Search

This article presents a novel approach for assessing the effects of residual stresses in laser-directed energy deposition (L-DED). The approach focuses on exploiting the potential of rapidly growing tools such as machine learning and polynomial chaos expansion for handling full-field data for measurements and predictions. In particular, the thermal expansion coefficient of thin-wall L-DED steel specimens is measured and then used to predict the displacement fields around the drilling hole in incremental hole-drilling tests. The incremental hole-drilling test is performed on cubic L-DED steel specimens and the displacement fields are visualized using a 3D micro-digital image correlation setup. A good agreement is achieved between predictions and experimental measurements.

Laser metal deposition is an additive manufacturing technology where the process parameters greatly influence the product quality and geometry. Therefore, it is generally concluded that real-time control and monitoring of the process is required to deliver quality parts. However, during the building process it is only possible to control the liquid state of the melt-pool. Monitoring of the melt-pool typically involves image processing and comes with an associated computational cost. CPU-based systems are limited in the real-time field due to their slower response than their FPGA counterparts for parallel computing and related latency. This paper presents an FPGA-based vision system that extracts geometric and intensity statistical features in real-time of the melt-pool based on a low-cost VNIR (Visible and Near-Infrared) camera. The extraction of features of the melt-pool is achieved by a thresholding approach and a 2D blob analysis using image moments. A pyrometer is used synchronously with thecamera to measure simultaneously the temperature in the melt-pool. The observed melt-pool shape and intensity features are compared to the temperature values, and the results are discussed and correlated.

One of today's ongoing challenges in directed energy deposition (DED) is controlling the geometry and material properties of parts. The objective of this paper is to investigate the relationship between several printing parameters of DED (laser power, laser speed, powder feed rate) and the melt pool temperature. Because DED is a complex and nonlinear process, well-established supervised-learning models such as support vector regression and artificial neural networks are particularly well suited to represent it. The MiCLAD machine, designed at the Vrije Universiteit Brussel, is equipped with a hyperspectral camera that monitors the light emitted at several wavelengths by the melt pool during the building process. A steady-state data set produced by the hyperspectral camera is postprocessed by an advanced temperature estimation method, and the limitations of the temperature estimation method are identified and discussed. The temperature data are used as training data for supervised-learning methods, and a studyis conducted to compare the performance of the considered methods using the measured optical data. This study demonstrates that the melt pool temperature of the DED process can be effectively modeled through the printing parameters thanks to supervised-learning methods.

Additively manufactured components are questioned for their fatigue performance and therefore not adopted for safety critical applications so far. The components might contain material imperfections and residual stresses due to the high temperature gradients during the printing process. These stresses alter the component's structural integrity and are one of the main sources of component deformation and cracking. The effective Structural Health Monitoring system (eSHM), developed and patented by the Vrije Universiteit Brussel, fully exploits the flexibility offered by the 3D printing process by integrating a smart continuous monitoring technology inside additively manufactured parts. The system is based on the detection of pressure changes in 3D-curved internal channels embedded in the fatigue critical regions of the component. The pressure is continuously monitored at the channel extremities by externally mounted pressure sensors. The system is capable of detecting the presence and finding the location of a fatigue crack. This paper proposes an assessment of the production of a fatigue sample with integrated eSHM-system using the MiCLAD in-house hybrid Directed Energy Deposition machine developed by the Additive Manufacturing Research Group of the VUB. It has the particularity to allow the combination of both additive and subtractive milling operations for the production of a part. Different printing strategies are presented and compared in terms of final geometry and internal channel roughness, known to be crucial to avoid undesired fatigue initiation and to obtain an accurate detection and localization of the fatigue crack. The encountered difficulties during the production process are identified and solutions are proposed.

This paper presents a numerical and experimental characterization of the powder-gas jet during the coaxial directed energy deposition process in terms of flow velocity, particle velocity, jet symmetry, and powder stream shape. The latter parameters are linked to the blown powder settings, known as the carrier and shielding gas volumetric flow rate, the powder mass flow rate, and the particle diameter. First, 2D axisymmetric steady-state computational fluid dynamics simulations of the turbulent gas flow are presented to illustrate the flow behavior and the effect of the presence of a substrate under the gas jet. Next, the free powder flow is visualized experimentally using high-speed imaging techniques. The particle velocities are extracted from the high-speed images by means of particle image velocimetry and compared with in situ measurements of the powder flow.

Directed energy deposition (DED) is a widespread laser additive manufacturing process characterized by the simultaneous laser and powder delivery. During the layerwise bounding of the material, many complex interconnected physical phenomena take place in a very short time. Presently, one of the main challenges faced by DED is to enhance the powder deposition efficiency. It is defined as the ratio of powder that has been effectively solidified in the part over the total amount of powder that flowed through the nozzle while the laser was on during the deposition process. Increasing the powder efficiency would allow us to minimize the powder waste, minimize the overall costs of the DED process, and therefore reduce the printed part cost. The present work, therefore, proposes to study the influence of laser beam diameter, stand-off distance, and gas/powder settings on the DED powder efficiency. The considered gas/powder settings are carrier and shielding gas volumetric flow rate, powder mass flow rate, and particle diameter. The efficiency is computed by means of the geometry of 316L stainless steel single-track deposits. The track geometries are evaluated based on the deposition width, deposition height, and area of deposition, all extracted from the deposition profiles measured by means of laser triangulation. Optical micrographs of the single-track transversal cross sections are linked to the obtained powder efficiencies. The final aim is to get insight into the effect of process parameters on the powder efficiency and print quality and to identify the optimal process parameter combination in order to maximize the powder efficiency.

The formation of a melt pool, during an additive manufacturing process, is a complex multi-physical process. Apparent factors, contributing to its complexity, are among others: the high temperatures, the existence of the material in multiple physical states and the involved high dynamics. However, it is a challenge to have full real-time control over all these factors. In this contribution, we will present our results and shed light, on the basis of measurements with a scanning laser doppler vibrometer, on how the vibrational behavior of the melt pool relates to its penetration depth into the substrate.

The MiCLAD machine designed at the VUB, Belgium, allows for closed-loop controlled laser metal deposition including various in-situ optical based measurement systems. These integrated sensors collect information on deposition geometry and temperature during the building process. Hence, each cubic millimeter of material that is either added or removed is mapped to its digital twin with a millisecond temporal resolution in the machines database. This paper introduces the platform and its capabilities by focusing on the procedure of obtaining the necessary training data for the future application of machine learning algorithms, with the goal of controlling the geometry and temperature history during additive manufacturing.

Additive manufacturing (AM) offers new manufacturing solutions for the integration of smart functionalities in engineering structures. In this paper, an analytical model is presented for an embedded load sensing element based on a liquid-filled capillary. During the additive manufacturing process, the capillary is integrated in the region where the strain is to be determined. The embedded capillary deforms as the structure deforms under an applied load, as such altering the pressure inside the capillary. The monitoring of the capillary pressure allows monitoring the loads and thus usage of the component. This paper presents a model describing the behavior of the sensing element under uniform tensile stress. The sensitivity of the load sensing element per unit longitudinal strain depends on the bulk modulus of the liquid inside the capillary and the Poisson coefficient of the surrounding material. The current work further compares the analytical model against static tension-compression tests of powder bed fused stainless steel (AISI 316L) test specimen with an integrated capillary filled with a liquid (water). Similarly, the validation of the model is then checked against a dynamic four-point bending test on a Ti-6Al-4V specimen produced by powder bed fusion.

One of the major challenges faced by laser-based Directed Energy Deposition (DED) is the process efficiency. This efficiency is significantly affected by the percentage of blown powder that effectively reaches the spherical metal melt pool. Increasing the powder efficiency would allow to reduce the overall costs of the DED process and consequently reduce the printed part cost. The present work focuses on the experimental characterization of the powder-gas jet in terms of powder stream shape, stand-off distance and powder-jet focus diameter by using 3 different high speed image-based approaches. The powder jet parameters are linked to the blown powder settings (carrier gas, shielding gas, powder feed rate, particle diameter) and optimized to minimize the powder waste. The different approaches give equivalent results and the reduction of the particle diameter turns out to be the most relevant parameter to decrease the powder-jet focus diameter and therefore increase the powder efficiency.

One of today’s ongoing challenges in directed energy deposition (DED) is controlling the geometry and material properties of parts. This manufacturing process is complex and nonlinear due to multiple physical phenomena at play and is therefore hard to model analytically. Machine learning (ML) on the contrary is particularly well suited to predict the behavior of a complex process with multiple inputs and outputs such as DED. A significant amount of data is required to train machine learning models, but experimental data are costly time-wise and should therefore be produced in an intelligent way. As a stepping stone for the future production of experimental training data, a finite element model of the process was developed in this study as an unlimited source of training data for the ML models. This model takes into account the printing parameters (laser speed, laser power, and powder flow rate) and outputs’ simulated process monitoring data thanks to a postprocessing method that is outlined in this article. A dataset was produced by simulating 102 tracks in 316L stainless steel with the model. From the analysis of this dataset, it was shown that K-nearest neighbors, support vector regression, decision tree regression, linear regression, and artificial neural network models are all capable of modelling the relationship between the printing parameters and the melt pool characteristics effectively.

Decentralized control design is commonly applied to design controllers for multi-variable control applications. The success of decentralized control design methodologies hinges on the quality of decoupling of the system. The aim of this paper is to develop a decoupling procedure that applies to multi-variable systems and only requires a frequency response function of the system. The proposed method builds on recent tensor decomposition methods. The potential of the method is shown both in a simulation and using experimental data. When applied on noisy data and on real-life systems, which are not possible to be exactly decoupled, the proposed method typically performs better compared to eigen-decomposition based method.

Vibration measurements of the fuel assembly of a nuclear reactor are a very useful tool to determine the health and lifetime of the reactor core. The importance of these measurements is exacerbated in the new generation of heavy liquid metal reactors, where the fuel assembly is exposed to a corrosive molten metal coolant at 300 $^\circ$ C and where the space between the individual fuel pins is limited to a few millimeters. In this paper, we consider fibre Bragg gratings as potential candidates for carrying out fuel pin vibration measurements in such an environment. We describe a dedicated method to integrate fibre Bragg gratings in a fuel pin, and we subject this pin to conditions close to those encountered in a real heavy liquid metal reactor. More specifically, we report on the performance of draw tower gratings used as a vibration sensor when the fuel pins are immersed in heavy liquid metal at 300 $^\circ$ C for up to 700 h. The performance evaluation is based on monitoring the signal-to-noise ratio of the grating's spectral response as a function of time. We show that accurate detection of the Bragg peak becomes very challenging after 400 h of exposure. Additionally, we succeed to extend the useful lifetime with a factor of two by using an appropriate integration of the fiber in the fuel pin and by using an alternate peak detection algorithm.

This paper addresses the influence on the fatigue life induced by the implementation of a capillary-based structural health monitoring methodology, patented under the name eSHM. It consists in integrating structurally small and pressurized capillaries into the component, so that when a fatigue crack breaches the capillary network, it results in a leak flow to the open atmosphere and loss of pressure in the galleries which is detected by a pressure sensor. The novelty of the proposed system resides in the opportunity to locate the capillary according to the designer’s need, as one resorts to additive manufacturing for the part production. However, the presence of these galleries in highly stressed regions raises concerns about crack initiation at the capillary itself and accelerated fatigue crack growth. This paper aims at the quantification of the influence the eSHM has on the fatigue behavior of the component and the determination whether this influence is significant or not. To that purpose, numerical simulations on a straight lug component, using the finite elements and eXtended Finite Elements Methods (XFEM), are performed. Various capillary sizes and shapes are assessed, so as to enable a general conclusion on the impact of the eSHM methodology in straight lugs.

Incore temperature measurements are crucial to assess the condition of nuclear reactor components. The sensors that measure temperature must respond adequately in order, for example, to actuate safety systems that will mitigate the consequences of an undesired temperature excursion and to prevent component failure. This issue is exacerbated in new reactor designs that use liquid metals, such as for example a molten lead-bismuth eutectic, as coolant. Unlike water cooled reactors that need to operate at high pressure to raise the boiling point of water, liquid metal cooled reactors can operate at high temperatures whilst keeping the pressure at lower levels. In this paper we demonstrate the use of optical fibre sensors to measure the temperature distribution in a lead-bismuth eutectic cooled installation and we derive functional input e.g. the temperature control system or other systems that rely on accurate temperature actuation. This first-of-a-kind experiment demonstrates the potential of optical fibre basedinstrumentation in these environments. We focus on measuring the surface temperature of the individual fuel rods in the fuel assembly, but the technique can also be applied to other components or sections of the installation. We show that these surface temperatures can be experimentally measured with limited intervention on the fuel pin owing to the small geometry and fundamental properties of the optical fibres. The unique properties of the fibre sensors allowed acquiring the surface temperatures with a resolution of 30 mK. With these sensors, we assess the condition of the test section containing the fuel assembly during different steps of the operation of the facility, including the heating and verification of the vacuum of the loop as well as the filling and draining of the LBE loop. We also identify a simulated electrical shut-down and heating circuit failure in the primed installation. (C) 2015 Elsevier B.V. All rights reserved.

Composite materials are nowadays widely used in several applications, especially in the aerospace field. Despite the numerous advantages that composites can offer, the health monitoring of this type of structures is challenging. A major challenge regards the detection and monitoring of delaminations. In the case of aeronautical structures, delaminations are often caused by impacts with external objects. Therefore, the capability of detection, localization and reconstruction of occurring impacts becomes an essential health monitoring tool for the estimation of the remaining structure lifetime. In this paper, we propose a new procedure for the reconstruction of impacts, based on the structural model and on measured vibration data. The procedure is tested over a carbon fiber reinforced composite plate, instrumented with surface mounted fiber Bragg grating sensors. The experimental results show that the proposed algorithm (VS-LS) can localize and reconstruct the impact forces with at least three times better accuracy than the classical pseudo-inverse method. (C) 2016 Elsevier Ltd. All rights reserved.

The Active Aeroelastic Test Bench (AATB) is a novel wind tunnel test setup that used higly dynamic electric linear actuators to command the pitch and plunge position of a cantilver wing. For the presented study the AATB is used to study the level of nonlinearities in the response of a cantilever wing to enforced 1-cos gust motion as well as multisine experiments. The cantilever wing being studied in the AATB is equiped with accelerometers and optical Fibre Bragg Grating (FBG) strain sensors. Measurements in the linear aerodynamic regime at small angles of attack, as well as at large angles of attack are studied and compared. It is illustrated how the Best Linear Approximation (BLA) can be estimated from multisine experiments using the Local Polynomial Method (LPM) as a first step to the modeling of the nonlinear aeroelastic response from wind tunnel tests.

This article introduces as new approach for flight envelope clearance of aircraft. This can now be performed as one continuous test, resulting in a major time saving. Both analysis of the current behaviour of the structure, and prediction towards higher velocities are important for flight flutter testing, and are dealt with in this article. The time-varying weighted non-linear least-squares estimator is used to obtain the frozen system poles of a cantilever wing under test in a low speed wind tunnel. The obtained smooth variation of the transfer function coefficients is used as basis for predicting the damping ratio towards higher velocities. Comparison with time-invariant measurements have shown that only one tenth of the measurement time is needed to obtain equal uncertainties on the estimates when performing a flight flutter test as one continuous test.

The Active Aeroelastic Test Bench is used for the study of low subsonic unsteady aerodynamics through forced motion experiments of rigid wings in pitch and plunge. The use of highly dynamic electric linear actuators allows for arbitrary motion in both the pitch and plunge degrees of freedom. In this paper, we discuss the forces measured on a constant section NACA 0018 wing subjected to isolated single and multisine motion in pitch or plunge. A data-driven approach is used, in which the importance of filtering and averaging of the noisy wind tunnel data is discussed. It is concluded from the forced motion experiments that a good agreement between the single and multisine experiments exists. However, only the pitch experiments agree with predictions from Theodorsen’s unsteady aerodynamics theory. The plunge experiments show the presence of major nonlinear contributions in the measured force response. Because of friction in the linear bearings of the setup, it cannot be excluded that these are generated by the plunging mechanism or the setup itself. Additionally, it is illustrated how the use of multisine signals for the identification of unsteady aerodynamic transfer function models allows for a reduction of wind tunnel test time by a factor of 10.

In-core vibration measurements are crucial in view of preventing excessive vibration already in the design stage or assessing component integrity during operation. This issue is exacerbated in new reactor designs that use liquid metals, such as for example a molten lead-bismuth eutectic, as coolant. In this paper we demonstrate the use of optical fibre sensors to measure the vibration induced by the coolant flow in a lead-bismuth eutectic cooled installation and we apply modal analysis techniques to derive modal parameters of the system. This first-of-a-kind experiment demonstrates the potential of optical fibre based instrumentation in these harsh environments. We focus on measuring the vibration of the individual fuel rods in the fuel assembly, but the technique can also be applied to other components or sections of the installation. We show that the vibration on the fuel rods can be experimentally measured with limited intervention on the fuel pin owing to the small geometry and fundamental properties of the optical fibres. With these sensors, we determine the vibration amplitude and modal parameters of the fuel assembly containing the fuel rods during different steps of the operation of the facility, including the initial start-up of the coolant circulation as well as during continuous operation.

In this article, we develop a nonlinear data-driven model of the aerodynamic force on an aerofoil made to oscillate in pitch in a wind tunnel under constant flow conditions. The proposed model structure is a polynomial nonlinear state-space model (PNLSS) which is an extension of the classical linear state-space model with nonlinear functions. In this work, we use experimental data of an aerofoil onto which a pitching motion is imposed, using a dedicated cyber-physical fluid dynamic (CPFD) setup. The pitch angle is considered as the input to the system, while the aerodynamic force is considered as the output of the system. Different input signals (single sines, multisines) are considered at different amplitudes and offset angles for the pitching aerofoil. Results show that the PNLSS model is three times more accurate than the linear model in minimising the least squares cost function. To the author’s knowledge, this constitutes the first instance of a fully nonlinear state space model of an oscillating aerofoil developed from experimental measurements.

In the field of additive manufacturing the implementation of real time process monitoring is required to enhance the quality and repeatability of the printed parts. Laser metal deposition (LMD) is a technique where the deposition of material is done in a melt pool created by a high-power laser beam. This paper presents the geometrical properties of such melt pool obtained during blown powder LMD. For achieving these results, the visual and hyperspectral information of the melt pool is used. An analytical model of the melt pool boundary is fitted to high-speed visual images recorded during 3D deposition. At the same time the hyperspectral information provides a way to estimate the temperature and emissivity profiles within the melt pool. The proposed paper discusses the correlation of the hyperspectral and visual high-speed camera results.

The dynamic response of a cantilever wing at small and large angles of attack was tested using the Active Aeroelastic Test Bench (AATB) in the wind tunnel at low wind velocity 8 m/s, Re ≈ 120, 000). The compliant wing was excited at the root of the wing with a small 2◦ pitch angle. Advantage was taken from the fact that the AATB allows for arbitrary excitation of the wing in pitch and plunge at its constrained root. This property is exploited in order to excite the wing with an odd random phase multisine. The response of the instrumented wing was recorded by using both fibre Bragg grating (FBG) optical strain gauges and piezoelectric accel-erometers. Using the spectral properties of the applied odd random phase multisine, the level of stochastic nonlinear distortions and the Best Linear Approximation (BLA) of the FRF was estimated for different angles of attack. A clear increase of the level of stochastic nonlinearities was observed when increasing the angle of attack up to 17.5◦. These distortions were mainlyfocussed around the resonance frequencies and in the very low frequency range. This latter could only be observed in the measurements of the FBG strain gauges due to the high pass filter effect of the piezoelectric accelerometers. Comparison of the identified BLA with gust response measurements showed that in general good predictions from the BLA model were possible when the Signal to Distortion Ratio (SDR) was above 10 dB. For the tested wing this was for angles of up to 12◦. Above this angle all resonance frequencies were still well approx-imated both in frequency and amplitude, however, a very important low frequency contribution was missing from the predicted data.

The Active Aeroelastic Test Bench (AATB) is designed for the study of low subsonic unsteady aerodynamics and aeroelasticity. The AATB is equipped with four linear motors offering separate control of the pitch and plunge position of the wing. The AATB allows to perform forced motion unsteady aerodynamic experiments when operated in open loop. In addition, when operated in closed loop, aeroelastic experiments can be performed, with both linear, or nonlinear variable structural behaviour. For this closed-loop operation the directly measured loads are fed back to the real-time controller. This controller then outputs the user desired position, emulating structural behaviour of the wing.

In aerodynamics, as in many engineering applications, a parametrised mathematical model is used for design and control. Often, such models are directly estimated from experimental data. However, in some cases, it is better to first identify a so-called nonparametric model, before moving to a parametric model. Especially when nonlinear effects are present, a lot of information can be gained from the nonparametric model and the resulting parametric model will be better. In this article, we estimate a nonparametric model of the lift force acting on a pitching wing, using experimental data. The experiments are done using the Active Aeroelastic Test Bench (AATB) setup, which is capable of imposing a wide variety of motions to a wing. The input is the angle of attack and the output is the lift force acting on the NACA 0018 wing. The model is estimated for two different types of input signal, swept sine and odd-random multisine signals. The experiments are done at two different pitch offset angles (5° and 20°) with a pitch amplitude of 6°, covering both the linear and nonlinear aerodynamic flow regime. In the case of odd-random multisines nonlinearity on the FRF is also estimated. We show that the level and characterisation of the nonlinearity in the output can be resolved through a nonparametric model, and that it serves as a necessary step in estimating parametric models.