Showing papers in "The International Journal of Advanced Manufacturing Technology in 1993"

Full-pose calibration of a robot manipulator using a coordinate-measuring machine

Abstract: The work reported in this article addresses the kinematic calibration of a robot manipulator using a coordinate measuring machine (CMM) which is able to obtain the full pose of the end-effector. A kinematic model is developed for the manipulator, its relationship to the world coordinate frame and the tool. The derivation of the tool pose from experimental measurements is discussed, as is the identification methodology. A complete simulation of the experiment is performed, allowing the observation strategy to be defined. The experimental work is described together with the parameter identification and accuracy verification. The principal conclusion is that the method is able to calibrate the robot successfully, with a resulting accuracy approaching that of its repeatability.

Integrated tolerance optimisation with simulated annealing

Abstract: Tolerance is one of the most important parameters in design and manufacturing. The allocation of design and machining tolerances has a significant impact on manufacturing cost and product quality. This article presents an analytical model for simultaneously allocating design and machining tolerances based on the least-manufacturing-cost criterion. In this study, tolerance allocation is formulated as a non-linear optimisation model based on the cost-tolerance relationship. A new global optimisation algorithm, simulated annealing, is employed to solve the non-linear programming problem. An example for illustrating the optimisation model and the solution procedure is provided.

Multiple-objective optimisation of machining operations based on neural networks

Abstract: Metal cutting plays an important role in manufacturing industries. Optimisation of cutting parameters represents a key component in machining process planning. In this paper, a neural network based approach to multiple-objective optimization of cutting parameters is presented. First, the problem of determining the optimum machining parameters is formulated as a multiple-objective optimization problem. Then, neural networks are proposed to represent manufacturers' preference structures. To demonstrate the procedure and performance of the neural network approach, an illustrative example is discussed in detail.

Predicting product quality with backpropagation: A thermoplastic injection moulding case study

Abstract: This application uses live data from a thermoplastic injection moulding manufacturer to examine the feasibility and effectiveness of using backpropagation artificial neural networks for predictive quality control. Preprocessing and post processing of live data, formulating neural predictive strategies, selecting architecture and parameters, and handling of temporal aspects are topics. Performance of the neural networks are compared to other quality control methods, including control charts and statistical techniques. This case study demonstrates that even manufacturers who have modest expertise in computing and limited hardware and software availability can successfully use neural networks for data analysis and modelling.

A matrix approach to tolerance charting

Abstract: This paper presents a unique method of deriving the dimensions and tolerances during tolerance charting. It is simple and practical. Unlike other methods, it does not require any representation of the machining sequence. The dimensions/tolerances between workpiece surfaces are organised into an input matrix. The input matrix then undergoes a series of operations to yield the final matrix. The final matrix contains the resultant dimensions/tolerances between any two surfaces of the workpiece. This method is easy to understand and implement on a microcomputer. An example is used to illustrate the method.

Automated fixture configuration using projective geometry approach

Abstract: The geometric and topological specification of a workpiece boundary is usually represented in a specific data format in a CAD database. To retrieve a set of workpiece data, to analyse its shape in addition to the machining requirements, and to determine the proper fixture configuration accordingly, are not trivial tasks when a part has a complicated shape. The real challenge is to recognise and synthesise the shape of a workpiece from its data representation. Consequently, the decision for fixturing can be made when the shape of a workpiece and the relationship of the shape and the fixturing configuration can be derived by a systematic methodology. In this paper, a projective spatial occupancy enumeration (PSOE) approach is applied as a representational and manipulating scheme for developing algorithms in automatic fixture configuration. The workpiece is projected onto the working plane of the fixture baseplate. A 2D projection is defined as a matrix of cells which can represent a workpiece with an arbitrary shape. Using a discrete search based upon the matrix of cells, the fixture types and their locations are generated according to a set of heuristic algorithms. This work is a generalisation and extension of previous works for prismatic parts. The same methodology is equally applicable in general robot grasping.

Distributed planning of collaborative production

Abstract: Collaborative production is viewed in two related dimensions: collaboration among machines to achieve increased reliability, quality and productivity; and collaboration between machines and human supervisors to combine internal and global control information. A structure, combining centralised planning and human-machine integration, is developed for distributed planning of such collaborative production. The centralised planning relies on game-theoretic optimisation techniques supported by knowledge-based rules for coalition formation of cooperating production operations. The product of the centralised planner is a general, common production plan. The feasibility of this model is demonstrated mathematically. If the centralised planner cannot reach a solution, then human input is sought. For integrating the human and machine knowledge to reach effective solutions, a neural network model is demonstrated. A simulation demonstrates that the distributed human and machine inputs can be combined under this framework. Such a distributed planning structure can overcome the major difficulties encountered in computerised production systems that seek full automation.

Diagnosis of roller bearing defects using neural networks

Abstract: A new approach for the diagnosis of bearing defects has been utilised. Artificial neural networks (ANN) were employed for the diagnosis of various kinds of bearing defects. The features selected for this purpose were: the average of the top five values of amplitude in the high-frequency region (5 kHz-22 kHz), the peak value of the amplitude in the high-frequency region, the average of the top five values in the prime spike region (340 Hz-3262 Hz), the autocorrelation function in the prime spike region, the autocorrelation function in the high-frequency region, and the cepstrum function in the high-frequency region.

A fast spatial representation method (applied to fixture design)

Abstract: A type of cellular-decomposition, solid-model representation based on variable orthorhombic cells and using a threedimensional matrix data structure is presented. It was developed to do two things that other representation methods cannot do easily or quickly. This method directly provides information concerning intersecting components, and it is fast enough to allow checks for intersection to be made in iterative routines. The method has application to layout design where interference checking is important. The method and the essential algorithms are described. An application to fixture design is presented.

Determining the cutting conditions for sculptured surface machining

Abstract: This paper deals with a new method for determining cutting conditions for the machining of 3D sculptured surface shapes with a ball endmill. Since the interaction area between workpiece and cutting tool changes during ball endmill machining, the cutting conditions have to be changed accordingly. To determine the spindle speed, the tool-workpiece contact area is obtained from the depth of cut, path interval, and cutter contact (CC) data. The spindle speed is determined so that the tangential cutting speed of the centroid of the tool-workpiece contact area becomes the value suggested in the machining data handbook. The feedrate is determined so that the material removal rate is maximised within the range of the cutting-force constraint. The practical feasibility of the proposed scheme has been verified through experiments performed on a workpiece with sculptured surfaces. The test result shows that the new method can be used to increase the tool life, and improve the production rate.

Parametric modelling and control of the robotic grinding process

Abstract: The modelling and control of the grinding force for the robotic disc-grinding process are described. Modelling experiments were conducted using an industrial robot equipped with a pneumatic grinder and a force sensor. The models relating the commanded depth of cut to the corresponding normal grinding force was developed based on nine sets of experimental data. These data were collected at three different grinding paths with three different pseudo-random binary sequence amplitudes in order to fully investigate the dynamics of the process. Using the extended-least squares method for recursive parameter estimation, a linear fourth-order autoregressive moving average with exogenous variable model was successfully developed for the robotic disc-grinding process.

Knowledge-based selection and sequencing of hole-making operations for prismatic parts

Abstract: This paper describes the rule structure used to select the appropriate sequence of operations to produce various types of holes in 2 1/2D prismatic components manufactured on machining centres. Operation selection is based on forward chained rules coded in Prolog. A different rule priority sequence is applicable depending on the rule fired last. This solves the conflict resolution problem and ensures reasonably fast execution. Rule conditions perform feature extraction, feature size checks, operation suitability checks, tool suitability checks and processing status checks. Actions perform the processing operations and update the model through feature processing tags. The knowledge base implemented for hole-making was not obtained from a practical industrial environment. Nevertheless it does produce sensible plans and it is relatively easy to modify.

A user-friendly, interactive expert system for the injection moulding of engineering thermoplastics

Abstract: An expert system for the injection moulding of engineering thermoplastics has been developed. The system performs off-line, on the factory floor, and can resolve operating deviations interactively with the user. The expert system is hierarchical and offers the corrective action which is easiest to achieve, rather than the most likely. This allows the operator the facility to attempt resolution of the deviation first, much the same as would occur currently. If the deviation persists then material changes will be offered. Finally, if the deviation is still present then mould changes will be offered. The corrective actions offered by the expert system are bounded by the recommended operating condition windows, but this restriction may be overridden by the user. Existing software packages are used to supplement the decision-making processes.

From fuzzy input requirements to crisp design

Abstract: Design is a highly knowledge-intensive and ill-structured problem. Most of the practical problems we face in a design are either too complex or too ill-defined to analyse with conventional approaches. By defining functional requirements as a fuzzy set, one can perform inexact reasoning during the conceptual or creative phase of the design process.

Path and orientation accuracy of industrial robots

Abstract: The automotive industry uses mainly on-line programming, i.e. teach-in programming, of welding and bonding robots for path planning purposes. This programming method is very time-consuming and expensive, and short-term program modifications cannot be easily performed. Computer-aided simulation of the working program and subsequent off-line programming seems to be one possible solution. Unfortunately, off-line versus on-line programmed paths entail completely different robot perfomances. As robot manufacturers are unable to provide data on the off-line path accuracy of their products, this type of programming is not suited for practical operation. This problem will be discussed on the basis of directive VDI 2861, and for the first time results of off-line and on-line path measurements will be compared.

Neural-networks-aided fault diagnosis in supervisory control of advanced manufacturing systems

Abstract: Fault diagnosis presents considerable difficulty to human operators in supervisory control of advanced manufacturing systems. A neural network which employs a symptomatic search strategy by pattern recognition has been developed to aid human operators in supervisory control of a simulated advanced manufacturing system. The abilities of learning-by-example, knowledge generalisation, and parallel processing of neural networks were examined, resulting in the superior performance of neural networks in comparison to expert systems and humans in terms of training, knowledge generalisation, knowledge updating, and taks performance. A more complex system of neural-networks-aided fault diagnosis, employing symptomatic search by hypothesis and test has been designed for the economy of collecting and using symptomatic information. Finally, it is suggested that an integrated supervisory control system with computer-aided fault detection and neural-networks-aided fault diagnosis will make advanced manufacturing systems more attractive in terms of increased productivity as well as improved human jobs.

A shape feature recognition framework and its application to holes in prismatic parts

Abstract: A feature recognition framework is proposed, consisting of a feature definition part and a feature matching part and operating on a B-rep model. Feature definition consists of a unique name, of a feature template yielding feature geometry, and of a set of parameter-extracting procedures yielding shape parameters. A feature template is defined in terms of building blocks and their relationships and properties. Building blocks are defined mainly in terms of profiles, their outward indexes, the types of their edges (solid/void) and their orientation. Features are classified as main and auxiliary, the former being further classified as unique and non-unique. Feature uniqueness is ensured by using key blocks that do not share elements with other feature blocks. Patterns of identical features are also identifiable. These concepts were implemented in Prolog for hole features and tested for real engineering components.

Thermometric design considerations for temperature monitoring in machine tools and CMM structures

Abstract: Temperature measurement is an essential element in monitoring and studying the thermal deformation response of machine tools and CMM structures. In this context, temperature measurement is required for model verification, for control purposes and for determining the thermal contact resistance and other thermal boundary conditions. In these applications, where heat flux and temperature rise levels are relatively low, a proper thermometric design is a must to ensure accurate and precise results. This paper addresses the problem of thermometric design in the case of temperature measurement inside a solid body and on its surface. Results and recommendations regarding the interactions between neighbouring thermocouples, the effect of the distance between the thermocouple and the contact interface (i.e. the heat input surface), and the effect of heat flow along the thermocouple wires on the measurement error are presented. The issue of the effect of the unknown thickness and thermal conductivity of the surface paint on the uncertainty in surface temperature measurement is also addressed.

Automatic generation of group technology families during the part classification process

Abstract: Initial part family formation and subsequent part classification are two important problems to be addressed in applying the group technology principle. Although these two problems are closely related, they have been treated separately. As an aggregate problem, the automatic creation of new part families during the classification process, is investigated. A two-layer neural network using the adaptive resonance theory is adopted. The capability of this neural network model of dealing with the stability-plasticity dilemma is utilised in classifying the parts into families and creating new families if necessary. A heuristic algorithm using the neural network is described, with illustrative examples.

Tool path planning for NC milling of convex polygonal faces: Minimisation of non-cutting area

Abstract: Face milling is one of the most important machining operations in industry for producing flat surfaces. Substantial research has been conducted in the field of optimisation of machining parameters for face milling, without considering the impact of tool path selection. Choosing an efficient tool path can reduce the tool travel time considerably.

An overview of JIT

Abstract: The term JIT (Just in Time), when used as a manufacturing control system, has become very confusing over the past few years. Everyone wants to claim they are users of JIT systems and because of this a variety of systems creations, all claiming themselves to be JIT systems, have sprouted up, none of which look anything like what Japan uses. This article attempts to clear up some of this confusion by showing four different variations of JIT and comparing them to the traditional MRP/MRP II environment.

Burr Detection by Using Vision Image

Abstract: In this paper, a practical force model for the deburring process is first presented. It will be shown that the force model is more general than Kazerooni's model and it is suitable for both upcut and down-cut grinding. In terms of this force model, an algorithm of burr detection by using a 2D vision image is proposed. In the burr detection algorithm, the relevant data of burrs, such as frequency, cross-section area, and height are simplified so that they are functions of the burr contour only. Then, a fast tracking method of the burr contour (BCTM) is developed to obtain the contour data. Experiments show that the BCTM of this passive (i.e. without lighting) image system can be as fast as 18.2 Hz and its precision is 0.02 mm, so online burr detection and control by using the vision sensor is feasible.

Modelling and forecasting industrial innovations via the transfer function S-shaped learning curve

Abstract: Algorithms are derived for the four-parameter transfer function S-shaped curve, using a least-squared-error (LSE) method for an exponential function. The S-shaped curve is just one in a family of industrial dynamics learning-curve models of increasing complexity which may be used to replicate and forecast the start-up of industrial innovations.

A PC-based system integrating CMM and CAD for automated inspection and reverse engineering

Abstract: A personal computer (PC)-based system integrating a manual CMM and a CAD system has been developed for on-line capture of 3D point data, resulting in the automation of the inspection process. The sequence of steps taken to measure a master component using the CMM is stored in a measurement program in ‘teach mode’ and replayed for repetitive measurement of a batch of identical components in the ‘replay mode’. The measurement program guides the CMM operator regarding the geometric entity to be measured, the number of points to be input and the direction to be followed for taking the points on the CMM. The operator has simply to follow the sequence displayed on the computer screen. The program automatically computes the dimensions and the deviations from the corresponding dimensions of the master component. The 3D measurement data from the CMM are transferred to a CAD system in real time. Programs have been developed to create 3D cubic splines and surfaces from the 3D point data taken on the CMM. The full features of the CAD software can be used to manipulate the 3D point data for CAD applications.

A neural information retrieval system

Abstract: Group technology is an approach to manufacturing that attempts to enhance production efficiency by grouping similar activities and tasks together. The results of this process are then used in the execution of similar tasks and activities. This concept can be applied to a variety of activities such as design retrieval, purchasing, sales, and process planning [1]. Traditionally, classification and coding has been used to implement group technology. In this paper, however, we discuss a novel approach using neural networks, a technology noted for its powerful pattern-matching capability. Although this approach can be applied to the entire spectrum of group technology applications, we focus on an application to the retrieval and reuse of engineering part designs.

Controlling a robot's position using neural networks

Abstract: In order to fully utilise the power of robots in factories, robot process capability (RPC) must be considered and improved. To improve the RPC in on-line processing by applying robot learning, the counterpropagation network was modified in this research. With two layers, the counterpropagation network was modified to control a robot's gross and fine motions. For the first layer, the network serves as a sensor-signal generator to control the gross motion. For the second layer, the network serves as a fine motion adjuster. Also, each layer can be separated functionally. By controlling both the gross and the fine motions, the RPC can then be improved. The modified two-layer counterpropagation network control scheme was validated by computer simulation and physical implementation on a RS-2200 robot system.

A formulation for collision identification and distance calculation in motion planning using neural networks

Abstract: The collision identification and object-to-object distance calculation play an important role in the motion planning for robots and manufacturing facilities. A formulation for collision identification and distance calculation in motion planning, using neural networks, is presented. The method calculates the distances between the vertices of an object and the given polyhedral obstacles using the modified Hamming net. This formulation is derived from the homogeneous geometric transformations. The method can be used to identify collision between the vertices of a moving object and the obstacles, to calculate the distance and interference between the moving object and the obstracle, and to find the optimal direction for collision removal. The parallel computation formulation is simple in form, and can be extended to line-to-object and object-to-object collision identification and distance calculation. The method can considerably decrease required computation time, and has the potential for being applied to on-line trajectory planning.

Guidelines for building 2D CAD models of turned components in CAD-CAPP integration

Abstract: EXCAP is a knowledge-based process-planning system for rotational components CADEXCAP is software which has been developed to integrate EXCAP with CAD systems via IGES (Initial Graphics Exchange Specification) CADEXCAP deciphers a full 2D model of a turned component developed on a CAD system and then via IGES it fully automatically generates a product model in a format suitable for EXCAP This paper describes the guidelines which have to be followed at the CAD modelling stage of the components which are to be used by CADEXCAP Following these guidelines will guarantee the automatic extraction of complete product models This not only includes the geometry but also includes nongeometric information such as toleranced dimensions, geometrical tolerances, manufacturing instructions and surface finish

CADEXCAP: Integration of 2D CAD models of turned components with CAPP

Abstract: EXCAP is a knowledge-based process-planning system for turned components. This paper describes a software package called CADEXCAP, which has been developed to integrate EXCAP and CAD systems via IGES. CADEXCAP deciphers a full 2 D model of a turned component developed on a CAD system and then, via IGES, fully automatically generates a product model in the EXCAP model format. It is clever enough to extract not only the geometrical information but also the other required information, such as geometrical and dimensional tolerances, surface finish and manufacturing instructions.

RDCAPP: A real-time dynamic CAPP system for an FMS

Abstract: Computer-aided process planning (CAPP) systems need to consider the uncertain nature of the shop-floor prior to the identification of job routes and processing sequences. This paper describes the design and development of a prototype, real-time dynamic CAPP (RDCAPP) system for a multimachining-centre flexible manufacturing system (FMS). Concepts relevant to this research include CAPP, profile input using group technology (GT), artificial-intelligence-based expert systems, and FMSs.