We experimentally obtain and analyse green density distribution in stainless steel compact samples and investigate the effect of compaction pressure on sample green density and density distribution. Experimental measurements of local density of stainless steel samples are conducted using scanning electron microscopy. For design purposes, the measured local densities, depth and planar location and compaction pressure are used to train an artificial neural network model to estimate the compaction density as a function of input parameters. Material parameters obtained experimentally are used to calibrate a finite element model. The results show that the artificial neural network and finite element modelling approaches are feasible and could be used in predicting the overall compaction density variations in powder metallurgy components. It is observed that the overall compact green density increases almost linearly with compaction pressure. Phenomena of particle interlocking and cold welding are observed and discussed.

Vibration control strategies strive to reduce the effect of harmful vibrations on machinery and people. In general, these strategies are classified as passive or active. Although passive vibration control techniques are generally less complex, there is a limit to their effectiveness. Active vibration control strategies, on the other hand, can be very effective but require more complex algorithms and are especially susceptible to time delays. The current paper introduces a novel vibration suppression system using non-linear optimization. The proposed methodology eliminates the need for a feedback loop and the sensitivity to time delays. The system has been evaluated experimentally and the results show the validity of the proposed methodology.

In this paper, a geometric method is presented for predictive modelling of surface finish in grinding. The chip removal process by successive grains is modelled as a three dimensional Boolean operation, from which the surface roughness is predicted. A solid modeller is used to model an individual chip as a partial ellipsoid. The measured topography of the grinding wheel, together with kinematic relationship in surface grinding, is used to determine the geometrical characteristics of the proposed ellipsoid. The surface roughness predicted by the model is compared with experimental results. The results show good consistency between the model and the actual surface properties.

An earlier energy-based fatigue damage model [1] has been examined to evaluate the durability of micro-sized silicon components subjected to cyclic loads. The components of this model are based on the physics of crack initiation and damage progress on the most damaging planes.Fatigue damage accumulation of silicon micro-components is accompanied with the formation of inclined micro-critical planes of {111} on the fracture surface as means of dissipating energy during the fatigue damage progress, which in principal coincides with the mechanism of fatigue damage accumulation in Varvani's approach [1]. Predicted fatigue lives based on the damage model were found in a good agreement with experimental fatigue life data of these components reported in the literature. Correlations are within a factor of +/-2.5 for short and long lives, which are within the limits of acceptance.

Laser welding is becoming increasingly important in the automotive industry and quality of the weld is critical for a successful application. In many cases the increase in welding speed provided by laser welding has resulted in the need for an automated, on-line weld monitoring system. This article describes a methodology for monitoring laser gear welding with the objective of detecting lack of fusion defects. The proposed system is evaluated experimentally. It is also shown that the proposed signal processing and modeling technique can be used to detect porosity in the welds.

This paper describes a system for on-line adjustment of the cutting conditions in a turning operation. Cutting conditions are set based on an adaptive model of the cutting operation taking into consideration the gradual wear of the tool. The objective is to control the rough turning operation for a predetermined tool life under varying cutting conditions. Two case studies show the feasibility of the proposed methodology.

A method for measuring the properties of laser weld metal in tailor welded automotive blanks is assessed. Tensile specimens in which the weld lies parallel to the axis of tension are pulled to failure and the weld metal properties are determined using a “rule of mixtures” type of calculation. Experiments are performed using two sets of similar gage welded steel specimens. The sensitivity of determined weld metal properties to specimen size and means of determination of weld metal cross-sectional area are assessed. The method is shown to be useful but in need of refinement, especially with respect to the measurement of weld metal cross-sectional area.

Laser welding is becoming more and more important in automotive industry and quality of the weld is critical for a successful application. In many cases the increase in welding speed provided by laser welding has caused the welding system operator to be unable to keep up with the production rate while inspecting each part. Therefore, either additional inspectors are required, or some form of real time on-line inspection of the weld must be provided. This is especially necessary where the laser weld properties are critical to final product performance. This paper describes architecture of such a system. The proposed system is based on a dynamic model comprised of static and dynamic neural networks.

This paper reports the latest progress in developing an "on-line" wear estimation system for turning operations. The system was designed to simultaneously estimate the important components of the wear encountered in turning. A hierarchical structure using multilayered feedforward static and dynamic neural networks is used as a specialized subsystem, for each wear component to be monitored. These subsystems share information about the tool wear components they are monitoring and their error in estimating the cutting force components is used to update the dynamic neural networks. The adaptability property of neural networks ensures that changes in machining parameters can be accommodated. Simulation studies are undertaken using experimental data available from manufacturing literature. Experimental verifications are also performed to ensure that the system could be implemented in real working conditions. The results are promising and show good estimation ability.

This paper describes a real-time tool condition monitoring system for turning operations. The system uses a combination of static and dynamic; neural networks with off-line and on-line training and cutting force components are used as diagnostic signals. The system is capable of monitoring several wear components simultaneously. The wear estimation system has been implemented experimentally to evaluate its suitability for use in shop floor conditions. The tests were performed in real time with different cutting conditions. The experimental results showed that the system was successful in predicting three wear components in real time. However, the accuracy of the wear prediction was not the same for all three wear components. The crater wear predictions were less accurate partly because of the opposing effects of crater and flank wear components on cutting force components. (C) 1999 Elsevier Science Ltd. All rights reserved.

This paper describes the development of a suitable algorithm to compute the potential of tipping-over for vehicles that carry manipulators. The energy method developed by Messuri and Klein (1985) is extended here to quantitatively reflect the effect of forces and moments arising from the manipulation of the implement. The amount of the impact energy that can be sustained by the vehicle without tipping-over, about each edge of potential overturning is computed. First, the instantaneous onset of instability configuration of the machine about the edge is determined by constructing an equilibrium plane. Next, the work done by all acting forces and moments when the machine is virtually brought to this unstable stance from the current state, is calculated. This work is the indication of the proximity of the machine to tipping-over around that edge. The application of this study is directed at industrial mobile machines that carry human-operated hydraulic manipulators. The algorithm is therefore used to study the stability of an excavator based log-loader. Simulation studies clearly show the importance of inertial loads in determining the stability of such machines.

Inherent to any heavy-duty hydraulic machine operation with a large number of interconnected components are nonidealities such as gear backlash, friction and leakage. The swing motion of the operator’s cabin in an excavator is a typical example. In this paper we conduct a study comprising experimental, mathematical and simulation components to determine the degree to which these nonlinearities affect the performance of such machines. The inclusion of the conventional model of backlash in the simulation of the excavator swing motion is shown to be inefficient and unnecessary in terms of computation time and the final results. A new model which combines the fluid-flow and the gear train dynamics is developed. The study of contact and non-contact cases brings about proper sets of static and dynamic equations which efficiently simulate this phenomenon for the class of excavator machines under consideration. The inclusion of stick-slip friction model in the simulation shows two effects. Firstly, it causes a noticeable time-delay at the beginning of the swing motion. Secondly, it results in an overshoot during velocity control experiments. It is also shown that dry friction and leakage (cross-port or external) are as significant as gear backlash in determining the pressure patterns in the connecting hydraulic lines and, therefore, should not be overlooked, especially when the excavator cabin is brought to a stop. Often, this is the most important state event when accurate positioning is crucial. The simulation results are qualitatively supported by the experimental evidence. The experiments were performed on an instrumented teleoperated Caterpillar 215B excavator.

In this paper, a neural network based system for ‘on-line’ estimation of tool wear in turning operations is introduced. The system monitors the cutting force components and extracts the tool wear information from the changes occurring over the cutting process. A hierarchical structure using multilayered feedforward static and dynamic neural networks is used as a specialized subsystem, for each wear component to be monitored. These subsystems share information about the tool wear components they are monitoring and their error in estimating the cutting force components is used to update the dynamic neural networks. The adaptability property of neural networks ensures that changes in machining parameters can be accommodated. Simulation studies are undertaken using experimental data available from manufacturing literature. The results are promising and show good estimation ability.

This paper presents a scheme to monitor the potential of tipping over for moving base manipulators. The method of energy stability developed by Messuri and Klein (1985) is extended here to quantitatively include the effect of all factors relevant to the stability of moving base manipulators. These factors include vehicle top-heaviness, rugged terrain conditions, inertial and external reactions arising from the manipulation of the implement. The application of this study is directed at teleoperated heavy-duty hydraulic machines that are used in forestry and construction industries. Simulation studies show that the inertial loadings are important in determining the stability of such machines and thus should not be overlooked