F. Guillet

A New Bearing Fault Detection Method in Induction Machines Based on Instantaneous Power Factor

Fault detection and diagnosis of asynchronous machine has become a central problem in industry over the past decade. A solution to tackle this problem is to use stator current for a condition monitoring, referred to as motor current signature analysis. This paper argues that bearing faults would have a negligible effect on motor currents and instead argues that the more likely reason why the faults can be detected in currents is because they entail a fluctuating resistive torque which acts immediately, in contrast to the radial displacement which takes time to integrate to a perceptible displacement even in response to a step change in velocity. In this context, we propose a new method for detecting bearing defects based on the exploitation of the instantaneous power factor that varies according to torque oscillations. Experimental results show the good performances of the proposed method which will be compared with the instantaneous power method to highlight the feasibility and advantages of this method.

Modeling and Detection of Localized Tooth Defects in Geared Systems

The early detection of failures in geared systems is an important industrial problem which has still to be addressed from both an experimental and theoretical viewpoint. The proposed paper combines some extensive numerical simulations of a single stage geared unit with localized tooth faults and the use of several detection techniques whose performances are compared and critically assessed. A model aimed at simulating the contributions of local tooth defects such as spalling to the gear dynamic behavior is set up. The pinion and the gear of a pair are assimilated to two rigid cylinders with all six degrees of freedom connected by a series of springs which represent gear body and gear tooth compliances on the base plane. Classical shaft finite elements including torsional, flexural and axial displacements can be superimposed to the gear element together with some lumped stiffnesses, masses, inertias, … which account for the load machines, bearings and couplings. Tooth defects are modeled by a distribution of normal deviations over a zone which can be located anywhere on the active tooth flanks. Among the numerous available signal processing techniques used in vibration monitoring, cepstrum analysis is sensitive, reliable and it can be adapted to complex geared system with several meshes. From an analytical analysis of the equations of motion, two complementary detection techniques based upon acceleration power cepstrum are proposed. The equations of motion and the contact problem between mating flanks are simultaneously solved by coupling an implicit time-step integration scheme and a unilateral normal contact algorithm. The results of the numerical simulations are used as a data base for the proposed detection techniques. The combined influence of the defect location, depth and extent is analyzed for two examples of spur and helical gears with various profile modifications and the effectiveness of the two complementary detection methods is discussed before some conclusions are drawn.

Blind separation of convolved cyclostationary processes

A non-parametric method is presented for the blind separation of convolved cyclostationary processes such as those typically observed at the output of MIMO systems driven by periodically modulated random processes. The approach is formulated in the frequency domain and is deductive in the sense that it follows the lines of optimal supervised filtering--from which the relationships to be used in the unsupervised situation are derived. This leads to an algorithm where the successive diagonalisations of some cyclic spectral density matrices give rise to unique separating filters. One important result concerns the proposal of solutions to unambiguously recover the exact source permutation at each frequency. A statistical performance analysis of the method is also conducted, with the results suggesting some strategies to increase the robustness of the separation. Examples of successful separation are finally provided on realistic convolutive mixtures, both synthetic and from the real world, where impulse responses of several thousands of coefficients are dealt with.

Cyclostationary modeling of ground reaction force signals

The importance of the measurement of human locomotion for the processes of diagnosis and treatment of locomotion disorders is increasingly being recognized. Human locomotion, in particular walking and running are defined by sequences of cyclic gestures. The variability of these sequences can reveal abilities or motorskill failures. The purpose of this study is to analyze and to characterize a runners step from the ground reaction forces (GRF) measured during a run on a treadmill. Traditionally, the analysis of GRF signals is performed by the use of signal processing methods, which assume statistically stationary signal features. The originality of this paper consists in proposing an alternative framework for analyzing GRF signals, based on cyclostationary analysis. This framework, being able to model signals with periodically varying statistics, is better at showing the development of runners fatigue.

Chatter Detection in CNC Milling Processes Based on Wiener-SVM Approach and Using Only Motor Current Signals

The CNC milling is one of the most common processes in modern manufacturing, which characterized by highly nonlinear behavior and chatter problems. As other complex processes, Chatter detection in this situation is a crucial step for improving surface quality and reducing both noise and rapid wear of the cutting tool. This paper proposes a new methodology for chatter detection in computer numerical control milling machines. The originality of this method consists for using only motor current signals picked from electrical cabinet of machine and artificial intelligent. The methodology can be decomposed into four general tasks: (1) data acquisition, (2) signal processing, (3), features generation and selection, (4) classification. In signal processing task, electrical signals are resynchronized according to the electrical cycle (60 Hz) by exploiting the cyclostationarity of electrical signals through their cyclic statistics. After that, synchronous average is computed and subtracted from original signals in order to obtain the residual part. Wiener filter is then applied on residual signals by taking as reference the residual electrical signals acquired in spindle free rotation. This procedure allows estimating a signals corresponding to the electrical part and extracting the mechanical part, which linked to a chatter phenomenon and cutting mechanism. The signal processing task is primordial in order to decrease the dynamics of the electrical fundamental component and its harmonics and also to increase the contribution of mechanical parts. Extracted features computed from mechanical parts are then ranked based on theirs entropies in which only best features are selected and presented to the system for classification. At the classification step, the selected features are classified into two classes: stable and unstable utilizing a support vector machine (SVM). The intelligent chatter detection has accuracy above 96 % for the identification of cutting state after being trained by experimental data. The results show that it is possible to monitor chatter behavior in milling process by using motor current signal.

Some New Diagnostic Parameters for Reciprocating Engines

Condition monitoring through signal analysis is not as well established for reciprocating engines as it is for rotating machines. One of the reason is that it has to deal with non-stationary vibrations. In this paper, some new statistical indicators are defined for the detection and separation of close transient events. They can also be used for diagnostic purposes, for instance for valve mechanics or for combustion rises. Their performances are compared on real vibrations of a small 4 cylinder diesel engine and it is finally verified that they allow the reconstruction of the engine cycle.

Blind separation of cyclostationary sources using non-orthogonal approximate joint diagonalization

This paper presents a new technique for the blind separation of cyclostationary signals by exploiting the cyclostationary nonstochastic temporal-probability models (fraction on time FOT) for signals (time-series) with periodic structure. The proposed approach is based on the joint diagonalization nonorthogonal of a set of matrices which have the same structure, then it can be simultaneously separating all sources without any restrictions and distributions to the number of cyclic frequencies of each sources. Simulation results are provided to illustrate the effectiveness of the proposed approach.

Blind Separation of Cyclostationary Sources Sharing Common Cyclic Frequencies Using Joint Diagonalization Algorithm

We propose a new method for blind source separation of cyclostationary sources, whose cyclic frequencies are unknown and may share one or more common cyclic frequencies. The suggested method exploits the cyclic correlation function of observation signals to compose a set of matrices which has a particular algebraic structure. The aforesaid matrices are automatically selected by proposing two new criteria. Then, they are jointly diagonalized so as to estimate the mixing matrix and retrieve the source signals as a consequence. The nonunitary joint diagonalization (NU-JD) is ensured by Broyden-Fletcher-Goldfarb-Shanno (BFGS) method which is the most commonly used update strategy for implementing a quasi-Newton technique. The efficiency of the method is illustrated by numerical simulations in digital communications context, which show good performances comparing to other state-of-the-art methods.

Weighted feature selection with growing neural networks for the FDD of rolling element bearings

This paper suggests an automated approach for fault detection, diagnosis and identification of roller bearings, which is based on optimized form of growing neural networks. In our recent work, we selected features according to their classification accuracy within supervised learning stage. Since each one of selected features has different effect on classification decision, a weighted feature selection is put forward in this paper to improve the network taxonomy. This is followed by a self adaptive growing neural network that optimizes its architecture by adding or updating hidden nodes to fulfill the training requirements. This pattern recognition procedure used to recognize between signals coming from normal bearings and those generated from different industrial bearing faults. The developed approach is compared with two different types of supervised neural networks. Results demonstrate that the developed diagnostic approach can reliably separate different bearing fault conditions at various rotational speeds.

An improper random vector approach for ESPRIT and unitary ESPRIT frequency estimation

The problem of estimating the frequencies of a complex signal corrupted by noise is addressed in this paper. Solving the problem by a subspace approach induce an inevitable maximum overlap between windowed observation vectors. It appears therefore that traditional second order statistics do not describe totally the second order behavior and the notion of improper random vector is recommended. Based on this, we analyze an ESPRIT and a Unitary ESPRIT-based methods established with improper random vectors assumption. Numerical simulations and a real application are brought for embellishing the discussion.