Hu Niaoqing

A Bearing Fault Detection Method Base on Compressed Sensing

For bearing fault detection in frequency domain, traditional methods estimate bearing fault condition based on mass data sampled by Nyquist sampling theorem, which will burden the storage. A new bearing fault detection method based on compressed sensing will be proposed in this paper in allusion to the problem mentioned above. The method presented here carried out compressive sampling and get a small set of incoherent projections, often the number of projections can be much smaller than the number of Nyquist rate samples. Then based on matching pursuit, the bearing condition will be estimated finally using these few measurements directly without ever reconstructing the signals involved. Sparsity of original signal is not demanded since the signal does not need to be recovered completely, which will also helped to expanded the method to other signals with similar characteristics in frequency domain. Related test will be achieved to verify the effectiveness of the method proposed in this paper.

Feature Selection Approach Based on Physical Model of Transmission System in Rotary Aircraft for Fault Prognosis

A majority of the mishaps of rotary aircraft are caused by the faults in drive train which is composed of some complex rotary mechanical systems. Planetary gear sets are common mechanical components and are widely used to transmit power and change speed and/or direction in rotary aircrafts. Planetary gear sets are epicyclical gear drive that is more complex compared to ordinary gear train, so the features of planetary gear sets is quite different from traditional features and hard to extract. This research focuses on the physical-model-based approach to extract features for planetary gear set. Physical model will be established for planetary gear set with fault. Then, the features suitable for severity estimation is selected based on the simulation signals of physical models. After that, the tests with faults seeded are carried out, and the validation has a promising result.

Damage level recognition for planetary gearbox in rotorcraft based on GRA and ANN

Planetary gearbox is a common mechanical component and is widely used to transmit power and change speed and/or direction in rotary aircrafts. The part failure of planetary gearbox is one of the main causes for the helicopter accidents. The need to identify the level of developing damage in part is central to reduce mechanically induced failures. An approach based on grey relational analysis(GRA) and artificial neural net (ANN) is presented to recognize the damage level quantitatively for planetary gearbox of rotorcraft. A particular emphasis is put on the feature selection based on GRA and the damage level recognition based on BP ANN. After that, the experiments with different-level-damage seeded are designed to validate the method above, and then the proposed method is used to identify the damage level based on test data. With the results of several experiments for damage level recognition, the feasibility and the effect of this approach are verified.

An Odor Discrimination Approach Based on Mice Olfactory Neural Network

Characteristic signals of the novelty volatile odor shed by equipments at abnormal state are often with higher dimension, and difficult to discriminate because of the complex background odorant noise in non-open space. An artificial olfactory neural network and its learning algorithm are introduced based on the anatomy of odor discrimination mechanism and olfactory neural model of mice. After the construction and training of an olfactory neural network for the discrimination of kerosene, gear oil and alcohol, it is verified through experiment data sets. The results indicate that the artificial neural network (ANN) based on mice olfactory model achieves a short time for training and the identification rate is feasible and effective.