Advances in Condition Monitoring of Machinery in Non-Stationary Operations

Condition monitoring is a central aspect in the health assessment and maintenance of industrial machinery. Vibration analysis is the most widely used technique for fault detection in rotating machinery. However, the technique can become difficult to apply in the case of machinery with non-stationary duty cycles due to the time-varying characteristics of the machine vibrations. The vibration signature of an incipient fault in rotating machinery is typically weak when compared to other sources of excitation. Due to these limitations, many methods have been proposed to increase the signal to noise ratio (SNR) of the signals as well as their applicability to non-steady operation. These include the separation of the random fault signatures from the deterministic components in the signal as well as techniques based on optimising the filtering of the signal to increase SNR. This work presents a method for extracting weak fault signatures from non-stationary signals using a reference signal from a parallel operating component on the same machine. The method, which is based on wavelet de-noising, employs a reference signal to adapt noise thresholds in the time and scale domain. Tests were performed using simulated non-stationary vibration signals. The proposed technique is shown to be effective at increasing the SNR when combined with envelope analysis to detect and diagnose faults.

The growing use of rotating machines operating in non-stationary conditions gave rise to a greater need to a higher precision for describing their dynamic behavior. The latter has always been based on a certain number of simplifying assumptions. In particular, the spinning speed is considered either constant or following a given law of variation as a function of time, resulting in a dynamic model that is limited to specific operating conditions. The aim of this work is to present a more general dynamic model of rotating machines, which accurately reflects its behavior in real working conditions. No assumption is made on the speed of rotation; it is included as an unknown of the dynamic problem by introducing a degree of freedom combining both the free body rotation and the torsional deformation. The instantaneous angular speed (IAS) is then deduced not only from the induced torque, but also from the whole dynamic behavior of the structure taking into account the periodic geometry (e.g.: gears, bearings) as well as the operating conditions (e.g.: going through the critical speeds). Making no assumption on the angular speed leads to a new formulation of the gyroscopic effect strongly present at very high speeds. This new formulation shows a coupling between the different degrees of freedom as well as a nonlinear behavior of the structure. The results of both classic and new formulations are compared for an architecture of a rotating machine to highlight the utility of the innovative approach in non-stationary operating conditions.

Gear wear is hard to monitor in lifting cranes due to the difficulties to provide appropriate models of such complex systems with varying functioning modes. Statistical machine learning offers an elegant framework to circumvent these difficulties. This work explores recent advances in statistical machine learning to provide a data-driven model-free approach to monitor lifting cranes, by investigating a large number of indicators extracted from vibration signals. The principal contributions of this paper are twofold. Firstly, it explores the recently introduced Neighbor Retrieval Visualizer (NeRV) method for nonlinear information retrieval. The extracted information allows to construct a low-dimensional representation space that faithfully depicts the evolution of the system. Secondly, it proposes a simple and efficient detection method to detect abnormal evolution and abrupt changes of the system at hand, using the distance measure with neighborhood retrieval in the same spirit as NeRV. The relevance of the proposed methods, for visualizing the evolution and detecting abnormality, is demonstrated with experiments conducted on real data acquired on a lifting crane benchmark operating for almost two years with more than fifty indicators extracted from vibration signals.

Planetary gear transmissions are widely utilized in rotating machinery which is running under stationary or non-stationary conditions.In this work, a numerical study of a planetary gear transmission is investigated in both run up the regime and stationary condition.The non-linear dynamic behavior of a single stage planetary gear during these two regimes was studied. The non-linearity is induced by the Hertzian contact force between teeth gear and it is implemented in a torsional lumped parameter model. In addition, the gear system is excited with the motor torque variation and the fluctuation of ring-planets and sun-planets mesh stiffness during the non-stationary regime. The system equations of motion were resolved by using the implicit type numerical integration technique Newmark-β with Newton-Raphson method. The obtained numerical results approve the influence of the Hertzian stiffness on the dynamic behavior of the system, especially in the run-up regime.

In this paper, a method using fully distributed optical fiber sensing is proposed for measuring the temperature profile induced during drilling in carbon fiber reinforced polymers (CFRP) laminates. By loosely inserting the sensing optical fiber into some array capillaries, the strain-free temperature distribution around the bored hole has been investigated experimentally while drilling CFRP plates. A full two-dimensional (2D) map of the dynamic temperature evolution is reconstructed continuously from the data acquired using optical frequency-domain reflectometry (OFDR). The developed OFDR sensor offers the possibility of distributed sensing with a spatial resolution of 2.61 mm. The data of the high spatial resolution distributed optical fiber sensor can be mapped to the temperature distribution in the vicinity of bored hole by two-dimensional reconstruction, similar to meshing a workpiece by using a finite element method. Therefore, 2D temperature distribution over the CFRP has been easily observed. Using an acquisition rate of 100 Hz, the time evolution of the 2D temperature map is captured with high precision, allowing us to analyze the impact of the carbon fiber and drilling directions on the temperature distribution. The method provides a reliable tool to study thermal damage of CFRP components and the dissipation of cutting heat while drilling.

Variable speed wind turbines use the available wind resource more efficiently than a fixed speed wind turbine, especially during light wind conditions. This enhancement forces the monitoring methods to deal with these large variations in speed and torque, since the conditions are seldom if ever stationary. The unsteady behavior of these wind turbines is also a difficulty in terms of long term diagnostic, since the comparison of successive measurements is usually performed under the same operating conditions. Normalization of the indicators according to well-chosen variables might bring a valuable tool regarding several aspects.In this paper, the attention is focused on the regression process using classical machine learning tools. The difficulty is to design a process able to efficiently estimate the behavior of any indicator depending on the environmental conditions. Indeed, indicator multivariate laws are expected to present extremely varied shapes, and using common linear regression technique can hardly solve this issue. Kernel machines are therefore presented in this paper as an efficient solution to normalize the indicators, and will be shown to ease the health monitoring of the wind turbine shaft line on a practical case based on instantaneous angular speed signals. The example presents the distinctive feature to have a defect visible only specific operating conditions. This operating conditions being unknown a priori, this example clearly enlightens the need of such a regression tool.

In the paper the long-term temperature data from LHD (load, haul, dump) machine from underground copper ore mine are analyzed. The main problem is to detect the moment when the temperature increases due to change of condition. Usually in condition monitoring system the problem is solved by selection of fixed threshold and observation if the temperature data exceeds this limit value. However this approach seems to be insufficient for the real data that are influenced by various factors related to harsh operating conditions in underground mine. In case of change of technical condition, events of exceeded temperature do not occur locally in time but affect the statistical properties of the temperature data for longer period of time. The key task could be defined as identification so called structural break point in raw signal based on statistical analysis in longer time window. In this paper a new method for detection of the structural break point of temperature data from LHD machine based on regime variance approach is presented. The data are investigated here as signals with two regimes behavior (good/bad condition). We select the most suitable critical point in order to separate different regimes. The introduced methodology is fully automatic and is based on simple statistics of the temperature signal.

Fault diagnosis is vital for the health management of rotating machinery. The non-stationary working conditions is one of the major challenges in this field. The key is to extract working-condition-invariant but fault-discriminative features. Traditional methods use expert knowledge on the machines and signal processing to extract fault features from vibration signals manually. This paper regards this issue as a domain adaption problem and utilizes deep learning technique to learn fault discriminative features automatically. We teach deep Convolutional Neural Networks to pronounce diagnostic results from raw vibration data and propose a Rotating Speed Normalization method to improve the domain adaption ability of the neural network models. A case study of rotor crack diagnosis under non-stationary and ever-changing rotating speeds is presented. Using 95600 signal segments, we compare the diagnostic performance of ours and reported Convolutional Neural Network models. The results show that our model gives solid diagnostic accuracy from non-stationary vibration signals, and the proposed Rotating Speed Normalization method can successfully boost the performance of all investigated CNN models.

This paper presents a framework for the condition monitoring of large non-linear electro-mechanical systems. Faults are assessed in different subsystems by means of joint state-parameter estimations. The different estimations are interconnected by a Digital Twin of the system, which provides background information for the physics neglected in the estimation models. In addition, the Digital Twin provides a benchmark for the estimations, which allows identifying the location and extend of faults. As application case, the condition of the guiding system and the electric machine of a vertical transportation system are evaluated. For this purpose, a scaled test bench of the vertical transportation system is used, showing the potential of this approach in the condition monitoring of complex industrial systems.

Condition monitoring performed directly from the estimated instantaneous angular speed has found some interesting applications in industrial environments, going from bearing monitoring to gear failure detection. One common way to estimate the angular speed makes use of angular encoders linked to a rotating shaft. At the opposite of traditional time-sampled signals, encoders describe purely angular phenomena often encountered in rotating machines. However, rotating encoders suffer from various geometric defects, corrupting the measurement with an angular periodic signature. The angular synchronous average is a very popular tool to estimate this systematic error, but is only adapted to constant speed conditions, which is rarely the case in real applications. We propose here two different estimators to compute a robust estimation of the synchronous component in variable speed conditions. The former, as a data-driven approach, is based on a local weighted least squares method, while the latter is a model-based approach. We study the behaviour of our estimators with both simulations and experimental signals, and show the relevance of the proposed method in an industrial context.

The application of machine learning to fault diagnosis allows automated condition monitoring of machines, leading to reduced maintenance costs and increased machine availability. Traditional approaches train a machine learning algorithm to identify specific faults or operational settings. Therefore, these approaches cannot always cope with a dynamic industrial environment. However, an industrial installation often contains multiple machines of the same type, which enables a fleet-based analysis. This type of analysis compares machines to tackle the challenges of a dynamic environment. In this paper a novel method is proposed for analyzing a fleet of machines operating under similar conditions in the same area by using inter-machine comparisons. The proposed methodology consists of two steps. First, the inter-machine difference is calculated using dynamic time warping by using the amount of warping as measure. This method allows comparing the measured signals even when fluctuations are present. Second, a clustering method uses the inter-machine similarity to identify groups of machines that operate in a similar manner. The generation of a fault usually causes a change in the raw signals and diagnostic features. As a result, the inter-machine difference between the faulty machine and the rest of the fleet will increase, leading to the creation of a separate group that contains the faulty machine. The methodology is evaluated and validated on phase current signals measured on a fleet of electrical drivetrains, where a phase unbalance fault is introduced in some of the drivetrains for a specific time duration.

A data-driven methodology is considered in this paper focusing towards the Remaining Useful Life (RUL) prediction. Firstly, diagnostic features are extracted from training data and an analytical function that best approximates the evolution of the fault is determined and used to learn the parameters of an Unscented Kalman Filter (UKF). UKF is based on the recursive estimation of the Classic Kalman Filter (CKF) and the Unscented Transform, presenting advantages over the Extended Kalman Filter (EKF) for high non-linear systems. The learned UKF is further applied on testing data in order to predict the RUL under different operating conditions. The influence of the starting point of the prediction is analyzed and a method for the automated parameter tuning of the Kalman Filter is considered. In the end, the result is evaluated and compared to CKF and EKF on experimental data based on dedicated performance metrics.

Centrifugal pumps are rotating machines which are widely used in process operations and other applications. Efficient and failure-free operation of these pumps is important for effective plant operation and productivity. However, the complexity of pumps, combined with continuous operation, can lead to failure and expensive maintenance requirements, thus motivating interest in on-line condition monitoring. This work investigates the application of frequency-domain analysis for classification of various centrifugal pump conditions. Seven conditions are considered, namely, healthy (non-faulty); five mechanical faults: misalignment, imbalance, faulty bearing, faulty impeller and mechanical looseness; and a hydraulic fault: cavitation. A centrifugal pump test rig was built specifically for this research to simulate each fault condition and acquire the resulting vibration measurements. Signals are acquired from the pump using an accelerometer which is mounted on its bearing housing, and a data acquisition device (DAQ) is used to acquire the signals. Fast Fourier Transform (FFT) is used to identify which types of faults generate stationary or non-stationary signals. However, the accuracy of detection for the non-stationary signals is shown to require further improvement, and alternate methods are suggested accordingly.

Rolling element bearings are one of the most important component in every rotating machinery. As a result, their diagnosis before occurrence of any catastrophic failure is of vital importance and vibration based diagnosis is very popular approach. In this paper, the performance of a recently proposed method, Autogram, will be investigated on different data sets provided by Politecnico di Torino and University of Cincinnati. The results will be compared with other well-established methods such as Fast Kurtogram and Spectral Correlation.

Vibration based analysis is a proven technique in condition monitoring of rotating machinery. Particularly in wind turbines, incipient fault detection based on vibration signature analysis plays a vital role in reliable operation and maintenance planning. The synergistic effect of non-stationary Vibration signals due to wide variation in wind speed and non-linearity due to inherent turbine load dynamics is a major challenging issue. Extraction of fault signatures from the non-stationary vibration depends on an efficacious decomposition of multi components into mono component intrinsic mode functions. In this work, an Adaptive Empirical Wavelet Decomposition (AEWD) based feature extraction technique with Quadratic Kernel Function Support Vector Machine (QKSVM) classifier is proposed to detect mechanical faults in Wind Turbine Planetary Epicyclic Gears. Instantaneous Amplitude and Frequency components are reconstructed from Empirical wavelet coefficients. Sample Entropy, Permutation Entropy, Signal descriptors (RMS, Peak, and Crest Factor) and Statistical moments are extracted to frame the hybrid feature space. Correlation analysis of hybrid features clearly shows non-linear distribution. The performance of proposed AEWD–QKSVM is analyzed with practical Wind Turbine Gearbox Condition Monitoring Vibration Analysis Benchmarking Datasets. The results are proven 91.5% accuracy with 2000 time segmented sampling vibration signals in scuffing fault detection in planetary stage with good performance index as 90% sensitivity and 93% specificity in fault detection.

In the paper we will show specific case study related to long-term temperature data from electric motor bearings (with progressing fault) used in belt conveyor operating in open-cast mine. Existing SCADA system for data acquisition has built-in simple decision making rules based on static thresholds. Due to time-varying environmental and operational conditions, i.e. machine is heavily influenced by ambient temperature (−20 up to +30 $$^\circ $$ C) and external load (no operation, idle mode, startup with heavily overloaded belt). Hence, basic analytical methods based on simple statistics are sometimes not sufficient to determine the change of technical condition of the bearing. In order to address this issue authors propose an analytical method based on multidimensional distribution analysis. Finally, a clustering method can be applied to multidimensional representation of the initial data. This approach allows to differentiate the technical condition across the investigated time period.

When a ship is design and developed, its engines are installed accordingly the environment conditions in a designated operation area and the owner requisites. Sometimes they are not design and adapted to sail in all seawaters temperatures. This implies that in some cases the engines power output cannot be the same for the safety of it. If an engine was conceived to operate on 16 ℃ of seawater temperature, when it navigates on a 36 ℃ sea, in some engines the power must be limited to not initiate major damage. Due the fact, in this study a diesel engine will be monitored with online data collection and statistical treatment. The statistic treatment will be done with the modified EWMA control charts. With the engine, operating on a range of seawater temperatures from 16 ℃ to 36 ℃, the auxiliary systems must be flexible due to ship performance with the aim of maintain it.

A proper maintenance of train basic systems is a key aspect in the comfort and safety of that train, especially in high-speed rail. One of the most critical systems in the operation of a train is the bogie, a very complex mechanical system made up of several elements that interact between them. Bogie vibrations are the result of multiple mechanical connections, which are generated by the different components involved in the dynamic or structural behavior of the bogie. The axle box is the element in which the sensors of the monitoring system are located and whose information is the essence of the predictive maintenance process of the train. In this work, it is studied the vibratory behavior of the railway running gear system of a high speed train, in commercial service, after a maintenance operation. Vibration signals are from sensors located in the axle box and will be processed using the Empirical Mode Decomposition (EMD) technique. The EMD technique decomposes the temporal signal into some elementary intrinsic mode functions (IMF), which are the result of progressive envelopes of the temporal signal and that work as bandpass filters. The spectral power of each IMF reflects the frequency behavior of the vibratory signal for the frequency band associated with each IMF. The evolution of these IMF spectral powers will be studied before and after the maintenance intervention, so we can determine if this evolution can be used as an indicator of the operating state of the railway mechanical system.

The paper presents the results of a study aimed to compare different signal processing techniques for the condition monitoring of a mechanical system for indexing motion. Artificial feed-forward neural networks (ANN) are used as classifiers. The mechanical system can work in different conditions (variable loads and velocities, lubricant oil with different viscosity) and the ANN identifies the working condition. The monitored variable is the acceleration signal of the rotating table, opportunely pre-processed. The signal processing techniques compared are: Power Spectral Density (PSD), Fast Fourier Transform (FFT), Wavelet, Amplitude Probability Density Function (PDF), Higher Order Spectra (HOS).

Until the last century, vibration analyses have been the most used method in monitoring the rotating machinery. But nowadays the electromechanical interaction in machinery is being the new scientific trend, mainly for its accuracy and facility.Several research papers have shown that investigations done on the motor can be used to describe the machines dynamic behaviour and give a deep overview of any anomaly.Within this context, this work reports the impact of an introduced pitting in the sun of one of the gearboxes of a back-to-back planetary gearbox configuration on the stator current signal. The gearbox monitoring was investigated through analysing the stator current measured experimentally by clamp meter.Later, the results presented in the frequency spectrum are obtained using Fast Fourier Transform (FFT) to distinguish the frequency of the defect and its harmonics besides to the mesh frequency which highlight the electromechanical coupling.Finally, the stator current signals were followed by the spectrum of acceleration signals. These signals were registered using an accelerometer mounted on the fixed ring in order to validate results foreseen in the current spectrum.

Comfort plays a very important role in the development of cars. The pistons firing order of an internal combustion engine, by nature, do not generate a constant torque on the crankshaft. At specific crankshaft speed, where the engine’s excitation frequency and the driven-system’s natural frequency coincide, torsional vibrations can excite crankshaft resonances and other dynamic phenomena within the engine or further down the driveline, potentially leading to fatigue failure of the crankshaft, or to important NVH problems. In order to achieve very high standards, the behavior of the driveline system needs to be investigated. For this reason a technique named Torsional-Order Based Modal Analysis (T-OBMA) has been developed in order to identify the torsional resonance frequencies and their related damping ratios of driveline systems during operating conditions. The technique is based on the rotational speed measurements acquired in two or more points along the driveline. The modal parameters are identified from torsional orders measured during an engine speed runup. The technique has been validated in a simulation and in an industrial test environment identifying crankshaft modal parameters on a Ferrari engine.

Independent cart conveyor system is an emerging technology in industries, trying to replace servo motors and kinematic chains in several applications. It consists of several carts on a closed-loop path, each of which can freely move with respect to the other carts. Basically, each cart is an servo linear motor, where the windings and the drives are on the frame and the magnets are on the moving carts together with a feedback device (e.g. a Hall sensor to track the position). The drive controls and actuates each cart independently according to the motion profile loaded. From a mechanical point of view, the carts are connected to the frame through a series of rollers placed on and under a mechanical guide. The rollers may be subject to a premature wear and the condition monitoring of these components is a no trivial challenge, due to non-stationary working conditions of variable speed profile and variable loads. This paper provides a bearing fault model taking into account the motion profile of the cart, the mechanical design of the cart, the geometry of the conveyor path, the expected loads and the type of fault on the roller bearings.

A methodology to detect if a fan matrix is working properly has been designed and is presented in this paper. This methodology is based on a Support Vector Machine (SVM) classifier that uses geometrical parameters of the acoustic images of the fan matrix. These acoustic images have been obtained using a 16 × 16 planar array of MEMS microphones working at different frequencies. A fan matrix that is not working properly implies that some of its fans have failed, that is, it does not work. The designed fault detection methodology supposes that these fans fail one by one. If one of the fans is not working, this fact can be detected rapidly with the purposed methodology, and the fan can be repaired or replaced by a new one. Although it is really unusual that more than one fan fails at the same time, this paper also studies how this methodology works if the number of faulty fans increases, in order to know if the methodology is robust enough in the presence of unexpected situations.

Metal mesh isolator is made of metallic wires. It has been widely used in vibration control engineering applications such as isolation mounting of machine tools. To investigate the performance of the metallic wires material, a set of dynamic tests was conducted for a range of frequencies and amplitudes of loading. The experimental results has demonstrated that the output of the isolator is revealed to the loading amplitude, however, slightly dependent to the loading frequency. As the loading amplitude increases, the dynamic mechanical property exhibits asymmetrical characteristic. Therefore, a model that includes the asymmetric non-linear elastic force, viscous damping and hysteretic coulomb friction is setup to describe the dynamic general restoring force. In this paper, an experimental identification methodology is presented to determine the unknown parameters of the constitutive mechanical model. The Bouc-Wen model was implemented to identify the unknown parameters of the hysteretic damping force. In order to measure the equivalent loss factor of nonlinear material, a damping capacity measurement method, based on the decomposition of the hysteresis loop, is brought forward. The equivalent loss factor of the metallic-wires material at different loading frequencies and amplitudes were measured through a damping capacity measurement method, based on the decomposition of the hysteresis loop. The results show that this material has excellent damping performance with loss factor about 0.4–0.5 for lower frequency and amplitude.

Planetary gearboxes failure has been a major problem in systems reliability, hence condition monitoring of this component is essential. In order to identify faults, a dynamic model of the gearbox is presented taking into account backlash and tooth cracks. Newtonian equations of motion developed for a planetary gear is solved and modal analysis for solving the lumped parametric differential equations of modelled single stage planetary gearbox is adopted to observe the vibrational responses and to compare with existing literature. This work focuses primarily on the effects of stiffness and damping. The eigenvalue problem has been solved to measure the natural frequencies and amplitude of the system. Parametric studies using stiffness and damping results indicate significant changes in vibrational signals when there is a crack present on tooth. Simulation results also show the relation between increment in backlash and high frequency responses in the resultant vibration.

The soil compaction is very important process during building and road construction. There are many various machineries used for improve the soil load baring capacities like rollers or vibrating plates. We can compact not only the soil but the concrete or very popular composite as asphalt - bituminous composition as well. Sometimes these procedures are implemented close to buildings or other sensitive to vibration construction. Currently we have some application of compaction level monitoring. But sometimes we need to create the model of ground behavior for future prognoses. In paper the problem of soil compaction and 3D cellular model of ground which describes process of ground flow, displacement and compaction will be presented. This model is made in MATLAB/SIMULINK software and describes dynamically changing soil feature. For better presentation model was simplified to 2D version. Basing on simulation results the concept of smart compactor was created and described.

The majority of structures are subject to dynamic loads resulting from vibrations and shocks. Studies conducted by the researchers show that transversal loads are the main causes of the loosening phenomenon. The purpose of this paper is to expose an experimental study of the loosening phenomenon of an assembly subjected to dynamic transversal load resulting from vibration. The assembly will be the subject to vibrations resulting from a shaker. It consists of an inertial mass, bolted on a bracket attached to the shaker, which is animated with a vertical sinusoidal movement at a fixed frequency. The inertia of the mass will be exploited to generate its sliding. In fact, under the influence of the imposed acceleration, the inertial mass is subject to a shear load that overcomes the friction one. Sliding of the mass will cause the loosening phenomenon. The Assembly will be instrumented by a high-speed camera to track the rotation of parts during the vibration; an accelerometer for measuring the imposed acceleration and a load washer for measuring the screw preload. Results obtained from this study showed the loosening phenomenon for an assembly with one and two fixations. A parametric study will be elaborated in order to identify the critical case for the loosening phenomenon.

Ferroresonance is a very studied phenomenon by the distribution electrical companies since it can cause damage to transformers and other line elements. This problem is generated by the association in series or in parallel of capacities (overhead and underground power lines) and saturable inductances (voltage transformers). During the ferroresonance, the Voltage Transformer (VT) becomes saturated and the magnetic field varies in a high way; all this produces that the ferromagnetic material dimensions of the transformer change due to the phenomenon of magnetostriction. Thus, magnetostriction generates a vibration on the transformer. The main effects produced by ferroresonance are as follow: core saturation, voltage and current distortion, overvoltages and overcurrents, temperature increase, vibration and noise. The aim of this paper is to detect the ferroresonance by analyzing the vibration on the VT.

Toward ever improvement technology, a diagnostic procedure making use of Dual Q-Factor wavelet decomposition (DQWD) and adaptive wavelet transform (AWT) is proposed for the extraction of weak bearing defect feature. The vibration signal of bearing consists of mix of transient impulse (low-Q factor) and oscillatory signal (high-Q factor signal). Therefore, to separate the two different behavioral signals, Dual Q-factor wavelet decomposition is carried out. The DQWD decompose any signal into low-Q factor and high-Q factor signal. Further, extraction of feature is carried out by adaptive wavelet transform. For this adaptive wavelet is extracted from the low-Q factor signal using least square fitting method. The generated wavelet is applied to low-Q factor signal to produce AWT scalogram. Then, coefficients of resulting scalogram are integrated with respect to scale for each time segment. Then, envelope demodulation is applied to the resulting waveform to spot the defect frequency. An experimental study is presented to show the effectiveness of the proposed method. The proposed method is also effective over EMD and EEMD technique in isolating the transient impulse of defect from the oscillatory part of the signal.

The instantaneous speed of a machine can be accurately determined by frequency demodulation of a tacho or shaft encoder signal, but can often be determined by frequency demodulation of a machine shaft order in a response vibration signal. It has recently been realised that the latter is not necessarily the same as the speed variation of the machine, as it varies depending on whether displacement, velocity or acceleration is being measured, and whether the band being demodulated falls on a spring line or mass line of the frequency response function. A widely used procedure for performing frequency demodulation of a signal using the Teager Kaiser energy operator (TKEO) actually gives an estimate of the frequency modulation of the derivative of the signal, and this paper discusses alternative more accurate ways of extracting the frequency modulation of the signal itself, and the errors involved in the different estimation methods. The discussion is illustrated using simulated signals, and measurements made on a variable speed gear test rig, with accelerometer response measurements as well as a tacho signal.

The purpose of this paper is to develop a methodology that utilises the order-frequency spectral coherence to detect novel (i.e. unobserved) second-order cyclostationary components. In the novelty detection methodology, a probabilistic model of the healthy data is utilised to detect, localise and trend novelties in the form of damage that manifest as second-order cyclostationary components in vibration signals. The methodology is unique in the sense that on the one hand, the spectral properties are retained and multiple harmonics of the fault frequency component are used during the condition inference process; while on the other hand, the methodology is simple and efficient to implement. A numerical gearbox model is used to generate vibration signals and to simulate bearing and distributed gear damage. The methodology is applied to the simulated vibration signals, generated under varying speed conditions, which demonstrates very promising results.

In modern industry, the wide use of reliable and sophisticated sensors with their connection to internet has introduced the phenomena of Big Data, especially in the field of condition monitoring systems (CMSs) in e-maintenance applications. In particular, in the case of vibration signals, high-performance acquisition systems are required, characterized by anti-aliasing filtering and high uniform sampling rate, in order to properly digitalize the meaningful frequency content of the signals. In this context, the capability of non-uniform random sampling (RS) is assessed in this work. While in different fields, such astronomy, structural and biomedical studies, the RS is a problem to be resolved, due to the unavailability of samples at specific instants (missing data problem), in the field of fault detection & diagnosis (FDD), RS can be a chosen sampling method thanks to its advantages: anti-aliasing property and low average sampling rate. Therefore, this paper focuses on studying the anti-aliasing property of the random sampled data, verifying the criterion proposed in literature for establish the Nyquist frequency, and analyzing its sensitivity to the sampling parameters. This study is carried out using simulated signals and computing the spectral window, giving the Nyquist frequency for different random sampling parameters; moreover, a spectral analysis method, the Schuster periodogram, is used to verify when the spectrum is actually free of alias. The results show that the Nyquist frequency depends on the numerical accuracy of the randomly generated time instants.

The control quality and technical condition of the engine control system (fuel control unit, FCU) plays a key role in the safe operation of aircraft. The article presents a comprehensive approach to diagnosing FCU in transient states. At the beginning, examples of damage to aircraft engines caused by FCU’s incorrect work were presented, and the difference in perception of FCU by the designer and diagnostician was indicated. A transfer of the status observer was proposed for a reliable diagnosis of the FCU technical condition (reverse problem). Instead of registering and analysing several parameters (including pressure and fuel flow, exhaust temperature, pressure, etc.) and temperatures at the inlet to the combustion chamber) in steady states, the registration of the instantaneous angular speed (IAS) in the transient states of the engine and the analysis of its parameters on the phase plane is proposed. Next, the methodology of active experiments was discussed, in which the functions of FCU components and new diagnostic symptoms were identified. The methodology for creating expert software has been approximated. Finally, the experience of the long-term operation of the method in Poland has been presented. The very high efficiency of the diagnostic method in identifying hidden FCU failures and regulation errors has been demonstrated, which is also influenced by human factors. The topics discussed will be illustrated by examples.

In this paper we discuss a problem of local damage detection based on the vibration signal analysis. One of the classical approach is to extract features of the analyzed signal that differ for damaged and healthy case. We propose to test the integration property in order to check if given signal corresponds to healthy or damaged machine. The integration issue is known from the econometric analysis. However actually this methodology is used in various fields including operational condition monitoring and fault detection. We say the signal is integrated with order d if after differentiation d times it becomes stationary. In the proposed procedure we extract the appropriate subsignals from the original raw signal and use the Kwiatkowski–Phillips–Schmidt–Shin (KPSS) statistics in order to test if they are integrated. We expect that for the healthy case the subsignals are integrated, therefore the KPSS test does not reject the H0 hypothesis of integration. For the damaged case the subsignals containing the impulses related to damage are not integrated therefore the H0 hypothesis is rejected. This approach is a continuation of the authors’ previous works and allows to detect the local damage by the inspection of the KPSS statistics. We apply the methodology to the real vibration signals from gearbox.

Condition monitoring of rotating machinery is a field of intensive research being closely related to the technological evolution in the area of energy, manufacturing and transport. The fault detection and diagnosis of planetary gearbox bearings present an interesting challenge even under steady operating conditions as often the relatively weak bearing signals are masked by other components such as the gear meshing. However, planetary gearboxes usually operate under time varying speed operating conditions (wind turbines, helicopters etc.), complicating even more the diagnostic procedure. The paper focuses on the diagnosis of bearing faults operating under steady and time varying speed conditions applying advanced cyclostationary and cyclo-non-stationary tools. Lately, spectral analysis based on the Cyclic Spectral Coherence (CSCoh) has shown improvement on bearing diagnostics, compared to classical methods, such as the Envelope Analysis. The integration over an optimal band of the CSCoh map leads to the estimation of an enhanced envelope spectrum with strong diagnostic capabilities. While this method seems to work successfully at constant speeds, varying speeds result in cyclo-non-stationary signals, rendering the method unsuccessful. Therefore, re-sampling the signals in the angular domain (using a measured or extracted speed signal) before further processing the signals with CSCoh is used in order to remove the influence of the speed. Furthermore criteria for the selection of the optimal band on the CSCoh (in the angular domain) are proposed and tested. The methodology is applied and evaluated on bearing signals captured over a planetary gearbox operating under steady and time varying speed conditions.

The technical condition of the power unit plays a key role in flight safety. The article presents concept of the diagnosis of the aircraft’s propulsion unit based on the vibration signal using lab-stand dedicated to HUMS – Health and Usage Monitoring System. At the beginning of the paper is presented the differences between the aircraft engine and the power unit. The stationary rotating machine which modify the use of the vibration signal in diagnostics are indicated. The theoretical foundations of vibration diagnostics of aircraft engines and power unit are discussed, indicating the main sources of excitation and the impact of constructional features of the machine on vibration propagation. Next, the methodology of active experiments was discussed, as part of which broadband vibration spectrum and new diagnostic symptoms were identified. The methodology of creating expert software for the needs of the HUMS has also been approximated. Finally operational experience was presented based on which the need for further research and development in the field of optimization of measurement data analysis methods, software and inference methods was demonstrated. The topics discussed will be illustrated by examples.

Nowadays, huge emphasis is given to research on diagnostic tools in order to prevent and monitor the health status of gears and bearings. However, the link between advanced signal processing techniques and ease of use is still missing in commercial software tools. Actually, softwares that implement advanced signal processing techniques leak in ease user interaction and automated diagnostic procedures. Authors have developed a commercial software tool, called UniVibe, that attempts to fill the gap between high sensitivity in the diagnostics of faults in complex geared transmission and user-friendly interface. This work focuses on the description of the UniVibe core, highlighting its diagnostic capabilities on the basis of a real industrial case. Specifically, the automated procedure that shepherds the user to the successfully fault diagnosis of a complex geared transmission is pointed out.

Condition monitoring of the rotating machines plays a key role in their maintenance. It is challenging task, especially in case of machines operating in the varying conditions (e.g. wind turbines, road headers). Typically the spectral analysis is used for the damage detection. Undoubtedly, this method is suitable for simple signals in order to analyse the energy of the signal. The local fault reveals in the vibration as a pulse train with high energy. However, for complex data with many different components and high contamination it can be insufficient to analyse only the envelope spectrum. In order to detect fault in such signals the cyclostationary approach can be applied, which gives a possibility to detect many different sources of faults. In this article the long term data is analysed, in particular there are presented results for the simulated and real case. For each observation the bi-frequency map is computed. It is shown that analysing the modulation frequency we are able to track the development of the damage. The results are compared with the classical spectral approach.

A problem of local damage detection for condition monitoring based on vibration data can be approached from many different angles. One of the most common ways is selective filtration of the vibration signal. There are many techniques allowing to construct digital filter for particular input data (e.g. spectral selectors). In previous articles authors proposed a technique called Progressive Genetic Algorithm (PGA) to optimally design digital filter for a given data set using no prior assumptions. It uses kurtosis as fitness function and local linear fit of fitness function progression vector as a global termination criterion (GTC), but local termination criterion (LTC) was defined as simple stall limit of fitness value. In this paper authors propose a new quantile-based way to terminate PGA locally for faster convergence. Initial testing phase shows that for comparable quality of obtained result, individual epochs terminate significantly faster without sacrificing the progress of local convergence. It results in more efficient optimization and faster global convergence which reduces the overall execution time of the program for about the order of magnitude.

Heavy-duty machines are often working in the harsh conditions. Components of such machine will suffer significantly higher stress and will tend to wear off more quickly. Thus, it is essential to detect fault in its early stages. One of the methods of detection is selection of the optimal frequency band (OFB) for the filter design. One can find such filter characteristic through statistical approach or iterative or adaptive methods. Authors in their research propose to use time-frequency decomposition via STFT as it is one of the quickest algorithms to apply to the data. However, due to the signals structure, there will be high energy in the lower frequency band. Thus, it is reasonable to perform normalization of the absolute value of STFT matrix. Knowledge of the signals structure allows us to distinct three main different signal components. First component is a noise, usually Gaussian, second - impulsive behavior related to the fault and last one - accidental high energy impacts which disturb performance of most of the algorithms. Therefore, authors propose to model each of the sub-signals of the decomposed signal with probability density functions (PDF). Different components will give different PDF. In the final step, authors have proposed to use k-means clustering algorithms to distinguish between different structure of the frequency bands and select optimal one for the filter characteristic design.

Acoustic Emission (AE) measurement has been long established as a sensitive tool for detecting damage and failure in engineering structures, where sensors are used to detect the elastic stress waves originating from crack growth, impact damage, plastic deformation and other failure mechanisms. This paper examines the sensitivity of AE to mixed lubrication conditions, in order to evaluate the technique for monitoring heavily loaded power transmission gear systems where roughness scale fatigue phenomena such as micro-pitting are a problem. Experiments were conducted using a power-recirculating twin disc rig designed to investigate elastohydrodynamic contacts. Speed and temperature were varied in order to instigate a range of lubrication conditions from full film to heavily mixed lubrication. The AE was found to be precisely dependent on the level of asperity contact and a general relationship between AE and the specific film thickness was determined for these results.

The paper analyzes the measurements of the soil compaction factor by vibroacoustic using a dynamic plate. The principle of operation consists in triggering a short-term force impulse, caused by the impact of the weight falling from a certain height. Then, using the accelerometer sensors, the amplitude values of the ground response to the given external extortion were recorded. The analysis of the signal using the fast Fourier transform (FFT) algorithm was aimed at its distribution into harmonic components. A correlation between the number of resonance frequencies occurring in the signal and the compaction factor of the considered soil was observed. For an loose soil there is one resonant frequency and with the change of the soil bulk density, i.e. an increase in the compaction factor, higher order components appear. Using this relationship, it was possible to develop an effective method for assessing the compaction factor of the considered ground material. In the paper three types of experiments were performed. The first type of research was to force the energy impulse with the sensors found on the measuring plate. In the second study, the signal generated by the rammer was read and in the last one the signal was recorded directly on the rammer.