Eric H. K. Fung

A real-time gyroscopic system for three-dimensional measurement of lumbar spine motion

This paper describes a new method of measuring the three-dimensional movements of the lumbar spine in real time. The measurement system consisted of solid-state gyroscopes which were attached to the trunk. They measured the angular rates of rotations in three dimensions, which were then integrated to obtain the orientation. The sensors contained gravitometers and magnetometers which provided additional information for eliminating any drift of the gyroscopes. The reliability of the data provided by the gyroscopic system was examined in a group of 19 young healthy subjects. The similarity of the movement-time curves obtained in three repeated measurements was assessed by the coefficient of multiple correlation. The coefficients were found to be high, ranging from 0.972 to 0.991. The reliability of the data was slightly lower for measuring axial rotation. The device did not only quantify the kinematic patterns in the primary plane of movements, but also the accompanying movements in the other planes. Flexion and extension was found to be mainly confined to the sagittal plane, whereas lateral bending and axial rotation always accompanied each other. It was concluded that the inertial tracking device would be a useful tool for clinical measurement as well as biomechanical investigations.

Stitching defect detection and classification using wavelet transform and BP neural network

In the textile and clothing industry, much research has been conducted on fabric defect automatic detection and classification. However, little research has been done to evaluate specifically the stitching defects of a garment. In this study, a stitching detection and classification technique is presented, which combines the improved thresholding method based on the wavelet transform with the back propagation (BP) neural network. The smooth subimage at a certain resolution level using the pyramid wavelet transform was obtained. The study uses the direct thresholding method, which is based on wavelet transform smooth subimages from the use of a quadrant mean filtering method, to attenuate the texture background and preserve the anomalies. The images are then segmented by thresholding processing and noise filtering. Nine characteristic variables based on the spectral measure of the binary images were collected and input into a BP neural network to classify the sample images. The classification results demonstrate that the proposed method can identify five classes of stitching defects effectively. Comparisons of the proposed new direct thresholding method with the direct thresholding method based on the wavelet transform detailed subimages and the automatic band selection for wavelet reconstruction method were made and the experimental results show that the proposed method outperforms the other two approaches.

A hybrid model using genetic algorithm and neural network for classifying garment defects

The inspection of semi-finished and finished garments is very important for quality control in the clothing industry. Unfortunately, garment inspection still relies on manual operation while studies on garment automatic inspection are limited. In this paper, a novel hybrid model through integration of genetic algorithm (GA) and neural network is proposed to classify the type of garment defects. To process the garment sample images, a morphological filter, a method based on GA to find out an optimal structuring element, was presented. A segmented window technique is developed to segment images into several classes using monochrome single-loop ribwork of knitted garment. Four characteristic variables were collected and input into a back-propagation (BP) neural network to classify the sample images. According to the experimental results, the proposed method achieves very high accuracy rate of recognition and thus provides decision support in defect classification.

Modelling and prediction of machining errors using ARMAX and NARMAX structures

Forecasting compensatory control, which was first proposed by Wu [ASME J. Eng. Ind. 99 (1977) 708], has been successfully employed to improve the accuracy of workpieces in various machining operations. This low-cost approach is based on on-line stochastic modelling and error compensation. The degree of error improvement depends very much on the accuracy of the modelling technique, which can only be performed on-line in a real-time recursive manner. In this study, the effect of the control input (i.e. the cutting force) is considered in the development of the error models, and the formulation of recursive exogenous autoregressive moving average (ARMAX) models becomes necessary. The nonlinear ARMAX or NARMAX model is also used to represent this nonlinear process. ARMAX and NARMAX models of different autoregressive (AR), moving average (MA) and exogenous (X) orders are proposed and their identifications are based on the recursive extended least square (RELS) method and the neural network (NN) method, respectively. An analysis of the computational results has confirmed that the NARMAX model and the NN method are superior to the ARMAX model and the RELS method in forecasting future machining errors, as indicated by its higher combined coefficient of efficiency.

Dynamic modelling of a rigid-flexible manipulator for constrained motion task control

Abstract A mathematical model governing the dynamics of a constrained rigid-flexible manipulator moving in a horizontal plane is derived using Hamiltons principle. A new variable is introduced before the procedure of modal expansion in order to convert the non-homogeneous boundary condition into a homogeneous one. The static tip deflection of the flexible link is allowed in order to maintain the contact force between the end effector and the constrained path and this tip deflection is considered in both the inverse kinematics and the order reduction procedures. The state vector of the proposed controller consists of joint angle of the rigid link, its derivative and integral, the first deflection mode and its derivative, and the integral of contact force. A multivariable controller is proposed for the simultaneous motion and force control of the manipulator. The controller consists of a feedforward term which contributes the torque for the expected joint angles and the contact force, and a feedback term with the time varying optimal gains obtained from the Matrix Riccati equation. Computer simulation results show that this proposed controller is capable of performing the straight line tracking task satisfactorily under four different conditions.

An approach to on-machine motion error measurement of a linear slide

A new system is proposed to measure the straightness motion error of a precision linear slide for the purpose of active error compensation (AEC). In the traditional method, where the slide profile is assumed constant, its motion error can be obtained simply by subtracting the profile data from the output of a displacement sensor. However, under on-machine conditions, such pre-calibrated profiles or references are not reliable, due to deviations from the calibration conditions. To overcome this difficulty, information on multiple displacement sensors is employed to construct an ideal software datum that can be used on-line in AEC. Such a software datum can act as an ideal reference for separating the profile from the motion errors. The characteristics of the proposed method are: (1) the profile is determined under on-machine conditions requiring no pre-calibrations; (2) a novel approach to represent the test section of the profile by one cycle of a periodic function is facilitated by specially arranging three sensors to form a stationary stage. Computer simulations are performed to confirm the feasibility of the method. A measurement system consisting of three non-contact capacitive sensors is built and evaluated experimentally on a laboratory lathe and good repeatability of profile results is obtained.

Fabric Stitching Inspection Using Segmented Window Technique and BP Neural Network

In the textile and clothing industry, much research has been conducted on fabric defect automatic detection. However, few have been specifically designed for evaluating fabric stitches or seams of semi-finished and finished garments. In this paper, a fabric stitching inspection method is proposed for knitted fabric in which a segmented window technique is developed to segment images into three classes using a monochrome single-loop ribwork of knitted fabric: (1) seams without sewing defects; (2) seams with pleated defects; and (3) seams with puckering defects caused by stitching faults. Nine characteristic variables were obtained from the segmented images and input into a Back Propagation (BP) neural network for classification and object recognition. The classification results demonstrate that the inspection method developed is effective in identifying the three classes of knitted-fabric stitching. It is proved that the classifier with nine characteristic variables outperformed those with five and seven variables and the neural network technique using either BP or radial basis (RB) is effective for classifying the fabric stitching defects. By using the BP neural network, the recognition rate was 100%.

Using ARMA models to forecast workpiece roundness error in a turning operation

Abstract The paper describes the methodology for developing autoregressive moving average (ARMA) models to represent the workpiece roundness error in the machine taper turning process. The method employs a two stage approach in the determination of the AR and MA parameters of the ARMA model. It first calculates the parameters of the equivalent autoregressive model of the process, and then derives the AR and MA parameters of the ARMA model. Akaikes Information Criterion (AIC) is used to find the appropriate orders m and n of the AR and MA polynomials respectively. Recursive algorithms are developed for the on-line implementation on a laboratory turning machine. Evaluation of the effectiveness of using ARMA models in error forecasting is made using three time series obtained from the experimental machine. Analysis shows that ARMA(3,2) with forgetting factor of 0.95 gives acceptable results for this lathe turning machine.

The implementation of an error forecasting and compensation systems for roundness improvement in taper turning

Abstract In this study, forecasting and compensatory control (FCC) techniques are employed to improve the workpiece roundness accuracy in taper turning on an experimental lathe. The paper focuses on the development and implementation of both hardware and software associated with the forecasting and compensation system. The tasks involve the in-process error measurement, off-line error modelling and simulation and on-line error forecasting and compensation. Experimental results have shown that an improvement of 24–38% was achieved for the roundness error of workpieces in the taper turning operations.

An error separation technique for measuring straightness motion error of a linear slide

A novel measurement system for on-machine determination of the straightness motion error and profile of a precision linear slide for active error compensation (AEC) purposes has successfully been developed. The characteristics of the proposed measurement method are: (i) the profile is determined under on-machine conditions requiring no pre-calibration; and (ii) a novel approach involving representation of the test section of the profile by one cycle of a periodic function is facilitated by specially arranging three sensors to form a stationary stage. In this paper, the principles and operation of the measurement method are presented. The approach of defining a periodic function to deal with the problem of non-periodicity of a profile is then described. A preliminary analysis of the sensor measurement errors is also given. Finally, the measurement system consisting of three non-contact capacitive sensors is built and evaluated experimentally on a laboratory lathe and good reproducibility of profile results is obtained.