James R. Ottewill

Stator circulating currents as media of fault detection in synchronous motors

Often found in critical, high power applications, synchronous machines require reliable condition monitoring systems. Large synchronous machines are typically designed with parallel connected windings in order to split the currents in parallel paths, delivering the total power at the terminals. Under ideal symmetrical conditions, no current will circulate between parallel branches of the same phase. However, when a motor fault breaks this symmetry, currents circulate between the branches. Thus, due to the fact that they are only non-zero under faulty conditions, circulating currents potentially represent a sensitive indicator of faulty condition. In this paper, the advantages of using the circulating current between parallel branches of the stator of a synchronous motor as an early indicator of motor faults are shown. Analysis is conducted both through simulation, via the use of finite element methods (FEM), and through experimentation using a specially-designed synchronous machine which allows various fault conditions to be investigated. Through comparison between experiment and simulation, the simulation tool is validated. Furthermore, it is shown that the circulating current is better suited for fault detection than either the branch or the stator current. It is concluded that an improved condition monitoring and protection system for a synchronous machine may be achieved if these currents are monitored.

Fault detection in electric motors by means of the extended Kalman Filter as disturbance estimator

In this paper, an approach for disturbance estimation in the stator phase currents of an induction machine is presented. The approach is based on the Extended Kalman Filter that uses the extended model of an electrical induction machine (IM) under healthy conditions. The extended model includes additional states (disturbances) that allow discrepancies between the model and the real system to be detected. It is demonstrated through simulation that the method is able to identify anomalies when unmodelled dynamics are induced. Subsequently, the values of the estimated disturbances may be used as inputs to a condition monitoring system in order to detect machine faults, helping to reduce the rate of spurious stops and false alarms, therefore improving the overall process efficiency. Additionally, the disturbances could be taken into account in the control system of the motor improving the machine performance.

A novel non-invasive method for detecting missing wedges in an induction machine

This paper presents a newly developed algorithm for evaluating the health of an induction machine. The proposed algorithm is based on spectrum analysis of an impedance calculated using measured stator current and voltage signals. The main idea is to calculate the frequency spectrum of the impedance for each power phase and compare specific differences between phases. Experimental investigations show that the method yields very accurate results and can form an important part of a machine monitoring system. In particular the presented method is shown to be successful in detecting missing wedges in electric motors.

Visual modeling of condition monitoring systems

In this paper a graphical approach for Condition Monitoring Systems (CM) based on Model Driven Architecture is presented; in particular, the software application “Smart Monitoring Agent” or SMA. This graphical approach starts from the idea of modeling, applying and visualizing non-hierarchical relations. The presented method makes use of well-defined data models that take the solution to a superior level of configurability, using only graphical tools. Finally it is explained how this approach enables the flexible knowledge exchange between condition monitoring experts and software engineers.

Ensemble Empirical Mode Decomposition and Sparsity Measurement as Tools Enhancing the Gear Diagnostic Capabilities of Time Synchronous Averaging

The paper introduces the concept of exploring the potential of Ensemble Empirical Mode Decomposition (EEMD) and Sparsity Measurement (SM) in enhancing the diagnostic information contained in the Time Synchronous Averaging (TSA) method used in the field of gearbox diagnostics. EEMD was created as a natural improvement of the Empirical Mode Decomposition which suffered from a so-called mode mixing problem. SM is heavily used in the field of ultrasound signal processing as a tool for assessing the degree of sparsity of a signal. A novel process of automatically finding the optimal parameters of EEMD is proposed by incorporating a Form Factor parameter, known from the field of electrical engineering. All these elements are combined and applied on a set of vibration data generated on a 2-stage gearbox under healthy and faulty conditions. The results suggest that combining these methods may increase the robustness of the condition monitoring routine, when compared to the standard TSA used alone.

A Frequency-Based Criterion for Automatic Selection of the Optimal Intrinsic Mode Function in Diagnostics of Localized Gear Tooth Faults

To date gearboxes remain one of the most important elements of virtually every power transmission system as far as a continuous operation of the shaft line is concerned. Any failure or breakdown may result in putting the whole production line, supply chain or even peoples life in jeopardy. Endeavours to detect an incipient fault within the system serve multiple purposes from increasing the safety of the people responsible for operating the machines, through to decreasing running and operational costs, allowing time to plan for the inevitable repairs and making sure that the downtime of the machine is kept to an absolute minimum. This, in turn, makes this branch of condition monitoring of rotating machinery one of the most intensively studied. The Empirical Mode Decomposition (EMD) is a relatively new method of signal decomposition, which breaks the original signal up into a number of so-called Intrinsic Mode Functions (IMFs). The decomposition represents a type of adaptive filtering which outputs a number of IMFs which, acquired according to two strict criteria, contain portions of the filtered version of the original signal and so can carry different information about the content of the signal. EMD has already been used in the field of condition monitoring of rotating machinery, but the selection of the optimal IMF for the task often requires the experience of a condition monitoring specialist. This paper proposes a frequency-based tool for automatic selection of the IMF that is best suited for the detection of localized gear tooth faults.

Diagnostic signal analysis using distribution of frequency gaps between spectral peaks

Many monitoring and diagnostics algorithms are based on analyzing the spectra of signals, such as vibrations, electrical currents and voltages, torques, etc. Often the analysis involves determining exact peak locations and looking at relations between various peak groups, such as harmonic trains or sidebands. These relationships can be more telling than the actual peak locations. Manual examination is aided by graphic tools for plotting the spectra, with features such as harmonic and sideband cursors. Automated analysis often relies on methods such as cepstra, or autocorrelation. This paper presents advantages of using another method: histograms of inter-peak gaps. In contrast to methods that operate on the whole spectrum (such as autocorrelation), this method relies on finding a discrete number of peaks and exploring the distribution of gaps between them. In particular, a histogram of frequency differences between all peak pairs is constructed. The paper presents details of this method, discusses its strong and weak points, and provides examples of its application to actual data: vibration and acoustic signals involving bearing defects, as well as a motor supply current.

Parallel Autoregressive Modeling as a Tool for Diagnosing Localized Gear Tooth Faults

One of the standard approaches widely used in the field of localized gear tooth fault diagnosis is the creation of residual signals i.e. signals obtained after removing the deterministic frequency components from a Time Synchronously Averaged vibration signals. Most of the time these components are removed based on the knowledge of the characteristic gearbox frequencies. Sometimes however such information is not available. AR modeling, a type of time series modeling, has been found to be capable of faithfully estimating the deterministic content of the signal allowing meaningful residual signals to be created. An improvement to the classic AR modeling approach is proposed in this text. The method is applied to experimental data taken from a gearbox in both healthy and faulty condition. The improvement derived from the new method is quantified through a comparison with results obtained by applying Time Synchronous Averaging and the classic AR modeling method to the experimental data.

On the detection of missing magnetic wedges using impedance analysis

In this paper, a method for detecting missing magnetic wedges in induction motors supplied both direct-online and via variable-speed-drive is presented. The proposed algorithm is based on spectrum analysis of an impedance calculated using measured stator current and voltage signals. The main idea of proposed algorithm is to first normalize current and voltage signals in order to remove variations due to non-stationary operating conditions. Once normalized, the signals are processed to obtain the frequency spectrum of the impedance for each power phase and, subsequently, specific differences between phases are analyzed to identify faults. To verify these algorithms three induction motor cases were examined. Two were supplied direct-on-line and one supplied by variable-speed-drives. In the case of direct-online-supply one of the motors was known to be healthy and the other was known to have missing magnetic wedges while in the case of variable-speed-drive-supply one motor was known to have missing magnetic wedges. Measurements were recorded and processed using an ABB portable condition monitoring tool dedicated for electric motors. Experimental investigations show that the method successfully identifies defective cases.