This paper addresses the design of an observer-based fault diagnosis scheme, which is applied to some of the sensors and actuators of a wind turbine benchmark model. The methodology is based on a modified sliding mode observer (SMO) that allows accurate reconstruction of multiple sensor or actuator faults occurring simultaneously. The faults are reconstructed using the equivalent output err...

We present a novel convolutional auto-encoder (CAE) for unsupervised feature learning. A stack of CAEs forms a convolutional neural network (CNN). Each CAE is trained using conventional on-line gradient descent without additional regularization terms. A max-pooling layer is essential to learn biologically plausible features consistent with those found by previous approaches. Initializing a CNN ...

Simplified broadband beamformers can be constructed by sharing a single tapped-delay-line within a narrowband subarray. This paper discusses the use of fractional delay filters to a steering in the digital domain. For the narrowband subarrays, an optimisation approach is proposed to maintain an off-broadside look direction constraint as best as possible across a given frequency range. We demons...

On-Line Transaction Processing (OLTP) workload is an important application in commercial service domain. Chip multiprocessor (CMP) is main stream of processor development. Performance characters of OLTP in CMP become a crucial benchmark for design and performance analysis of CMP server system. System level performance characters of OLTP on AMD quad-core system is presented in the paper. Accordi...

We solve the dynamics of on-line Hebbian learning in large perceptrons exactly, for the regime where the size of the training set scales linearly with the number of inputs. We consider both noiseless and noisy teachers. Our calculation cannot be extended to non-Hebbian rules, but the solution provides a convenient and welcome benchmark with which to test more general and advanced theories for s...

This study focuses on performance assessment and monitoring of model predictive control systems. A methodology is proposed to determine a benchmark and monitor MPC performance on-line. A performance measure based on the ratio of historical and achiev ed performance is used for monitoring and a ratio of design and achiev ed performance is used for diagnosis. Case studies with linear and nonlinea...

A new diagnostic fault simulator is described that diagnoses both feedback and nonfeedback bridge faults while using information from fault simulation of single stuck-at faults. A realistic fault model is used which considers the existence of the Byzantine Generals Problem. The approach has been demonstrated for two-line bridge faults on several large combinational benchmark circuits and has ac...

The research reported in this paper is concerned with assessing the usefulness of reinforcment learning (RL) for on-line calibration of parameters in evolutionary algorithms (EA). We are running an RL procedure and the EA simultaneously and the RL is changing the EA parameters on-the-fly. We evaluate this approach experimentally on a range of fitness landscapes with varying degrees of ruggednes...