Real Time Bearing Defect Classification Using Time Domain Analysis and Deep Learning Algorithms

<div class="section abstract"><div class="htmlview paragraph">TRIADE is a European Union project that focuses on the development of technological building blocks for Structure Health Monitoring (SHM) sensing devices in aeronautics. It is funded under the 7th framework program. In terms of objectives, the TRIADE project focuses on providing these technological building blocks and fully integrated prototypes in order to achieve power generation, power conservation, energy management and embedded powerful intelligence for data processing and storage for SHM sensing devices.</div><div class="htmlview paragraph">The principal technological building blocks that the TRIADE project will provide are: - A low profile battery with high energy density which will be able to function in a harsh environment, - An energy harvester from vibration and electromagnetic RF, - Ultra low power sensors which will be designed in SOI technology, and - A neural network for data recording and damage assessment.</div><div class="htmlview paragraph">The smart tag will be the size of a credit card (dimension 85x55x4 mm) and will support remote sensors. It will be embedded under the last composite layer or attached on the structure of the aircraft. It will be autonomous as all the power needed will be provided by the onboard rechargeable battery. It is wireless and the data stored will be transferred to the base station by using the RFID communication protocols through an RFID reader. It will record temperature, relative humidity, pressure, strain, vibration and acoustic emissions whenever the flight domain is overrun.</div><div class="htmlview paragraph">The structure health monitoring capability that the TRIADE project offers is planned to be applied in 2 application domains: on helicopters with Eurocopter and PZL Swidnik as end-users, and on airplanes with EADS Deutschland GmbH and Dassault Aviation as end-users. The different requirements from each end-user have been unified on the smart tag with minimal compromises to accommodate of.</div></div>

<div class="section abstract"><div class="htmlview paragraph">The bolt is one of the most standardized and most commonly used machine elements. On the other hand, since the mechanics of highly stressed bolted joints and the thread fatigue are complex issues, the design and evaluation of such joints is frequently carried out with major simplifications and assumptions, leading to either over-engineered solutions or to premature failures of the prototypes.</div><div class="htmlview paragraph">The simulation techniques and the computing power that are now available theoretically allow a precise evaluation of the fatigue safety in the most heavily loaded areas through the application of very fine FE models of the thread regions. However, due to the modeling and calculation effort, this is still only acceptable in practice for structures with a limited number of bolts. Furthermore, uncertainties related to material specification including the behavior in the plastic range as well as difficulties in considering thread manufacturing tolerances or residual torsional stresses from the assembly state, makes this approach difficult to apply widely in the engineering practice.</div><div class="htmlview paragraph">An interesting alternative method is a combination of the FE analysis of highly stressed bolted joints with historically well-proven analytical considerations, as proposed in the initially released second part of the VDI 2230 standard. In this way, most of the aspects that are difficult to be effectively simulated by FEA are considered in the VDI 2230 limits defined by hardware tests, while in parallel the unique behavior of the joint including the tension, bending and shearing of the bolt can be precisely simulated.</div><div class="htmlview paragraph">The effective application of the above approach is shown in this paper using the example of a straight-split, inclined-split and a marine type connecting rod. The differences in joint behavior and thus the design requirements as well as the limitations and benefits of the applied evaluation technique are discussed in detail.</div></div>

<div class="section abstract"><div class="htmlview paragraph">According to the advance of engine control development, demands for direct sensing of physical quantity have been growing. Regarding pressure sensing, key properties for direct sensing are robustness against high temperature and pressure, and response time in addition to accuracy. In this work, a pressure sensor module with these key properties was developed. First of all, a piezoelectric device was selected as a suitable sensing principle for the required properties because of its thermally stable piezoelectric effect and potential for simple installation structure. Regarding robustness against temperature, the sensor module was designed to form thermal isolation layer with outer housing which is optimized according to its application. Regarding robustness against pressure and response time, breakage of the piezoelectric element is the main technical issue. Based on FEM analysis, a bounce of piezoelectric element in the sensor structure was proven to be the main cause of the element breakage. To prevent piezoelectric element bounce, a pre-load design method was developed as a countermeasure. Regarding high accuracy, the piezoelectric sensor module was modeled as an equivalent spring model. Based on this model, a degradation mechanism of the sensor accuracy was established by converting the sensor component degradation to spring constant changes. Then the contact state between each component was proven to be the important factor for sensor accuracy. Finally, a cylinder pressure sensor was designed as an application example of this pressure sensor module. As a result of engine bench tests, the potential of the module to measure high frequency pressure under high temperature was confirmed.</div></div>

<div class="htmlview paragraph">The effects of engine speed and load on nitric oxide emission were investigated with a single-cylinder engine operated at conditions representative of those found in the 7-mode, 7-cycle Federal Emissions Test Cycle. The fuel-air mixture was constant at 83% of stoichiometric and the spark settings were optimized.</div> <div class="htmlview paragraph">Nitric oxide emission increased as engine speed and load were increased. The effects were interdependent, in that, the effect of speed was greatest at low load and the effect of load was greatest at low speed. In addition, both speed and load effects were larger when valve overlap was higher.</div> <div class="htmlview paragraph">Engine speed and load affected exhaust nitric oxide concentration primarily via influences on charge dilution. This finding explains the interdependency of the speed and load effects, and also why they depended on valve overlap.</div>

<div class="section abstract"><div class="htmlview paragraph">Addressing climate issues is a key aspect of good global governance today. A key aspect of managing the threats caused to the environment around is to ensure a sustainable transportation system so that humans exist in peace with nature. According to sources, in 2020 alone, cars accounted for approximately 23% of global CO<sub>2</sub> emissions. In addition, they also emit dangerous pollutants thus damaging the ecosystem.</div><div class="htmlview paragraph">To keep pollutants in check there are emission level testing strategies in place in each country. However, we can do better for a sustainable future. On one hand, the huge volume of vehicles around the world makes it an excellent choice and source for a vast emission level dataset comprising of input features as well as the target variable representing the emission band of the vehicle.</div><div class="htmlview paragraph">In addition to the big data available as mentioned above, major advancements in the machine learning algorithms are done today. The advent of algorithms such as Artificial Neural Networks (ANN) has made it possible to develop models with very high accuracy.</div><div class="htmlview paragraph">In this paper, the authors therefore propose the application of Extreme Learning Machines (a type of feedforward neural networks) to solve the pressing challenge of classifying the emission band of a vehicle which can be used by agencies to ensure that healthy vehicles operate on road at large. Extreme Learning Machines (ELM), by definition, is an excellent choice for emission band prediction task as it offers a significant reduction in training time and thereby is scalable such as to the task confronted with in this paper. Results from the metric, namely classification accuracy, are discussed at length. The highly accurate trained model, thus developed, can be used by agencies to then predict the emission band for any given vehicle scenario thus complementing the strategies already in place today for a greener earth.</div></div>

<div class="section abstract"><div class="htmlview paragraph">Industries have been increasingly adopting AI based computer vision models for automated asset defect inspection. A challenging aspect within this domain is the inspection of composite assets consisting of multiple components, each of which is an object of interest for inspection, with its own structural variations, defect types and signatures. Training vision models for such an inspection process involves numerous challenges around data acquisition such as insufficient volume, inconsistent positioning, poor quality and imbalance owing to inadequate image samples of infrequently occurring defects. Approaches to augmenting the dataset through Standard Data Augmentation (SDA) methods (image transformations such as flipping, rotation, contrast adjustment, etc.) have had limited success. When dealing with images of such composite assets, it is challenging to correct the data imbalance at the component level using image transformations as they apply to all the components within an image. Even though existing methods would aid in increasing the dataset volume, the continued data imbalance leads to a biased model whose accuracy is adversely impacted. This paper presents an approach to synthesizing composite asset images using Deep Learning, particularly focusing on organically inserting specific components with defect signatures in a composite image without compromising the integrity of the composite asset. The described approach leverages diffusion-based models to generate images, enabling granular control over the location and appearance of the synthesized components in the image. The exercise of this granular control over individual components of an image leads to the creation of a refined and balanced dataset across multiple-components and defect categories. The utility of the approach is established through experiments conducted with real-world asset images. The synthetic data is assessed through metrics that measure how well it represents the real data distribution as well as its impact on downstream asset inspection models through accuracy metrics. The paper concludes by contrasting the efficacy of the presented approach against other family of algorithms including Generative Adversarial Networks.</div></div>

<div class="htmlview paragraph">The people reports data on block program fatigue testing of a simple cantilever specimen with a stress concentration, the specimen being made of SAE 1530 steel. The various programs investigated were based on a normal distribution with a crest factor of 4 and were aimed at accelerating fatigue testing.</div> <div class="htmlview paragraph">From tests on programs with 4,8, and 16 stress levels, it was concluded that at least 8 stress loads were required to stimulate accurately the normal distribution. Tests in which various stress levels were omitted showed that stress levels S/σ< 1.75 contributed nothing to fatigue damage, and consequently they could be omitted from the program, resulting in a saving of 87% in testing time. Stress level in range 1.75 S/σ<=4 had a damaging effect, and even the one of S/σ=4 in each full program of 4000 cycles caused significant damage. This indicates the great importance of the choice of crest factor.</div>

<div class="section abstract"><div class="htmlview paragraph">Fatigue damage sensing and monitoring of any structure is a prerequisite for reliable and effective structural health diagnosis. The designed sensor has alternate slots and strips with different strain magnification factor with respect to the nominal strain at its location. The strips experience the strains which closely resemble the actual strain distribution in the critical area of the component. One of the major advantages of this sensor is that it can be placed at any convenient location, still experiencing the same fatigue damage as a critical location. It can be used on various structures from ground civilian and military vehicles to steel bridges. This can predict the remaining useful life of a component or the number of miles (for any automobile) left for the component before it needed replacement. This paper mainly describes the design aspects of this sensor following analytical and finite element analysis (FEA) approaches. Further, the test results for single ligament sensors were discussed in the context of the designed strain ratios. Overall, the role of fatigue damage (FD) sensors in structural health monitoring is presented.</div></div>

<div class="section abstract"><div class="htmlview paragraph">Aviation industry is striving to leverage the technological advancements in connectivity, computation and data analytics. Scalable and robust connectivity enables futuristic applications like smart cabins, prognostic health management (PHM) and AI/ML based analytics for effective decision making leading to flight operational efficiency, optimized maintenance planning and aircraft downtime reduction. Wireless Sensor Networks (WSN) are gaining prominence on the aircraft for providing large scale connectivity solution that are essential for implementing various health monitoring applications like Structural Health Monitoring (SHM), Prognostic Health Management (PHM), etc. and control applications like smart lighting, smart seats, smart lavatory, etc. These applications help in improving passenger experience, flight operational efficiency, optimized maintenance planning and aircraft downtime reduction. Intra Aircraft WSNs (IAWSN) used for such applications are expected to provide robust and reliable communication performance. However, IAWSNs, when deployed, must co-exist with other wireless devices and networks based on Wi-Fi and Bluetooth technologies. Sharing of the ISM frequency band (2.4 GHz) among these networks makes the co-existence problem more challenging and has a significant impact on the Quality of Service (QoS) in terms of throughput, latency and Packet Error Rate (PER) of IAWSN. In addition, they are subjected to additional electromagnetic interference from other electronic and avionic systems onboard aircraft. There are various channel management, priority-based scheduling and time-sharing techniques that are deployed currently to address the co-existence problem. However, these methods perform a trade-off among one or more operational parameters like channel bandwidth, number of nodes per channel, throughput, latency, PER, etc. of IAWSN. Code Division Multiple Access (CDMA) technique is a proven one in cellular networks for providing reliable interference free communication performance especially in large scale networks. This paper evaluates the feasibility of deploying CDMA for IEEE 802.15.4 based IAWSN onboard aircraft. A CDMA based communication schema is proposed and simulated for IAWSN operating as per IEEE 802.15.4 protocol. The communication performance of the CDMA based IAWSN is evaluated in comparison with the performance of a standard IEEE 802.15.4 protocol implementation in the presence of interference from the co-nodes of the IAWSN and Wi-Fi devices by analyzing the QoS parameters like signal to noise ratio (SNR), operational bandwidth, Bit-Error Rate (BER) and Process Gain through simulation. Simulations results are evaluated against the desired performance level for a reference use case application. Improvement opportunities are identified and areas for future research are proposed.</div></div>

<div class="htmlview paragraph">This paper will detail the development and validation of the world's first integrated smart automotive lead-acid battery tester. As today's automobiles add more electronic and electrical systems to the vehicle the need for reliable battery power is more important than ever. The need to determine the battery's functional life span has demanded a product to reliably test the battery's state of health (SOH).</div> <div class="htmlview paragraph">After two years of development backed up by over 17 years experience in testing lead acid batteries, onGUARD™ delivers the ability to accurately test and predict the batteries life expectancy.</div> <div class="htmlview paragraph">The paper details the elements behind the smart tester technology and its ability to measure the battery's voltage, temperature and conductance to accurately diagnose the health of the battery. A background of conductance testing technology and its approach to measuring the health of a battery by measuring the internal resistance of the battery is included. The benefits of conductance as a safe and accurate alternative to load testing in determining the Cold Cranking Amps of a battery are included.</div> <div class="htmlview paragraph">The mechanical configuration of onGUARD will be described, as it presently is available in two platforms. One platform is fully integrated into the battery, designed for a specific manufacturer's battery. A second is an external sensor module, which can be, applied to any lead acid battery in the 300 to 1500 CCA range. The integrated onGUARD model is designed to fit inside a battery lid connected internally to the battery's posts. The onGUARD unit can be programmed to specifically test a particular battery after a battery characterization has been established, and this process will be explained. The aftermarket onGUARD unit is a universal model to be attached to the outside case of the battery. The customer interface process of pushing and holding a pushbutton on the onGUARD unit perform measurements, and a display of test results as shown by a simple series of LEDs (A pattern of green, yellow and red LEDs indicate the batteries state of health) will be included.</div> <div class="htmlview paragraph">A full-production lead-acid SLI automotive battery will launch in October 2002 in Japan featuring an integrated onGUARD conductance test module with SOH and voltage state of charge (SOC) diagnostics. This paper provides some technical details of that platform, and as onGUARD has application in automotive, marine and battery backup power systems, a brief mention of potential benefits from the integration of additional technologies is also presented.</div>

<div class="htmlview paragraph">This research proposes an automatic measuring method for the impulse noise of the valve system in engine production line. This research is composed of the following two parts. (1) The most suitable method for indicating the impulse noise of the valve system - the representative characteristic values - is selected from the general measuring methods for impulse noise. As the result, the crest factor in the frequency band above 1kHz became optimal. (2) By detailed sensory characteristic analysis it was found that impulse noise can be heard better with increasing frequency and that there is little influence in the frequency band with the same frequency as the background noise. Thus the crest factor was obtained for each frequency, and the sensory test for the impulse noise of the valve system is deduced by this linear coupling. As the result of multiple reguression analysis, a high accuracy prediction equetion with a multiple correlation coefficient of 0.91 has been obtained.</div>

<div class="section abstract"><div class="htmlview paragraph">In order to meet increasingly stringent emission regulations, it is significance to establish a control- oriented transient NOx and PM emission prediction model and improve the accuracy and real-time performance. In this study, the prediction model of transient PM emissions based on Transformer is established. In terms of model accuracy and real-time performance, Transformer emission prediction model is compared with Multilayer perceptron (MLP) neural network and Long-Short Term Memory (LSTM) emission prediction model. The results show that the performance of Transformer transient emission prediction model is superior to other model structures, it can be used for real-time prediction.</div></div>

<div class="section abstract"><div class="htmlview paragraph">Current studies about battery pack bottom strike usually focus on one test condition individually. To study the relation between quasi-static and dynamic crush in battery pack bottom strike, the paper combined quasi-static crush result and dynamic strike preset kinetic energy value with the same displacement damage on the battery pack bottom plate and cell. First, based on the finite element model of the battery pack, the quasi-static crush is applied. Several dynamic crush tests with different initial kinetic energy sets are also introduced. Then based on the same displacement damage, the pressure in quasi-static and kinetic energy in dynamic conditions are summarized. Fitting methods including polynomial regression, support vector regression (SVR), extreme learning machine (ELM), multilayer perceptron (MLP), Gaussian process regression (GPR), and K-nearest neighbor (KNN) regression are used to study the relation between the two different test load. The result shows that they have a strong relation. Compared with the case test of GPR and KNN, polynomial regression with a degree of 4 could be the best choice to predict the dynamic load value from quasi-static results globally for bottom plate and degree of 3 for cell.</div></div>

<div class="htmlview paragraph">Artificial Neural Network(ANN) techniques are used to develop a system to detect High Impedance Faults(HIFs) in electric power distribution lines. Encouraging results were observed with a simple Multi-layer Perceptron(MLP) trained with the backpropagation learning algorithm. Although the results are not significantly better than those reported with other algorithmic approaches, ANN techniques have potential advantages over the other approaches; namely, ability to train the system easily to accommodate different feeder characteristics, ability to adapt and so become a better detector with experience and better fault tolerance. When these features are incorporated, the system is expected to perform better than existing systems.</div> <div class="htmlview paragraph">The system we developed for the current phase, the training strategies used, the tests conducted and the results obtained are discussed in this paper. Also background discussions on existing HIF detection techniques, and ANN techniques can be found in this paper. A feature extraction method based on fuzzy logic and a bin structure is discussed.</div>

<div class="section abstract"><div class="htmlview paragraph">The first cross-industry guidelines for the implementation of structural health monitoring for aerospace applications have been created as a SAE International Aerospace Recommended Practices document: SAE ARP 6461 ‘Guidelines for Implementation of Structural Health Monitoring on Fixed Wing Aircraft’ [<span class="xref">1</span>]. These guidelines have brought together manufacturers, operators / users, systems integrators, regulators, technology providers and researchers to produce information on the integration of SHM into aircraft maintenance procedures, generic requirements and advice on validation, verification and airworthiness.</div><div class="htmlview paragraph">The take-up of SHM in the aerospace industry has been slow, in part due to the lack of accepted industry practices surrounding not just the technology itself (sensors and sensor systems) but also the associated issues arising from the introduction of new methods into aircraft maintenance. The SHM community responded by forming an Industry Steering Committee on the topic which then became a Technical Group within the SAE International standards organization (G-11 SHM). The initial task of this group was to produce guidelines to address some of the issues that currently prevent adoption of SHM in commercial and military aircraft operations.</div><div class="htmlview paragraph">This paper is a summary of the key content to the guidelines and the rationale behind their development. This is the first document of its type and as such is not concerned with details of the SHM technologies themselves but instead, focuses on the key steps needed to implement the technologies within the current regulatory environment and prevailing aircraft structural design and maintenance practices.</div></div>