Fault Detection and Tolerance on Robot Manipulators Locked Joint Failure Using Anfis

Transcription

1 Fault Detection and Tolerance on Robot Manipulators Locked Joint Failure Using Anfis TRaja prathab AP/ECE, Satyam College of Engineering & Technology, Nagercoil, India Prathabngl@gmailcom Dr RSuja Mani Malar Prof/Head, EEE Department, CAPE Institute of Technology,Levengipuram, Tamilnadu, India rsujamonimalar@gmailcom, Abstract Robot manipulator play important role in the field of automobile industry, mainly it is used in gas welding application and manufacturing and assembling of motor parts In complex trajectory, on each joint the speed of the robot manipulator is affected For that reason, it is necessary to analyze the noise and vibration of robot's joints for predicting faults Also improve the control precision of robotic manipulator In this paper, we will propose a new fault detection system for Robot manipulator The proposed hybrid fault detection system is designed based on fuzzy support vector machine and artificial neural networks (ANNs) In this system the decouple joints are identified and corrected using fuzzy SVM, here nonlinear signal are used for complete process and treatment, the artificial neural networks (ANNs) are used to detect the freeswinging and locked joint of the robot, two types of neural predictors are also employed in the proposed adaptive neural network structure The simulation results of a hybrid controller demonstrate the feasibility and performance of the methodology Keywords: Fault tolerance, Robotics, Co-operation, Fuzzy SVM, ANN Introduction In the recent years, various Neural networks based controller is developed for robotic manipulators Fault detection, diagnosis and accommodation play a key role in the operation of autonomous and intelligent robotic systems System faults, which typically result in changes in critical system parameters or even system dynamics, may lead to degradation in performance and unsafe operating conditionsrobotic manipulators have been deployed in an ever growing number of unstructured and/or hazardous environments, such as in outer space and in deep sea Robots are used in these environments to limit or eliminate the presence of human beings in such dangerous places, or due to their capability to execute repetitive tasks very reliably Faults, however, can put at risk the robots, their task, the working environment, and any humans present there Faults in robots are mainly due to their inherent complexity There are several sources of faults in robots, such as electrical, mechanical, and hydraulic [1] In fact, the meantime-to-failure of industrial robots can be considered small for their intended life expectancy and cost According to studies published in the decade of 1990, the recorded meantime-to-failure of industrial robots was only h [2] This number is probably smaller in unstructured or hazardous environments due to external factors, such as extreme temperatures, moving obstacles, and radiation Therefore, there are good reasons to research and develop fault detection and isolation (FDI) systems for robots Robotic systems with actuation redundancy are inter- esting in applications where fault tolerance is needed because the number of degrees of freedom (dof) in these systems is generally higher than the dof required to execute the task Furthermore, as in the human case where the use of two arms means an advantage over the use of only one arm in several cases, two or more robots can execute tasks that are difficult or even impossible for only one robot [3] Fault tolerant systems for single manipulators generally employ the residual generation scheme The residual vector is generated by comparing the measured states of the arm with their estimates obtained by a mathematical model of the fault-free arm This method, however, does not work well in the presence of modeling errors, generating false alarms or hiding the fault effects Robust techniques [4] and Artificial Intelligence techniques [5] have been used to avoid these problems In the approach presented in Vemuri and Polycarpou [6], the off-nominal behavior due to faults is mapped utilizing an ANN trained using a robust observer, based on the robot s physical model Overall, one problem with FDI methods which rely on the mathematical system model is that, for some real robots, detailed modeling is difficult The major advantage of the neural network approaches over the traditional control methods is that they requires less prior knowledge about the controlled systemrather,they involve using neurol networks to complete modeling and control task autonomously Kiguchi [7] proposed a fuzzy vector method that enables the controller to deal with force sensor signals including noise and/or unknown vibrations caused by working tool This type of controller exploits the possibilities of neural network for learning nonlinear functions as well as for solving certain types of problems where massive parallel computation is requiredthe learning capability of NN s is used to make the controller learn a certain function,highly nonlinear most of the time,representing direct dynamics,inverse dynamics or 887

2 any other charecterics of the processthis is usally done during a,normally long,training period when commissioning the controller is supervised or unsupervised manner[8]if the learning capability of the NN is not switched off after the training period,once the controller is commissioned The NN based hyprid controller is worked as an adaptive controllerthe ability of NNs for parallel computation has been exploited to implement controllers which require a substantial amount of computation[9] This article is organized as follows: in Section 2 we briefly explain the concept of and proposed hybrid controller In Section 3 we introduce the -based FDI methodology for robotic manipulators Sections 4 and 5 present, respectively, simulation and experimental results to validate the proposed framework We conclude the article in conclusion Section 6 Proposed Hyprid Adaptive Controller This paper investigates the problem of fault diagnosis in decouple joints of robotic manipulators A learning architecture, with neural networks as on-line approximations of the off-nominal system behavior, is used for monitoring the robotic system for faults The approximation (by the neural network) of the off-nominal behavior provides a model of the fault characteristics which can be used for detection and isolation of faults is a hybrid system of Neural networks and fuzzy logic concept The neural network is good at learning and highly interpretable and can function better than Regression methods [10] and fuzzy logic is good in handling imprecision The advantage of both the neural networks and fuzzy logic is combined in the Neuro-fuzzy model The advantage of Neurofuzzy approach is that the training time is reduced, due to its smaller dimensions and the capability of the network to be initialized with parameters relating to the problem domain[11] A specific approach in Neuro-fuzzy development is the adaptive Neuro-fuzzy inference system (), which has shown significant results in modelling nonlinear functions[12] uses a feed forward network to search for fuzzy decision rules that perform well on a given task Using a given input-output data set, creates a FIS whose membership function parameters are adjusted using a back propagation algorithm alone or a combination of a back propagation algorithm with the least square method This allows the fuzzy systems to learn from the data being modelled (Adaptive Neuro-Fuzzy Inference system) used in this paper is on sugeno type Fuzzy inference system It applies the combination of Least square method and the back propagation gradient descent method for training FIS membership function parameters to emulate the given training dataset [13] A typical architecture is shown in Fig1 The square represents the nodes that are adaptive and the circle represents the nodes that are fixed The fuzzy inference system consist of a first order Takagi- Sugeno fuzzy model with two inputs and,and a single output The inference system is governed by the following rule base Rule1: If is and y is, then Rule2: If is and y is, then Figure 1: The proposed Architecture of Layer 1: The functioning of is a five-layered feedforward neural structure, and the functionality of the nodes in these layers can be summarized as: for i=1, 2 (1) for i=3,4 (2) Where x or y is the input to the node, Ai or Bi-2 is a linguistic label, is the membership grade of a fuzzy set A associated with this node and it specifies the degree to which the given input satisfies the quantifier A Membership function for A can be represented in terms of generalized bell function as Where is the parameter set The Bell shaped function varies as the parameter set changes Layer 2: The product operation is used to estimate the firing strength of each rule, i=1,2 (4) Layer 3: The third layer is used to estimate the ratio of i th rule s firing strength to the sum of all rule's firing strengths Layer 4: (3) i=1,2 (5) Where w i is the output of layer 3 and {p i, q i, r i} is the parameter set Parameters in this layer will be referred to as consequent parameters Layer 5: The final layer computes the overall output as the summation of all incoming signals from layer 4 Overall output = (7) (6) 888

3 i Subtractive clustering: Subtractive clustering is proposed by Chiu et al [14] in which data points are considered as the candidates for cluster centers Subtractive clustering operates by finding the optimal data point to define a cluster centre based on the density of surrounding data points By this method the computation is simply proportional to the number of data points and independent of the dimension of the problem under consideration ii Fuzzy C Means clustering: Fuzzy C means clustering (FCM), also known as fuzzy ISODATA, is a data clustering algorithm in which each data point belongs to a cluster to a degree specified by a membership grade [5] Bezdek proposed this algorithm in 1973 as an improvement over earlier hard C means (HCM) clustering The Cluster center can also be initialized first and then the iterative procedure is carried out No guarantee is ensured that the FCM converges to an optimum solution The performance depends on initial cluster centers, there by using another fast algorithm to determine the initial cluster center or to run FCM several times starting with a different set of initial clusters Fault Detection and Isolation Scheme In this section, we present the fault detection and isolation technique based on artificial neural networks We initiate with a performance of residual generation for a generic dynamic system, and then concentrate it for the case of a robotic manipulator The dynamic of a fault free robotic manipulator with actuators in each joint is given by 1 ( t M (, V (,, G(, W (,,, H( T) Where q is the vector of joint angular positions, J is the vector of joint torques, M is the inertia matrix, T is the time index, V is the vector of Coriolis and centripetal terms, G is the vector of gravitational terms, W is the vector of friction terms and other nonlinearities and H is the vector of external uncorrelated disturbances The MLP mapping should reproduce the above eqn(1) in order to generate the residual vector However, the joint accelerations usually are not measured in real robotic manipulators Considering a small sample rate is submitted ) ( t ( t ( t in above eqn We get 1 ( t M (, V (,, G(, W (,,, H( T) t ( This is the function that should be replicated by the MLP mapping The residual generation system is expatriate in Figure 2 Figure 2: Residual generation process Where r is the residual generation vector The next step is to categorize the residual vector and the joint velocities by the The is employed now because, typically in FDI problems, it produces a more interesting classification than the MLP Other advantages over the MP,there are not a local minimum in the calculation of the weights and the training time is smaller However, the high number of radial units calls for more memory and the choice of the radial units and their parameters are suboptimal The outputs are trained to present signal 1 in the occurrence of the fault and 0 otherwise It is given in Figure 3 Figure 3: Residual analysis of proposed system Simulation Results The simulation study of the proposed system is carried out using the model of PUMA-560 robot, with the gear ratio of 30, the controller was designed for two of the joints,shoulder and elbow Assuming uncertainty of the payload and graspring center-point,parametrization of the robot model is accomplished with a vector of four components with m: payload mass and L ; Position of center of mass of load in x coordinate of the link fixed systemnow, asset of N=4 NNs is used to produce the feedforward control signaleach NN has 6 inputs,two hidden layers of 50 and 25 neurons,and one output layerthe learning method for training each NN consists of two phases In a first learning phase, the general learning structure, second phase is feed-forward error learning structure is used The input data set is presented normalized to the MLP The number of training epochs is high because the residual desired should be low for the fault-free operation After trained, the MLP reproduces the robot input/output behavior of trajectories utilized and for trajectories do not utilize the training procedure Meanwhile, when the joint 889

4 velocities are large, the residual is larger This can be minimized increasing the number of training epochs However, this can be absorbed using the joint velocity measurements in the input Result Analysis In the training, 9 trajectories with 40 samples each is selected During the training procedure, the trajectories are presented to the three times: one of the fault-free case, one when the fault 1 occurs and one when the fault 2 occursthe three algorithms described in above section are employed The Table 1 shows the sum-of-squared errors in the training patterns considering two outputs (faults 1 and 2) For validation test 17 untrained trajectories with 100 samples are presented three times (faults 1 and 2 and faultfree case) for the RElFN The Table 2 shows the sum-ofsquared errors in the test patterns Applying the fault criteria for RElFN trained with the three algorithms, all faults in the test set are detected and isolated For the GRR and LRR not any false alarm occurred in the test set For the FS, only one false alarm occurredthe experimental results are visualized in Figure 4 Table 1: Sum-of-squared errors in the training patterns Methods output 1 FS GRR output 2 Table 2: Sum-of-squared errors in the test patterns Methods output 1 output 2 FS GRR LRR CONCLUSION In this article a neural network founded hybrid adaptive controller for robot manipulators are proposed The controller uses a and back propagation based artificial neural network Each network has two stages learning stages and detection stages The proposed scheme produces a feed forward signal which adapts to changes in the payload and other operating parameters According to the simulation results the proposed system easier to manipulate and have a better performance after failure on one of its joints, also notable characteristics for the application of fault tolerance, which protest the reach-ability, operates easily without mechanical braking its functionality and performs better without new architectural configuration after recovery To show the feasibility and performance of the proposed control scheme, it has been tested on the model of a PUMA-560 robot, with a gear ratio of 30 LRR Figure 4: The encoded output error of different values 890

5 Reference International Journal of Applied Engineering Research ISSN Volume 11, Number 2 (2016) pp [1] Visinsky, M L, Cavallaro, J R, & Walker, I D (1995) A dynamic fault tolerance framework for remote robots IEEE Transactions on Robotics and Automation, 11(4), [2] Dhillon, B S, & Fashandi, A R M (1997) Robotic systems, probabilistic analysis Microelectronics and Reliability, 37(2), [3] Vukobratovic, M, & Tuneski, A (1998) Mathematical model of multiple manipulators: Cooperative compliant manipulation on dynamical environments Mechanism and Machine Theory, 33, [4] McIntyre, M L, Dixon, W E, Dawson, D M, & Walker, I D (2005)Fault identification for robot manipulators IEEE Transactions on Robotics, 21(5), [5] Schneider, H, & Frank, P M (1996) Observer-based supervision and fault-detection in robots using nonlinear and fuzzy logic residual evaluation IEEE Transactions on Control Systems Technology, 4(3), [6] Vemuri, A T, & Polycarpou, M M (2004) A methodology for fault diagnosis in robotic systems using neural networks Robotica,22(4), [7] Jayachandran A, Dhanasekaran R, "Automatic detection of brain tumor in magnetic resonance images using multi-texton histogram and support vector machine", International Journal of Imaging Systems and Technology, Vol 23, no 2, pp , (2013) [8] Psaltis D,SiderisA and Yamura AA A Multi layered neural network controller, IEEE control systems Mag, vol 8,no2,pp 17-21,1998 [9] Quero JM and Camacho EF, Neural generalized predictive self-tuning controller, Proc,IEEE infconf SystEng, Pittsburgh,PA,Aug 2011 [10] Muktha Palival, UshaAKumar, Neural networks and statistical techniques A review of applications, Expert system with application, vol36, no1, pp2-17, 2009 [11] Jyh-Shinng Roger Jang, : Adaptive-Network- Based fuzzy Inference system, IEEE ransaction on Ststem, Man and cybernetics, Vol23, no3, pp , May/June 1993 [12] Jayachandran, A & Dhanasekaran, R, Severity Analysis of Brain Tumor in MRI Images using Modified Multi-Texton Structure Descriptor and Kernel- SVM, The Arabian Journal of science and engineering October 2014, Volume 39, Issue 10, pp ,(2014) [13] JangJSR, Chuen-Tsai Sun, Eiji Mizutani, Neuro Fuzzy and Soft Computing, Prentice-Hall, 2006 [14] ChiuS, Method and Software for Extracting Fuzzy Classification Rules by Subtractive Clustering, Fuzzy Information Proceding Society, Biennial Conference of the North American, pp , 1996 [15] Sathyananda Reddy, KVSVN Raju, Improving the accuracy of effort estimation through fuzzy set combination of size and cost drivers, WSEAS Transaction on Computers, Vol8, no6, PP , June