Showing papers by "Hee-Jun Kang published in 2019"
TL;DR: The three popular Deep Learning algorithms for Bearing fault diagnosis including Autoencoder, Restricted Boltzmann Machine, and Convolutional Neural Network are briefly introduced and their applications are reviewed through publications and research works on the area of bearing fault diagnosis.
379 citations
TL;DR: This paper proposes a method for diagnosing bearing faults based on a deep structure of convolutional neural network which does not require any feature extraction techniques and achieves very high accuracy and robustness under noisy environments.
281 citations
TL;DR: The proposed adaptive terminal sliding mode control algorithm for robot manipulators can enable the advantages of non-singularity, high robustness, small transient error, and finite time convergence, as well as employing the adaptive self-tuning rules with no prior information regarding the upper bounds of undefined parameters.
Abstract: This paper presents an adaptive terminal sliding mode control (TSMC) algorithm for robot manipulators. The contribution of our control method is that the suggested controller can enable the advantages of non-singular TSMC such as non-singularity, high robustness, small transient error, and finite time convergence. To develop the suggested system, a non-singular terminal sliding variable is selected and does not have any complex-value or constraints of the exponent in conventional TSMC. Therefore, it prevents the singularity that occurs in the conventional TSMC and eliminates the reaching phase glitch. Accordingly, the suggested system can ensure that the controlled variables reach the desired values within a randomly known finite time using an efficiently smooth and chattering-free definite control input. In addition, sliding motion in finite time can be achieved by employing the adaptive self-tuning rules with no prior information regarding the upper bounds of undefined parameters (e.g., friction, disturbances, and uncertainties). Furthermore, the finite-time convergence and global stability of the proposed algorithm are proved by the Lyapunov stability theory. Finally, the proposed control algorithm is applied to the joint position tracking control simulation for a 3-DOF PUMA560 robot. The trajectory tracking performance of the proposed method is compared with those of the conventional terminal sliding mode control and the conventional continuous sliding mode control. This comparison shows the efficiency and superiority of the proposed algorithm.
40 citations
TL;DR: The proposed method models and identifies determinable error sources, for instance, geometric errors and joint deflection errors, and uses an artificial neural network for compensating for the robot position errors, which are caused by these non-geometric error sources.
Abstract: Robot position accuracy plays a very important role in advanced industrial applications. This article proposes a new method for enhancing robot position accuracy. In order to increase robot accurac...
35 citations
TL;DR: To track the specified trajectories with high accuracy, a control approach is developed for the class of general nonlinear second-order systems by utilizing an IFOTSM surface and an adaptive compensator and the unknown dynamic model is derived based on a radial basis function neural network.
Abstract: This paper reports the design of a control system for a class of general nonlinear second-order systems. The significant problems of singularity and chattering phenomenon, which limit the use of the conventional terminal sliding mode control (TSMC) in real applications due to the order of the sliding surface, need to be addressed. In addition, the effects of disturbances and uncertainties need to be removed, and the response rates are increased. Therefore, the integral full-order terminal sliding mode (IFOTSM) surface was proposed. To track the specified trajectories with high accuracy, a control approach is developed for the class of general nonlinear second-order systems by utilizing an IFOTSM surface and an adaptive compensator. The unknown dynamic model is derived based on a radial basis function neural network (RBFNN). Consequently, our controller provides good performance with minimum position errors, robustness against uncertainties, and work without a precise dynamic model. The simulated examples were performed to analyze the effectiveness of the control approach for position pathway tracking control of a 2-DOF parallel manipulator.
23 citations
TL;DR: A chattering-free, adaptive, and robust tracking control scheme for a class of second-order nonlinear systems with uncertain dynamics and an integral of a switching term and an adaptive updating law to compensate the lumped system uncertainty is introduced.
Abstract: This paper introduces a chattering-free, adaptive, and robust tracking control scheme for a class of second-order nonlinear systems with uncertain dynamics. First, a proportional-integral-derivative control-fast terminal sliding function is proposed to enable the advantages of both the PID and non-singular fast terminal sliding mode approaches in the field of non-singularity, fast convergence time, defined time convergence, and stability with small steady-state errors. Second, to obtain the desired control target without chattering behavior, the proposed controller with a continuous approach has been applied. In detail, the proposed controller uses an integral of a switching term and an adaptive updating law to compensate the lumped system uncertainty (e.g., disturbances, unmodeled dynamics, nonlinearities, or unmeasurable noise). Our proposed controller does not require knowledge about bound values of those anonymous components. The robust behavior and the defined time convergence have been demonstrated rigorously by the Lyapunov principle. Finally, the position tracking computer simulations have been performed to demonstrate the effectiveness and practicality of the suggested controller.
18 citations
14 Dec 2019
TL;DR: A pre-trained deep neural network in the domain of image classification is adopted and modified to extract features from vibration signals measured by multiple sensors to solve the problem of bearing fault diagnosis.
Abstract: Rolling element bearings are among the most important components in rotary machines. The reliable operation of rotary machines highly depends on the performance of bearing. Therefore, bearing fault diagnosis is a critical task in the industry. Signal-based fault diagnosis for bearings has applied extensively deep learning algorithms because of their ability to automatically extract features from fault signals measured from rotary machines. However, designing a deep learning model for any fault diagnosis problem is not a trivial task since each deep model has a complex structure and a huge number of hyper-parameters and trainable parameters. Each hyper-parameter of a deep learning model has a profound impact on the performance of that model. The selection of appropriate hyper-parameters is often conducted manually based on the Trial & Error method and experiences of the designer. Transfer learning is a technique that adopts already existing machine learning models into new domains. This technique helps to save the designing and training time of machine learning models, especially deep neural networks. In this paper, transfer learning technique is exploited to the problem of bearing fault diagnosis. A pre-trained deep neural network in the domain of image classification is adopted and modified to extract features from vibration signals measured by multiple sensors. The effectiveness of the proposed method is verified by experiments conducted with actual bearing data set supplied by Case Western Reverse University Bearing Data Center.
9 citations
03 Aug 2019
TL;DR: A continuous PID sliding mode control strategy based on a neural third-order sliding mode observer for robotic manipulators by using only position measurement that provides finite-time convergence, high accuracy, chattering reduction, and robustness against the dynamic uncertainties and faults without the need of velocity measurement.
Abstract: This paper proposes a continuous PID sliding mode control strategy based on a neural third-order sliding mode observer for robotic manipulators by using only position measurement. A neural third-order sliding mode observer based on radial basis function neural network is first proposed to estimate both the velocities and the dynamic uncertainties and faults. In this observer, the radial basis function neural networks are used to estimate the parameters of the observer, therefore, the requirement of prior knowledge of the dynamic uncertainties and faults is eliminated. The obtained velocities and lumped uncertainties and fault information are then employed to design the continuous PID sliding mode controller based on the super-twisting algorithm. Consequently, this controller provides finite-time convergence, high accuracy, chattering reduction, and robustness against the dynamic uncertainties and faults without the need of velocity measurement and the prior knowledge of the lumped dynamic uncertainties and faults. The global stability and finite-time convergence of the controller are guaranteed in theory by using Lyapunov function. The effectiveness of the proposed method is verified by computer simulation for a PUMA560 robot.
8 citations
25 Sep 2019
TL;DR: A combination between a neural network and an adaptive sliding mode control for trajectory tracking control of a 3-DOF planar parallel manipulator, which has a complicated dynamic model, including modelling uncertainties, frictional uncertainties and external disturbances.
Abstract: This paper proposes a combination between a neural network and an adaptive sliding mode control for trajectory tracking control of a 3-DOF planar parallel manipulator. It has a complicated dynamic model, including modelling uncertainties, frictional uncertainties and external disturbances. The proposed control algorithm is to use a PID sliding mode surface, an adaptive sliding mode controller with a neural network to overcome the drawback of the traditional sliding mode controllers, such as slow response rate with variation of uncertainties and external disturbances, chattering, and upper bound values of undefined dynamics which affects system performance, high wear of moving mechanical parts and high heat losses in power circuits. The radial basis function neural network is designed to compensate for uncertainties and external disturbances, which allows small switching gain. Hence, the chattering can be significantly reduced. In addition, an adaptive control law is used to adaptively converge small switching gains of the sliding mode controller as the neural network reduces model uncertainties. The effectiveness of the proposed control strategy is demonstrated by simulations which are conducted by using the combination of Sim-Mechanics and SolidWorks.
7 citations
03 Aug 2019
TL;DR: This paper proposes a method for diagnosing faults in a rotary machine based on vibration signal using convolutional neural network with batch normalization technique, which is an advantaged variant of traditional convolutionAL neural network.
Abstract: Deep learning or deep neural network is a type of machine learning which exploits deep structures of neural networks with many layers of nonlinear data processing units. Deep neural networks have the ability to automatically extracting abstract data presentation for raw data such as time series signals, texts, and images. Deep learning has been extensively applied in vibration signal-based fault diagnosis for rotary machine since it can learn features from the raw signal of a rotary machine without requiring hand-crafted feature extraction. Batch normalization is a technique proposed to accelerate the training process and convergence of deep neural networks. In this paper, we propose a method for diagnosing faults in a rotary machine based on vibration signal using convolutional neural network with batch normalization technique, which is an advantaged variant of traditional convolutional neural network. The effectiveness of the proposed method is then verified with experiments on bearings and gearboxes fault data.
3 citations
03 Aug 2019
TL;DR: The suggested controller provides strong properties of high tracking accuracy and quick response with minimum tracking errors, and the simulated performances verify high effectiveness of the proposed controller in trajectory tracking control of a 3-DOF robot manipulator.
Abstract: In this paper, a full-order sliding mode tracking control system is developed for industrial robots. First, to dismiss the effects of perturbations and uncertainties, while to improve faster response time and to eliminate the singularity, a full-order sliding function is selected. Next, to reach the prescribed tracking path and to remove the chattering, a control method is designed for robot manipulators by using a combination of full-order sliding function and a continuous adaptive control term. Additionally, the unknown dynamic model of the robot is estimated by adopting a radial basis function neural network. Due to the combination of these methodologies, the proposed controller can run free of exact robot dynamics. The suggested controller provides strong properties of high tracking accuracy and quick response with minimum tracking errors. In simulation analysis, the simulated performances verify high effectiveness of the proposed controller in trajectory tracking control of a 3-DOF robot manipulator.
03 Aug 2019
TL;DR: The results were illustrated that the proposed control can tolerate the relatively bigger faults due to the design of the observer and then show the better performances than the conventional super-twisting controller does.
Abstract: In this paper, real implementation of an active fault-tolerant control for a robot manipulator based on the combination of an external linear observer and the super-twisting algorithm is proposed. This active fault-tolerant scheme uses an external linear observer to identify faults. Then, the fault information is used to compensate the uncertainties/disturbance and faults with the super twisting controller. Finally, the effectiveness of proposed control is verified by simulation and implementation for a 3-DOF robot manipulator. The results were illustrated that the proposed control can tolerate the relatively bigger faults due to the design of the observer and then show the better performances than the conventional super-twisting controller does.
03 Aug 2019
TL;DR: A new hybrid calibration method for improving the absolute positioning accuracy of robot manipulators is proposed, where the geometric errors and joint deflection errors are simultaneously calibrated by robot model identification technique and a radial basis function neural network is applied for compensating the robot positions errors, which are caused by the non-geometric error sources.
Abstract: Though the kinematic parameters had been well identified, there are still existing some non-negligible non-geometric error sources such as friction, gear backlash, gear transmission, temperature variation etc. They need to be eliminated to further improve the accuracy of the robotic system. In this paper, a new hybrid calibration method for improving the absolute positioning accuracy of robot manipulators is proposed. The geometric errors and joint deflection errors are simultaneously calibrated by robot model identification technique and a radial basis function neural network is applied for compensating the robot positions errors, which are caused by the non-geometric error sources. A real implementation was performed with Hyundai HH800 robot and a laser tracker to demonstrate the effectiveness of the proposed method.
25 Sep 2019
TL;DR: This paper proposes an adaptive integral sliding mode tracking control for robotic manipulators with the elimination of the reaching stage to provide better trajectory tracking accuracy and to stabilize the closed-loop system.
Abstract: This paper proposes an adaptive integral sliding mode tracking control for robotic manipulators. Our proposed control method is developed based on the benefits of both integral sliding mode control and adaptive control, such as high robustness, high accuracy, and estimation ability. In this paper, an integral sliding mode controller is designed with the elimination of the reaching stage to provide better trajectory tracking accuracy and to stabilize the closed-loop system. To reduce the computation complexity, an adaptive controller with only one simple adaptive law is used to estimate the upper-bound values of the lumped model uncertainties. As a result, the requirement of their prior knowledge is eliminated and then decrease the computation cost. Consequently, this controller provides better tracking accuracy and handles the dynamic uncertainties and external disturbances more strongly. The system global stability of the controller is guaranteed by using Lyapunov criteria. Finally, the effectiveness of the proposed control method is tested by computer simulation for a PUMA560 robotic manipulator.