Chair: Prof. Dongmin Kim (Korea Research Institute of Standards and Science)
/ Dr. Jay KIM (University of Cincinnati)
Physics-based prognostics-promises and challenges Prof. Nam-Ho KIM, Yiwei WANG and Raphael T HAFTKA (University of Florida, Université de Toulouse)
In this paper, an interesting observation on the noise-dependent performance of prognostics algorithms is presented, as well as a method of evaluating the accuracy of prognostics algorithms without having the true degradation model is presented. This paper compares the four most widely used model-based prognostics algorithms, i.e., Bayesian method, particle filter, Extended Kalman filter, and nonlinear least squares, to illustrate the effect of random noise in data on the performance of prediction. The mean squared error (MSE) that measures the difference between the true damage size and the predicted one is used to rank the four algorithms for each dataset. We found that the randomness in the noise leads to very different ranking of the algorithms for different datasets, even though they are all from the same damage model. In particular, even for the algorithm that has the best performance on average, poor results can be obtained for some datasets. In absence of true damage information, we propose another metric, mean squared discrepancy (MSD), which measures the difference between the prediction and the data. A correlation study between MSE and MSD indicates that MSD can be used to estimate the ranking of the four prognostics algorithms without having the true damage information. This may be particularly useful when information is available from multiple sites of damage for the same application.
A Modelling Approach to Monitor Friction within Electromechanical Actuator Ballscrews using Motor Current Mr. Yameen Monsur HUSSAIN, Stephen BURROW, Leigh HENSON and Patrick KEOGH (University of Bristol, Stirling Dynamics, University of Bath)
In this paper a modelling approach to monitor ballscrew friction within Electromechanical Actuators (EMA) using motor current is presented along with subsequent fault diagnostics using data classification of simulated data for healthy, degrading and fault states.
There is a move towards ‘more electric’ aircraft within the aerospace industry which has in turn prompted aircraft manufacturers to consider replacing traditional hydro-mechanical solutions for EMAs in actuation systems to fulfil the need for better maintainability and precision control. These drivers are also true for safety critical applications such as primary flight control systems and landing gear systems. The absence of reliable fail-safe mechanisms and adequate redundancy to mitigate the single point of failure (ballscrew jamming) has made it challenging to introduce EMAs in such safety critical systems. It is widely understood that Prognostics & Health Monitoring (PHM) of such systems could be one way to mitigate the risk of ballscrew jamming, however, research in this area has revealed that the issue is complex for PHM designers due to the limitations in sensing. Aircraft manufacturers are reluctant to add more sensors due to weight, reliability and cost implications and so PHM designers have to rely on motor current alone to detect the onset of ballscrew jamming. The purpose of this paper is to demonstrate a means to identify friction build up within the ballscrew using motor current for feature classification and fault diagnostics.
The approach used was to model a baseline linear EMA system to a high level of detail. The modelling placed emphasis on the Permanent Magnet Synchronous Motor (PMSM) where a greater understanding of the drivetrain could be achieved by modelling the PMSM in detail. The PMSM was modelled using ‘dq axis’ transformation theory, reducing 3-phase Alternating Current (AC) quantities (IA, IB, IC) to Direct Current (DC) quantities (ID, IQ). This simplifies the analysis of these reduced quantities before performing the inverse transform to recover from the actual 3-phase results. Modelling of the motor in this way will aid in-depth learning for condition monitoring and fault detection of the whole EMA drivetrain. The mechanical parts of the EMA were also modelled to a high fidelity. This included detail of the ballscrew by acknowledging the most contentious areas of friction within the ballscrew that would lead to a jamming case. Interaction between the screw thread and nut thread was considered the main source of friction within the ballscrew and so sliding velocities between the ball and nut, and ball and screw were calculated with subsequent values of velocity dependent friction generated using the ‘Stribeck’ friction model. Contact angles (between ball and nut, and ball and screw) and other mechanical efficiencies were varied to analyse the effect on the torque producing current for healthy, degrading and jamming conditions. The simulated data were then trained for each condition for data classification using a k-Nearest Neighbour (k-NN) algorithm.
The first part of the analysis revealed that ballscrew degradation should be detectable using motor current by monitoring changes to the torque producing q-axis current for each failure state in the ballscrew damage model. Instantaneous loads through a single cycle were also modelled to replicate external disturbances which in effect could cause fluctuations to the q-axis currents thus making it troublesome to isolate deteriorations to the ballscrew. The simulated datasets were processed for classification as training data using the k-NN algorithm where a classification accuracy of ~74% was achieved. This was deemed to be of reasonably high accuracy given the induced variability in the test conditions during simulations. Overall, the in-depth modelling of the EMA system presented a comprehensive approach to monitoring ballscrew friction through use of motor current analysis with reasonable accuracy from classifications from different test cases. It is envisaged that employing a hybrid approach (combination of model based and data driven techniques) to fault diagnostics of this problem by utilising real-life test data can further improve the classification accuracy.
Bayesian inferences of damage index and damage growth model based on on-site measurement data of steam turbine Mr. Woosung CHOI, Hyunseok OH, Byeng D. YOUN and Nam-Ho KIM (Seoul National University, GIST, University of Florida)
Accurate prediction of remaining useful life (RUL) of plant turbine is a major scientific challenge for effective operation and maintenance in the power plant industry. This paper proposes an RUL prediction methodology that such that measurement is relatively easy and unavoidable uncertainties are considered or reduced. Especially, discrepancy reduction and damage growth model considering uncertainties, combined with aleatory and epistemic uncertainties by irregular and discontinuous measurement and non-homogeneous samples, is developed to accurately predict RULs of aged components in power plants. Bayesian inference and MCMC technique is used to estimate the probability distribution of a damage index from on-site hardness measurements with uncertainties. A Bayesian approach to the damage growth model is proposed for aged steam turbines and the predictive distribution of the damage index is estimated using its mean and standard deviation.