritized production orders can be controlled and the state of the individual order can easily be derived. The control structure consists of a set of generic controllers with interfaces that are independent of the particular application. Only the control system structure is application dependent in a distributed environment comprising several computing entities connected in a network. A graphical interface has been established for the purpose of testing the prototype control system.<<ETX>>

Model based control of a one degree of freedom workpiece manipulator for welding of nozzles

This paper presents an automatic control system for welding the root-pass in a seam joining a nozzle perpendicular onto a large diameter pipe. The system is based on the use of a geometry model of the welding seam and on inverse welding process models (i.e. models that map desired weld quality, joint shape, and equipment parameters onto appropriate welding control variables). By using these models a workpiece manipulator is controlled in such a way that an appropriate welding position (determined from the inverse process models) is obtained while welding. A robot, that is carrying the welding torch, is then positioned so that the position and orientation of the torch relatively to the seam become as specified in the inverse welding process models. The paper presents the overall concept of the system. Furthermore, algorithms for controlling the workpiece manipulator and the robot are presented.

Model of Root-Bead Welding for Off-line Programming and Control

In this paper, a model is presented which, based on a description of the shape of a welding seam (the gab size and bevel angles), can be used to select variables controlling the welding of a root-bead. The model can be used in an off-line programming system as well as in a geometry-sensor based control system.

A New Heuristic Approach to the Robot Free-Path Problem by Stepwise Path Improvement

Abstract The wide variety of robotic applications of the Find-Path Problem calls for a generalized approach. The general solution should be practically implementable in a broad range of cases. This paper reports the progress made by the CIMLAB group at the University of Aalborg towards the solution of this problem. A general Free Path Solver Module was developed, which can be integrated with any system for robot programming and simulation. The paper describes the overall considerations taken for the solution of the problem, with a detailed description of the algorithm, and presents as a special case of its implementation the problem of a 2-D polygonal robot which moves amidst 2-D polygonal obstacles. The developed algorithm solves the problem in (joint) configuration space. The concept “pseudodistance”, which is a function (PD(C)) which expresses the relative positioning between the robot and its environment at any configuration (C) (also from those which bring about collision). The properties of a function (PDL(t)), which describes the behaviour of the function PD(C) along a hyperline in the configuration space traversed by a travelling robot were also used. By properly capturing the concepts of adaptive discretization of configuration space, approximate description of free space, stepwise path improvement, heuristic-driven goal search and geometrical coherency, a general implementable Find-Path Algorithm was developed.

Information structures for design of feedback resource distribution control systems for discrete part manufacture

Increased pressure for efficiency improvement on discrete part manufacture justifies investment in design and implementation of control systems for distribution of resources in a manufacturing plant. Design of such systems is, however, to day based on intuition. In the paper a systematic method for design of such control systems is described. The paper presents the generic structure of such control systems including definition of input, state and output vectors. Definition of the set of events, which cause choice of input and corresponding transition from one state into another one, is presented. Additionally the paper presents a method for identification and representation of the degrees of freedom, which are available for the choice of system input in reaction to valid classes of appearing production events. The paper finally presents the structure of the state space of the control system and the principle according to which it is expanded and searched for optimal control input.

Closed loop scheduling and control of One-of-a-Kind Production

In this paper, a closed loop cybernetic system is proposed for real-time scheduling and control of production in an automatic shop floor or on an automatic production line for One-of-a-Kind Production. By a concept of the so-called Product Production Structure, the designs and production processes of different types of products can be formulated into a set of data. These data can be ‘read’ by the control system and thereby the control system makes schedule of as well as real-time controls the production of the products. In this manner, the control system gains enough robustness to cope with the great diversity of product types (or kinds) in One-of-a-Kind Production. By the simulation module and experience time estimates of production processes, the control system is designed as an estimator to be able to make the production schedule before starting a production of a product or a batch of products with different types. By its closed loop, the control system is also designed to be able to real-time control the production and update the production schedule according to states of a shop floor or a production line during the production.

Comparison of rule-based and optimization-based control of resource allocation on a production line

The effect of disturbances to task execution in manufacturing cells can be minimized by control. A rule-based and an optimization-based controller is developed for a production-line at Odense Steel Shipyard and implemented in a computer-based simulation model. The differences in performance between the two controllers is found in a two-level five factor factorial experiment. The simulations show that on average the optimization-based controller is 2-5% better than the rule-based controller. It is also shown that the rule-based controller is generally more sensitive to changes in parameters on the production-line.

Hybrid state plant model constraints for design of control systems for discrete part manufacture resource distribution

Material- and product transformations of discrete part manufacture are performed through sequences of continuous processes. The sequences of the transformations are in most humanly operated manufacture today determined partly by conscious planning partly by non conscious planning. Basis of the planning is deterministic transformation durations. This paper establishes a rational system framework upon which rational methods for design of systems for the control of distribution of resources for discrete parts manufacture can be developed.