A wavelet-based multi-dimensional temporal recurrent neural network for stencil printing performance prediction

The solder paste printing (SPP) is a critical procedure in a surface mount technology (SMT) based assembly line, which is one of the major attributes to the defect of the printed circuit boards (PCBs). The quality of SPP is influenced by multiple factors, such as the squeegee speed, pressure, the stencil separation speed, cleaning frequency, and cleaning profile. During printing, the printer environment is dynamically varying due to the physical change of solder paste, which can result in a dynamic variation of the relationships between the printing results and the influential factors. To reduce the printing defects, it is critical to understand such dynamic relationships. This research focuses on determining the printing performance during printing by implementing a wavelet filtering-based temporal recurrent neural network. To reduce the noise factor in the solder paste inspection (SPI) data, this research applies a three-dimensional dual-tree complex wavelet transformation for low-pass noise filtering and signal reconstruction. A recurrent neural network is utilized to model the performance prediction with low noise interference. Both printing sequence and process setting information are considered in the proposed recurrent network model. The proposed approach is validated using practical dataset and compared with other commonly used data mining approaches. The results show that the proposed wavelet-based multi-dimensional temporal recurrent neural network can effectively predict the printing process performance and can be a high potential approach in reducing the defects and controlling cleaning frequency. The proposed model is expected to advance the current research in the application of smart manufacturing in surface mount technology.     

Additive manufacturing of zirconia ceramics by material jetting

The possibility of additive manufacturing of ceramics has been reported widely in scientific literature. This study investigates the potential of direct inkjet printing or material jetting of 3Y-TZP ceramics by assessing the microstructure and mechanical properties of the sintered printed parts. The technique allows to print in layers of 10.5 μm, with an as-printed green density of 58 % and nearly fully sintered density of 6.03 ± 0.1 g/cm³ (99.7 % TD). The dimensions of the green and sintered parts were highly accurate but showed an anisotropic roughness in function of the building direction, mainly due to the support structures. The biaxial bending and 4-point bending strength of the sintered material was found to be substantially higher in the XY direction than in the building (Z) direction. SEM and X-Ray computed tomography revealed the presence of delamination cracks, agglomerates and spherical pores, which were identified as fracture origins on fractured surfaces.     

Feasibility of manufacturing of Al2O3–Mo HTCC by hybrid additive process

Additive manufacturing processes make it possible to produce increasingly complex 3D parts. In addition, these numerical processes can be usefully used to manufacture ceramic/metal parts of high dimensional resolution with thermal, electrical and electronic properties of interest for applications in the field of power electronics.In this context, a hybrid additive machine was developed to manufacture ceramic/metal parts. This machine consists in the combination of two additive manufacturing processes: stereolithography and robocasting.Using this hybrid process, the feasibility of HTCC components has been demonstrated by building dielectric alumina by stereolithography and molybdenum conductive network by robocasting. Molybdenum-based metallic formulation adapted to the process and allowing to obtain a high conductive metallic network has been developed. The co-debinding and co-sintering cycles have been optimized to minimize the content of residual carbon and to prevent the oxidation of molybdenum. The alumina/molybdenum interface has also been observed to conclude about a possible delamination between these two materials with different thermal expansion coefficients (CTE). Sintered HTCC parts have been characterized in the domain of hyperfrequency. The frequency responses deviate from the simulation due to a lack of dimensional accuracy of the metallic network.     

The effect of particle size distribution on the microstructure and properties of Al2O3 ceramics formed by stereolithography

Stereolithography (SL) shows advantages for preparing alumina-based ceramics with complex structures. The effects of the particle size distribution, which strongly influence the sintering properties in ceramic SL, have not been systematically explored until now. Herein, the influence of the particle size distribution on SL-manufactured alumina ceramics was investigated, including bending strength at room temperature, post-sintering shrinkage, porosity, and microstructural morphology. Seven particle size distributions of alumina ceramics were studied (in μm/μm: 30/5, 20/3, 10/2, 5/2, 5/0.8, 3/0.5, and 2/0.3); a coarse:fine particle ratio of 6:4 was maintained. At the same sintering temperature, the degree of sintering was greater for finer particle sizes. The particle size distribution had a larger influence on flexural strength, porosity and shrinkage than sintering temperature when the particle size distribution difference reached 10-fold but was weaker for 10 μm/2 μm, 5 μm/2 μm and 5 μm/0.8 μm. The sintering shrinkage characteristics of cuboid samples with different particle sizes were studied. The use of coarse particles influenced the accuracy of small-scale samples. When the particle size was comparable to the sample width, such as 30 μm/5 μm and 5 mm, the width shrinkage was consistent with the height shrinkage. When the particle size was much smaller than the sample width, such as 2 μm/0.3 μm and 5 mm, the width shrinkage was consistent with the length shrinkage. The results of this study provide meaningful guidance for future research on applications of SL and precise control of alumina ceramics through particle gradation.     

Direct additive manufacturing of melt growth Al2O3-ZrO2 functionally graded ceramics by laser directed energy deposition

Functionally graded ceramics (FGC), which combine properties of different ceramics in one part, usually have better comprehensive function and structural efficiency. In this study, four different gradient transition Al₂O₃-ZrO₂ FGC samples were prepared by laser directed energy deposition (LDED) method. The results show that there is an obvious interface in direct transition sample. The transition section bears tensile stress caused by difference of thermophysical properties of materials, resulting in significant longitudinal cracks. Element transition in interface region shows a step sharp transition. The direct transition sample shows intergranular fracture and the bonding strength is very low. Gradient transition mode can effectively suppress cracks, and avoid the step transition of microstructure and elements. Elements, microhardness of 25, 20 wt% FGC samples realized a nearly linear smooth transition. The interface fracture of FGC samples changed to transgranular fracture, bonding strength was significantly improved, and the maximum flexural strength reached 160.19 MPa.     

Effect of lognormal particle size distributions on particle spreading in additive manufacturing

Additive manufacturing (AM) has attracted much attention worldwide in various applications due to its convenience and flexibility to rapidly fabricate products, which is a key advantage compared to the traditional subtractive manufacturing. This discrete element method (DEM) study focusses on the impact of particle polydispersity during the particle spreading process on parameters that affect the quality of the final product, like packing and bed surface roughness. The particle systems include four lognormal particle size distribution (PSD) widths, which are benchmarked against the monodisperse system with the same mean particle diameter. The results reveal that: (i) the solid volume fraction of the initial packed particle bed in the delivery chamber increases then plateaus as the PSD width increases; (ii) regardless of PSD width, the solid volume fraction of the particle bed increases with spreading layer height before compression, but decreases with layer height after compression; (iii) the bed surface roughness increases with PSD width or layer height both before and after the compression of the spreading layer; (iv) the extent of increase in solid volume fraction during compression is correlated with the extent of decrease in bed surface roughness; and (v) the broader PSDs exhibit larger fluctuations of solid volume fraction of the particle bed and bed surface roughness due to greater variability in the arrangement of particles of different sizes. The results here have important implications on the design and operation of particle-based AM systems.     

Lightweight reactor design by additive manufacturing for preheating applications using metal hydrides

Thermal energy storage systems based on metal hydrides can be a solution for preheating fuel cells (FCs). They can provide thermal energy at temperatures below −20 °C during startup, while heat at 50 °C during operation is sufficient for regeneration. The challenge of such a system in mobile applications is the final weight specific thermal power. In this study, a reactor design based on additive manufacturing techniques for ~300 g of metal hydride is presented. Here, a reactor (passive) to hydride (active) mass ratio of 0.97 is realized, still reaching high weight specific thermal power of up to 2.1 kW/kg_MH at −20 °C and 8 bar (LmNi_4.91Sn_0.15). Considering the example of preheating a FC from −20 °C in ~120 s, the performance of LaNi₅ and LmNi_4.91Sn_0.15 is studied. While LaNi₅ requires higher regeneration temperatures than LmNi_4.91Sn_0.15 (>40 °C compared to >20 °C), its performance is less sensitive to operative variations due to its nearly ideal thermodynamic characteristic.     

Study of aroma compound formations and transformations during Jinxuan and Qingxin oolong tea processing

We characterised the formation and dynamic changes in the main aroma compounds produced during the manufacturing process of Jinxuan and Qingxin oolong tea. Thirty-five aroma compounds were investigated. Subsequent principal component and cluster analyses showed that the fresh leaves, spread leaves and rocked leaves of the two varieties of tea were distinguished from each other. Particularly, the aldehyde and ‘other’ compounds showed the highest correlation coefficients (0.71 and −0.70) among the principal. The heat map showed that the proportions of acetic acid, 3-methylbutanal and 2-methylbutanal significantly changed during the manufacturing process of the two tea varieties. The key amino acids (isoleucine, leucine, valine and aspartic acid) and enzymes (branched-chain amino acid aminotransferase, aspartate aminotransferase and acetaldehyde dehydrogenase) involved in the synthesis of 3-methylbutanal and 2-methylbutanal differed between the two varieties during manufacture. Our study reveals the characteristics of different varieties of oolong tea and their aroma formation after manufacture using the same process.     

犹豫模糊环境下准则具有偏好关系的TODIM方法

针对智慧制造评估时专家的决策信息具有犹豫模糊不确定性问题，提出了一种关于准则具有犹豫模糊偏好关系的改进交互式多准则决策（TODIM）方法。首先提出了准则间的犹豫模糊偏好关系概念，并证明了其基本性质。在TODIM方法优势度的计算过程中，将准则权重犹豫模糊偏好关系替代原有的精确值权重，使信息的准确性最大化。将该方法用于智能制造的评估上，实例分析结果表明所提方法是可行和有效的。