No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
A novel neural network with variable-activation-function for analytical redundancy of variable cycle engine sensors
ResearchGate has not been able to resolve any citations for this publication.
The variable cycle engine selects the appropriate working mode through mode switch to meet the requirements of low specific fuel consumption and high unit thrust. When the variable cycle engine switches between different modes, it should ensure the smooth transition of working modes and obtain the expected performance at the same time. Consequently, a general control schedule design method for mode switch is investigated in this paper. The feasible switching domains of single-bypass working mode and double-bypass working modes are obtained based on multifurcating tree traversal, and the switching point in the feasible switching domain is determined by solving the collaboration-game optimization problem based on sequential quadratic programming (SQP). Finally, the schedule of control variables is designed to realize smooth and fast switching between different working modes of variable cycle engine. The simulations show that based on the general control schedule of mode switch proposed in this paper, smooth transition of thrust and rotor speed can be achieved under the premise of safe engine operation, with small fluctuations and a short switching time, which is about 2.5 s.
Reducing the total mission time is essential in wildlife surveys owing to the dynamic movement of animals throughout their migrating environment and potentially extreme changes in weather. This paper proposed a multi-UAV path planning method for counting various flora and fauna populations, which can fully use the UAVs’ limited flight time to cover large areas. Unlike the current complete coverage path planning methods, based on sweep and polygon, our work encoded the path planning problem as the satisfiability modulo theory using a one-hot encoding scheme. Each instance generated a set of feasible paths at each iteration and recovered the set of shortest paths after sufficient time. We also flexibly optimized the paths based on the number of UAVs, endurance and camera parameters. We implemented the planning algorithm with four UAVs to conduct multiple photographic aerial wildlife surveys in areas around Zonag Lake, the birthplace of Tibetan antelope. Over 6 square kilometers was surveyed in about 2 h. In contrast, previous human-piloted single-drone surveys of the same area required over 4 days to complete. A generic few-shot detector that can perform effective counting without training on the target object is utilized in this paper, which can achieve an accuracy of over 97%.
The safety and reliability of the measuring elements of an aero-engine are important preconditions of the stable operation of the engine control system. The number of control parameters of a variable cycle engine increases by 20%–40% compared to traditional engines. Therefore, it is important to conduct study on the analytical redundancy, design fault diagnosis and isolation of the sensors, as well as the signal reconstruction system, so as to increase the ratability and fault-tolerant capability of the variable cycle engine control system. The analytical redundancy method relies on the accuracy of the mathematical model of the engine. During the service cycle of the engine, it is inevitable that the engine performance will deteriorate, resulting in a mismatch with the model. In this paper, the adaptive model of the variable cycle engine is built with a Kalman filter. Based on this, the strategy of analytical redundancy logic is built and the dynamic adaptive calculation of the threshold is introduced. Simulation results reflect that this method can effectively increase the reliability of sensor fault diagnosis and the accuracy of the analytical redundancy when there is performance degradation of the variable cycle engine.
The effective selection of Variable Cycle Engine (VCE) parameters plays a key role in achieving low specific fuel consumption (SFC) of fighters. However, the selection of VCE parameters is a continuous multimodal issue involving substantial local optima, so that most swarm intelligence (SI) algorithms are easily trapped into local optimal solutions, and cannot obtain satisfactory performance. To address this problem, an improved moth flame optimization algorithm with adaptive Lévy-Flight perturbations (ALFMFO) is proposed. In ALFMFO, the current population aggregation status can be accurately judged based on the difference in fitness variance between two successive moth generations. According to the population aggregation status, the Lévy-Flight disturbance strategy can adaptively adjust the perturbation probability to enhance the ability of ALFMFO to escape from local optimal solutions and realize the minimum SFC optimization of VCE. Experimental results suggest that ALFMFO is effective and superior to other compared SI algorithms in terms of accuracy and robustness.
This paper investigated the problem of fault estimation and fault-tolerant control (FTC) against sensor faults for aircraft engines. By applying a second order sliding mode observer (SOSMO) to the engine on-board model, estimations of the system states and sensor faults could be obtained simultaneously, and the result of state estimation was unaffected when using the reduced-order sliding mode system. This result gave rise to the idea to use the estimated states instead of physical sensor signal in the engine close-loop feedback control. Unlike those using passive FTC concepts, the tradeoff between control performance and robustness was inherently unnecessary. Meanwhile, compared to active FTC approaches, because any classical state/output feedback method can be directly applied to the proposed scheme without any controller reconfiguration, extra undesired dynamic responses caused by parameter reconfiguring were avoided. In this paper, the proposed FTC scheme was tested on the nonlinear model of a civil aircraft turbofan engine, and numerical simulation results showed satisfactory sensor FTC performance.
Analytical redundancy technique is of great importance to guarantee the reliability and safety of aircraft engine system. In this paper, a machine learning based aeroengine sensor analytical redundancy technique is developed and verified through hardware-in-the-loop (HIL) simulation. The modified online sequential extreme learning machine, selective updating regularized online sequential extreme learning machine (SROS-ELM), is employed to train the model online and estimate sensor measurements. It selectively updates the output weights of neural networks according to the prediction accuracy and the norm of output weight vector, tackles the problems of singularity and ill-posedness by regularization, and adopts a dual activation function in the hidden nodes combing neural and wavelet theory to enhance prediction capability. The experimental results verify the good generalization performance of SROS-ELM and show that the developed analytical redundancy technique for aeroengine sensor fault diagnosis based on SROS-ELM is effective and feasible.
Multi-sensor systems are very powerful in the complex environments. The cointegration theory and the vector error correction model, the statistic methods which widely applied in economic analysis, are utilized to create a fitting model for homogeneous sensors measurements. An algorithm is applied to implement the model for error correction, in which the signal of any sensor can be estimated from those of others. The model divides a signal series into two parts, the training part and the estimated part. By comparing the estimated part with the actual one, the proposed method can identify a sensor with possible faults and repair its signal. With a small amount of training data, the right parameters for the model in real time could be found by the algorithm. When applied in data analysis for aero engine testing, the model works well. Therefore, it is not only an effective method to detect any sensor failure or abnormality, but also a useful approach to correct possible errors.
The hard support vector regression attracts little attention owing to the overfitting phenomenon. Recently, a fast offline method has been proposed to approximately train the hard support vector regression with the generation performance comparable to the soft support vector regression. Based on this achievement, this article advances a fast online approximation called the hard support vector regression (FOAHSVR for short). By adopting the greedy stagewise and iterative strategies, it is capable of online estimating parameters of complicated systems. In order to verify the effectiveness of the FOAHSVR, an FOAHSVR-based analytical redundancy for aeroengines is developed. Experiments on the sensor failure and drift evidence the viability and feasibility of the analytical redundancy for aeroengines together with its base—FOAHSVR. In addition, the FOAHSVR is anticipated to find applications in other scientific-technical fields.
Kalman filters are often used to estimate the state variables of a dynamic system. However, in the application of Kalman filters some known signal information is often either ignored or dealt with heuristically. For instance, state variable constraints (which may be based on physical considerations) are often neglected because they do not fit easily into the structure of the Kalman filter. This article develops an analytic method of incorporating state variable inequality constraints in the Kalman filter. The resultant filter truncates the probability density function (PDF) of the Kalman filter estimate at the known constraints and then computes the constrained filter estimate as the mean of the truncated PDF. The incorporation of state variable constraints increases the computational effort of the filter but also improves its estimation accuracy. The improvement is demonstrated via simulation results obtained from a turbofan engine model. It is also shown that the truncated Kalman filter may provide a more accurate way of incorporating inequality constraints than other constrained filters (e.g. the projection approach to constrained filtering).
Performance forecasting of aeroengines is crucial for achieving better operational efficiency, ensuring safety, reducing costs, and minimizing environmental impact in the aviation industry. It enables engineers and researchers to make informed decisions, leading to advancements in technology and the overall evolution of aviation. This study is focused on the effects of flight conditions on the performance of a turbojet, which in consequence affects the environmental aspect of operation. By investigating the relationships between aeroengine efficiency and performance indicators such as thrust, shaft speed, and exhaust gas temperature (EGT), and flight characteristics expressed in terms of environmental and operational conditions, the study seeks to elucidate these connections. The article's significance lies in its successful application of Long Short-Term Memory (LSTM) networks to predict thrust, shaft speed, and EGT variations in turbojet engines under varying flight conditions. Experimental data from a turbojet test bench is processed with deep learning, specifically LSTM recurrent neural networks that are developed based on Matrix Laboratory (MATLAB). The model inputs are free stream air speed, compressor inlet pressure, combustor inlet temperature, combustor inlet pressure, turbine inlet temperature, turbine inlet pressure, nozzle inlet pressure and fuel flow, and the outputs are thrust, shaft speed and EGT. Predicted thrust closely aligns with actual thrust values, though with minor discrepancies. Shaft speed predictions exhibit a similar trend, while EGT predictions showcase a comparable pattern with slight variations. Despite the prediction errors, a thorough evaluation of median values, box plots, and probability density functions confirms that the models effectively capture available information, though discrepancies may arise from measurement inaccuracies and initial engine conditions. These results show that it is possible to accurately predict turbojet performance using LSTM recurrent neural network. This research paves the way for enhanced aeroengine performance prediction, particularly in scenarios requiring off-board or high-performance applications.
The integrated design of a variable cycle engine (VCE) and an aircraft thermal management system (TMS) is investigated. The integrated system is designed using the multiple design point approach in order to achieve required performance metrics at points other than the cycle design condition. The VCE architecture is a three stream design where the third stream is split off after the fan, exhausting through a separate third-stream nozzle. The primary air stream passes through a low-pressure compressor before splitting into an inner bypass stream and a core stream. The inner streams mix aft of the low-pressure turbine and exhaust through a core nozzle. The variable cycle engine utilizes variable compressor inlet guide vanes, a variable area bypass injector at the core stream mixing plane, and variable throats in the two exhaust nozzles. The TMS architecture is an air cycle system using air bled from the high-pressure compressor. The effect of integrating the TMS into the engine design loop is investigated. A comparison is made to prior studies where the same TMS architecture was connected to a low bypass ratio turbofan engine. The comparison shows that the variable cycle engine is able to improve heat dissipation capability versus a ram air cooled system, while eliminating the airframe integration impact that comes with a separate ram-air stream. Lastly, the impact of modulating the variable geometry features on overall cooling capability is investigated. Results are presented for individual operating points as well as at the aircraft mission level.
With the rapid development of the aviation industry, videoscope inspection of aeroengines has become crucial for ensuring aircraft flight safety. Recently, deep learning, particularly convolutional neural networks (CNNs), have shown remarkable efficacy in videoscope inspection tasks. However, these methods usually require large-scale training labels with accurate annotation. In videoscope images, defects are often present at the micrometer level and require manual labeling by professional inspection personnel, posing further challenges to model training. To address these issues, Weakly-Supervised Deep Level Set (WDLS), a fully automatic framework for aeroengine defect segmentation, was proposed in this work. WDLS employs a multi-branch structure and an iterative learning strategy to combine a high-performance CNN architecture with level set evolution. First, the Similarity-Guided Region (SGR) detector was designed to generate class-specific segmentation proposals and initialize the level set function. Second, the Adaptive Local Parameter Prediction (ALPP) algorithm was proposed to integrate local priors and constraints into the energy function and optimize the iterative process. Finally, a self-supervised objective function named Convexified Level Set (CLS) was introduced to represent an improved level set formulation and obtain the final segmentation. The proposed framework is continuously differentiable and unified. Thus, it can be seamlessly embedded in any CNN without postprocessing. Furthermore, the method was evaluated on a new aeroengine dataset Turbo19 and a benchmark dataset. Experimental results demonstrated that the proposed framework meets real-time requirements and achieves better performance than state-of-the-art methods.
Sensors are the primary information source of the aeroengine control system, their measurement accuracy is closely related to whether the engine can operate safely and efficiently. Aiming at the direct thrust control system of aeroengines, this paper proposes an intelligent prediction algorithm combining feedforward and recurrent networks and a fault-tolerant control strategy combining analytical redundancy and controller switching. First, the cycle reservoir with regular jumps is introduced into the online sequential extreme learning machine. The CR-OSELM is developed, which maps the inputs from the low-dimensional space to the high-dimensional space. Then the CR-OSELM is adopted to build the sensor measurement prediction module, and identify and reconstruct the sensor bias and drift fault. Furthermore, in allusion to the rotor speed sensor and temperature/pressure sensor faults, analytic redundancy and controller switching strategies are designed respectively, realizing the fault-tolerant control of thrust feedback. The main contribution of this paper is to come up with the CR-OSELM algorithm with the ability to save historical input information, which overcomes the limitations of traditional feedforward neural networks in identifying time series. And the compound fault-tolerant control scheme is put forward based on the fault characteristics and their impact on the direct thrust control system, which can deal with different types of sensor faults. Simulations are performed on some benchmark datasets and a turbofan engine model to investigate the time series prediction accuracy, sensor fault diagnosis, and thrust fault-tolerant control effect. The results show that the system can detect various faults rapidly, realize the analytic redundancy with high accuracy, reduce the influence of sensor faults and restore the engine performance.
Airbreathing aero-engines are regarded as excellent propulsion devices from ground takeoff to hypersonic flight, and require control systems to ensure their efficient and safe operation. Therefore, the present paper aims to provide a summary report of recent research progress on airbreathing aero-engine control to help researchers working on this topic. First, five control problems of airbreathing aero-engines are classified: uncertainty problem, multi-objective and multivariable control, fault-tolerant control, distributed control system, and airframe/propulsion integrated control system. Subsequently, the research progress of aircraft gas turbine engine modelling, linear control, nonlinear control, and intelligent control is reviewed, and the advantages and disadvantages of various advanced control algorithms in aircraft gas turbine engines is discussed. Third, several typical hypersonic flight tests are investigated, and the modelling and control issues of dual-mode scramjet are examined. Fourth, modelling, mode transition control and thrust pinch control for turbine-based combined cycle engines are introduced. Followed, significant hypersonic airframe/propulsion integrated system control is analysed. Finally, the study provides specific control research topics that require attention on airbreathing aero-engines.
Intelligent data-driven fault diagnosis based on conventional machine learning techniques has been extensively studied in recent years. However, these methods often assumed that the data used for training and testing are drawn from the identical distribution, which is impractical in real application. Such idealized hypothesis may confine these promising data-driven techniques to well-designed experimental environments rather than actually putting them into real-world applications. In practice, the distribution discrepancies between source domain and target domain will degrade the diagnostic performance. To this end, this work introduces a transfer learning based extreme learning machine to align the distribution discrepancies of the data collected from a turbofan engine, which is rarely studied in the fault diagnosis for aero-engine. The proposed method is capable of learning the transferable cross domain features while preserving the properties and structures of source domain as much as possible. Meanwhile, the marginal distribution and conditional distribution discrepancies are matched. Through these transfer data representations, a relatively high diagnostic accuracy is guaranteed. Finally, extensive experiments have been performed on gas path fault diagnosis of turbofan engine, including hybrid transfer cases and complete transfer cases, to verify the effectiveness and feasibility of the proposed method.
In this paper, variable-weights neural network is proposed to construct variable cycle engine’s analytical redundancy, when all control variables and environmental variables are changing simultaneously, also accompanied with the whole engine’s degradation. In another word, variable-weights neural network is proposed to solve a multi-variable, strongly nonlinear, dynamic and time-varying problem. By making weights a function of input, variable-weights neural network’s nonlinear expressive capability is increased dramatically at the same time of decreasing the number of parameters. Results demonstrate that although variable-weights neural network and other algorithms excel in different analytical redundancy tasks, due to the fact that variable-weights neural network’s calculation time is less than one fifth of other algorithms, the calculation efficiency of variable-weights neural network is five times more than other algorithms. Variable-weights neural network not only provides critical variable-weights thought that could be applied in almost all machine learning methods, but also blazes a new way to apply deep learning methods to aeroengines.
Gas-path model (GPM) plays a key role in the application of sensor fault tolerant control for the gas turbine engine (GTE). However, the modeling accuracy of traditional physics-based GPM is restricted by modeling inaccuracies and measurement uncertainty of sensors, etc., making fault-tolerant control of sensors difficult. In this paper, an improved input and output self-tuning hybrid modeling (IOSTHM) method is proposed for improving the GPM modeling accuracy of a dual shaft turbofan GTE. The proposed IOSTHM consists of an input self-tuning model (ISTM) and an output self-tuning model (OSTM). In the framework of ISTM, a fundamental traditional physics-based model (PBM) is constructed by using the component level modeling method firstly. Based on the PBM and the actual measurement of rotor speed of low-pressure shaft (N1) which has high measurement accuracy, the ISTM is conducted by tuning the measurement of fuel flow in real-time according to the deviation of N1 between engine and PBM, aiming to deal with the measurement uncertainty of fuel flow. Based on the ISTM, the IOSTHM is constructed by combining with an OSTM which is constituted by a bank of residual learning models to reduce the modeling inaccuracies further. The effectiveness of the proposed IOSTHM is evaluated with the verification of simulated flight data and actual ground test data. Both flight and ground verification results reveal that the proposed hybrid model IOSTHM achieves the best modeling performance when compared with other models such as traditional PBM, ISTM, and OSTM, it shows the superiority of this methodology.
Data-driven fault diagnosis for aircraft engine has recently been widely studied, most of which utilized traditional machine learning tools. However, these approaches often assumed that the data used for training and testing come from the identical distribution, which is unpractical in practice, because even for the same type of engine, there may be some discrepancies in the data collected between different engine individuals. This idealized assumption may confine these promising data-driven technologies to well-design experimental environments instead of applying them to practical applications. Meanwhile, the research on data-driven engine fault diagnosis methods based on transfer learning is relatively limited. To overcome this issue, this paper proposes a methodology using unilateral alignment strategy to address the issue of cross domain engine gas path fault diagnosis, where the training data extracted from source domain and the testing data extracted from target domain are assumed to come from different distributions. More importantly, the proposed methodology is capable of maintaining the inter-class relationship of source domain, instead of aligning the discrepancies for source and target domain by mapping both of their feature spaces into an unknown intermediate space, because the label space between two domains may be different, and the label set of target domain might be only a subset of that of source domain. Meanwhile, both the marginal and conditional distribution discrepancies are considered. Finally, the proposed approaches are evaluated by extensive experiments on engine cross domain fault diagnosis, including hybrid transfer experiments, complete transfer experiments and missing class experiments, and the overall results demonstrate the feasibility of the proposed methodology.
Various Kalman filter approaches have been presented for performance estimation of aircraft engines provided that sensor measurement numbers are sufficient and all of them are available. However, it is difficult to collect all physical parameters along the gas path since the complex structure limits sensor installation, especially around the high-pressure turbine. The contribution of this article is to create a virtual sensor in the hot-section based on the component physical characteristics and aerothermodynamic theory and develop a dual hypothesis test strategy combined with a state estimator to track abrupt degradation of the engine component. A novel couple algorithm of the unscented Kalman filter using a virtual sensor is developed that has three primary advantages: (i) The performance degradation of four rotatory components can be estimated without sensor P 43 . (ii) The accuracy of state estimation using the virtual sensor is equivalent to that of sensor P 43 , and it adapts to the abrupt change of engine operating condition in the whole flight envelope. (iii) If the sensor can be installed in the future, it can be engaged to its analytical redundancy. Simulations are carried out on typical abrupt degradation datasets of aircraft engines, which demonstrate the superiority of anomaly detection of gas path component performance using virtual sensor measurements, especially hot-section components.
In order to study the performance seeking control of double bypass variable cycle engine, a Feasible Sequential Quadratic Programming (FSQP) algorithm with global and local superlinear convergence characteristics is adopted. The engine model established a backward infrared radiation intensity prediction method for exhaust system, which could calculate the infrared radiation intensity of the high temperature components and nozzle stream in the system. The infrared radiation absorption effect of the nozzle stream on the high temperature components is considered in the method. According to the above model, a minimum infrared characteristic performance seeking control based on infrared prediction model is proposed. The optimal control simulations of the maximum thrust, minimum specific fuel consumption and minimum infrared characteristic mode on the double bypass variable cycle engine are carried out. Finally, the simulation results show that the algorithm could obtain the optimal operating conditions of double bypass variable cycle engine in different working conditions, which achieves the steady-state performance seeking control of the engine.
An onboard aero-engine model with low computation burden and high accuracy is urgent for online parameter prediction in some model-based control applications. However, current engine models, including component level models, linear models and data-driven models cannot meet the demands of model-based control and diagnostics well. Thus, a novel hybrid adaptive model is proposed, which consists of a component level model, an improved incremental linearized Kalman filter and a state space model. The improved linearized Kalman filter is used to adapt the component level model to the performance degradations of the engine, and the state space model is implemented for online prediction with a low computation burden. Both the linearized Kalman filter and the state space model are obtained by linearizing the adaptive component level model in real-time. Thus, an exact derivative calculation method is extended and an improved transient calculation and linearization joint algorithm is proposed to achieve real-time property. Experiments are conducted to demonstrate the effectiveness and real-time property of proposed model in terms of estimation accuracy, prediction accuracy and time consumption. The results show that the proposed adaptive model can not only estimate the health parameters and track the performance degradations well but also predict interesting outputs over next several sample periods accurately and quickly. In addition, the proposed adaptive model can complete the estimation and prediction task in a sample period (25 ms) in a micro-controller unit, which suggests the great real-time property of the proposed model.
By establishing a three dimensional model of a double bypass variable cycle compression system, the flow patterns and matching characteristics of each working component during mode transition are investigated using numerical simulation. Results show that transition from the single bypass mode to double bypass mode by opening the mode selector valve (MSV) alone would increase the fan operation point while decreasing that of the core driven fan stage (CDFS) and the high pressure compressor (HPC). It would also incur outer bypass flow recirculation and bring radial inflow distortions to the CDFS. Deviations of the fan aerodynamic performance lie mainly in its aft stage, while the HPC first stage undertakes most of the inlet distortion. Reducing the bypass backpressure during mode transition is an effective way to alleviate the outer bypass flow recirculation but would further choke the CDFS. Closing the forward variable area bypass injector (FVABI) could raise the CDFS matching point so as to improve the stator performance without influencing the outer bypass ratio. It is recommended to decrease the bypass backpressure and close FVABI simultaneously in the real transition process.
For next generation aircraft, Variable Cycle Engine (VCE) is a candidate to fulfil the multi-mission requirements of flight. This new concept is promising to complete deficiencies of conventional low by-pass mixed turbofan engines because the VCE model incorporates different thermodynamic cycles (turbojet and turbofan) on the same system. Namely, having single by-pass (SBM) and double by-pass modes (DBM), these engines can modulate by-pass ratio, ranging from 0.3 to 1.8 during flight with respect to operating condition requirements. In this study, parametric equations for components of the developed VCE model are determined by means of cycle equations of a mixed-flow turbofan with afterburner. The first aim is to compare energetic performance results of the developed VCE with the results of F100-PW-100 (as called F100) used on F-15 or F-16 for SBM and DBM. Secondly, the main goal of this study is to perform second law analysis for the VCE model and compare it with the results of the F100 engine for off-design, at which flight altitude (9–20 km) and Mach number (0.3–1.9) changes. For this aim, thermodynamic data (pressure, temperature, mass flow etc.) of this engine main components are obtained from GasTurb program for above stated flight conditions. On the other hand, exergy analysis along with five sustainability parameters is carried out at SBM and DBM as comparing with those of F100 engine. Off-design analysis shows that the specific fuel consumption (SFC) varies from 19.97 g/kN.s to 28.25 g/kN.s for VCE and from 23.91 g/kN.s to 31.14 g/kN.s for F100 at DBM. Moreover, at SBM, SFC changes from 52.68 g/kN.s to 59.19 g/kN.s for VCE and from 58.04 g/kN.s to 63.57 g/kN.s for F100 throughout the entire flight conditions. As for exergy-based sustainability analysis, exergetic sustainability index (ESI) of the whole VCE is found to be in the range of 0.14 and 0.51 for DBM and to be in the range of 0.26 and 0.38 for SBM. For F100 engine, ESI is calculated to be in the range of 0.10 and 0.39 for DBM and to be in the range of 0.15 and 0.34 for SBM. It is thought that this study can help in understanding the effect of the working cycle on engine performance and sustainability parameters.
Prognostics technique aims to accurately estimate the Remaining Useful Life (RUL) of a subsystem or a component using sensor data, which has many real world applications. However, many of the existing algorithms are based on linear models, which cannot capture the complex relationship between the sensor data and RUL. Although Multilayer Perceptron (MLP) has been applied to predict RUL, it cannot learn salient features automatically, because of its network structure. A novel deep Convolutional Neural Network (CNN) based regression approach for estimating the RUL is proposed in this paper. Although CNN has been applied on tasks such as computer vision, natural language processing, speech recognition etc., this is the first attempt to adopt CNN for RUL estimation in prognostics. Different from the existing CNN structure for computer vision, the convolution and pooling filters in our approach are applied along the temporal dimension over the multi-channel sensor data to incorporate automated feature learning from raw sensor signals in a systematic way. Through the deep architecture, the learned features are the higher-level abstract representation of low-level raw sensor signals. Furthermore, feature learning and RUL estimation are mutually enhanced by the supervised feedback. We compared with several state-of-the-art algorithms on two publicly available data sets to evaluate the effectiveness of this proposed approach. The encouraging results demonstrate that our proposed deep convolutional neural network based regression approach for RUL estimation is not only more efficient but also more accurate.
Deeper neural networks are more difficult to train. We present a residual
learning framework to ease the training of networks that are substantially
deeper than those used previously. We explicitly reformulate the layers as
learning residual functions with reference to the layer inputs, instead of
learning unreferenced functions. We provide comprehensive empirical evidence
showing that these residual networks are easier to optimize, and can gain
accuracy from considerably increased depth. On the ImageNet dataset we evaluate
residual nets with a depth of up to 152 layers---8x deeper than VGG nets but
still having lower complexity. An ensemble of these residual nets achieves
3.57% error on the ImageNet test set. This result won the 1st place on the
ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100
and 1000 layers.
The depth of representations is of central importance for many visual
recognition tasks. Solely due to our extremely deep representations, we obtain
a 28% relative improvement on the COCO object detection dataset. Deep residual
nets are foundations of our submissions to ILSVRC & COCO 2015 competitions,
where we also won the 1st places on the tasks of ImageNet detection, ImageNet
localization, COCO detection, and COCO segmentation.
Deep and recurrent neural networks (DNNs and RNNs respectively) are powerful models that were considered to be almost impossible to train using stochastic gradient descent with momentum. In this paper, we show that when stochastic gradient descent with momentum uses a well-designed random initialization and a particular type of slowly increasing schedule for the momentum parameter, it can train both DNNs and RNNs (on datasets with long-term dependencies) to levels of performance that were previously achievable only with Hessian-Free optimization. We find that both the initialization and the momentum are crucial since poorly initialized networks cannot be trained with momentum and well-initialized networks perform markedly worse when the momentum is absent or poorly tuned. Our success training these models suggests that previous attempts to train deep and recurrent neural networks from random initializations have likely failed due to poor initialization schemes. Furthermore, carefully tuned momentum methods suffice for dealing with the curvature issues in deep and recurrent network training objectives without the need for sophisticated second-order methods.
With the continuous increase in complexity and expense of industrial systems, there is less tolerance for performance degradation, productivity decrease, and safety hazards, which greatly necessitates to detect and identify any kinds of potential abnormalities and faults as early as possible and implement real-time fault-tolerant operation for minimizing performance degradation and avoiding dangerous situations. During the last four decades, fruitful results have been reported about fault diagnosis and fault-tolerant control methods and their applications in a variety of engineering systems. The three-part survey paper aims to give a comprehensive review of real-time fault diagnosis and fault-tolerant control, with particular attention on the results reported in the last decade. In this paper, fault diagnosis approaches and their applications are comprehensively reviewed from model- and signal-based perspectives, respectively.
Engine diagnostic practices are as old as the gas turbine itself. Monitoring and analysis methods have progressed in sophistication over the past 6 decades as the gas turbine evolved in form and complexity. While much of what will be presented here may equally apply to both stationary power plants and aero-engines, the emphasis will be on aero propulsion.
Beginning with primarily empirical methods centering around monitoring the mechanical integrity of the machine, the evolution of engine diagnostics has benefited from advances in sensing, electronic monitoring devices, increased fidelity in engine modeling and analytical methods. The primary motivation in this development is, not surprisingly, cost. The ever increasing cost of fuel, engine prices, spare parts, maintenance and overhaul, all contribute to the cost of an engine over its entire life cycle. Diagnostics can be viewed as a means to mitigate risk in decisions that impact operational integrity. This can have a profound impact on safety, such as In-Flight Shut Downs (IFSD) for aero applications, (outages for land based applications) and economic impact caused by Unscheduled Engine Removals (UERs), part life, maintenance and overhaul and the overall logistics of maintaining an aircraft fleet or power generation plants.
This paper will review some of the methods used in the preceding decades to address these issues, their evolution to current practices and some future trends. While several different monitoring and diagnostic systems will be addressed, the emphasis in this paper will be centered on those dealing with the aero-thermodynamic performance of the engine.
This paper provides an overview of the controls and diagnostics technologies, that are seen as critical for more intelligent gas turbine engines (GTE), with an emphasis on the sensor and actuator technologies that need to be developed for the controls and diagnostics implementation. The objective of the paper is to help the “Customers” of advanced technologies, defense acquisition and aerospace research agencies, understand the state-of-the-art of intelligent GTE technologies, and help the “Researchers” and “Technology Developers” for GTE sensors and actuators identify what technologies need to be developed to enable the “Intelligent GTE” concepts and focus their research efforts on closing the technology gap. To keep the effort manageable, the focus of the paper is on “On-Board Intelligence” to enable safe and efficient operation of the engine over its life time, with an emphasis on gas path performance.
Fault-tolerant control scheme for the sensor Fault in the acceleration process of variable cycle engine
- Li
Model-based fault tolerant control
- A Kumar
- D Viassolo
Research on Adaptive Model-Based Fault Diagnosis and Fault Tolerant Control of Variable Cycle Engine
- S C Wang