Search

Data analysis requires a high level of expertise for domain workers, and AutoML aims to make these decisions in an automated way. But it is still a difficult problem to automatically generate machine learning pipelines with high performance in acceptable time. This paper presents a DFSR (Data Feature and Service Association) approach to automatically generating machine learning pipelines utilizing data features and service associations. The experimental results showed that the performance of the generated pipelines reached the satisfactory level of current AutoML tools, and the time consumption is reduced to the minute level.

From the perspective of financial management, financing risk comes as one of the primary hazards. The multinational corporation will be faced will more risk in international trade, such as fluctuations in exchange, friction of investment, interference from foreign governments and international economic sanctions. It is of great importance for a multinational corporation seeking its survival and development overseas to avoid this kind of risk and to minimize the loss. Therefore, it is of absolute necessity to establish the Financing Risk Warning System. The Financing Risk Warning System for multi-national corporations is based on the corporate financial warning theory and gives the warning from the financial perspective with the management of the timely financial figures and relevant variations. The system include at least managing center, risk signals collecting center, and risk dealing department.

In order to meet wide angle scanning requirements of the satellite antenna, a side-fed parabolic torus reflector antenna is designed in this paper. The ellipsoid is adopted as sub-reflector with the purpose of reduction the mechanical equipments. The sub-reflector is rotated in azimuth to scan the antenna beam in the direction (±20°) for wide swath coverage. The calculated results guiding the antenna optimization are based on physical optics (PO) analysis and agree well with our design goals.

The design and performance of an novel edge-fed microstrip array antenna are introduced, which generates sum and difference far field patterns for monopulse tracking. Each microstrip line array is fed at the edge side with the middle of the row shorted. A compact slot patch antenna element loaded with a pair of spirals is proposed, which is 40% smaller than the conventional half-wave patch antenna. The -19dB side-lobe level of sum pattern and 37dB null depth of difference pattern have achieved in the experiment.

A novel millimeter wave monopulse slotted waveguide array antenna with a central hole for infrared sensor use is presented. The amplitude distribution is optimized, and a monopulse comparator is designed to feed the proposed array. This millimeter wave array antenna has been fabricated and measured. The gain of the sum beam is higher than 32dB, whereas the zero depth of the difference beam is lower than -28dB. The side-lobe level is -15.0dB in E-plane and that in H-plane is -13.0dB.

Over the past several years, we have developed several fluxless bonding processes using the silver-indium (Ag-In) system. With innovative design, the joint achieves high melting temperature, higher than 600°C, even when the bonding is performed at low temperature, 180°C. The joint has the structure of Ag/(Ag)/Ag2In/(Ag)/Ag where (Ag) is the solid solution of Ag with In. Shear test results show that Ag2In is the weak region in the structure. To increase the joint strength, an idea is to convert Ag2In compound into (Ag) by continuous reaction with the adjacent (Ag) regions. When this happens, the melting temperature of the joint will increase to 790°C. In this research, we studied this converting process with systematic experiments. The joint is made between two copper (Cu) substrates to ensure that the samples break in the joint rather than in the substrate during shear test. The upper Cu was electroplated with 15μm thick Ag layer. The lower Cu was electroplated with 15μm thick Ag layer, followed by 8μm indium, which was the capped by 0.1μm Ag layer to avoid In oxidation. The samples were bonded at 180°C in 100 millitorr for 5 minutes. SEM & EDX analysis show that the joint has structure of Ag/(Ag)/Ag2In/(Ag)/Ag, where the Ag2In region is 15μm. Samples are well bonded between Cu substrates without crack or voids. To convert Ag2In region into (Ag), samples were annealed at 200°C for 200, 300, 500 hours, respectively. 200°C was chosen because it is more compatible with the surviving temperature of most devices. The samples were cut in cross section, polished, and examined using scanning electron microscopy (SEM) with energy dispersive X-ray spectroscopy (EDX) to study its microstructure and composition. At 200 hours, the sample is still well bonded. The joint turned into Ag/(Ag)/alloy/(Ag)/Ag, where alloy is a mixture of Ag2In and Ag3In compounds. At 300 hours, the joint structure has not changed but the alloy region becomes thinner. At 500 hours, the alloy region is almost gone. The annealing process continues. After the alloy region turns into (Ag), six samples will go through shear test. Shear strength will be measured and fracture mode identified. We expect the (Ag) phase to be significantly stronger than the Ag2In compound. This bonding and annealing processes thus provide a new way to produce strong and high-temperature joints at low temperature, 200°C.

The operating temperature of packaged devices is often constrained by the melting temperature of the die-attach material. Taking the most popular lead-free solder SAC as an example, its melting temperature is 217°C. Using 80% of the homologous as a guideline of maximum operating temperature, it corresponds to 119°C. Many advanced semiconductor chips can operate at a much higher temperature. It is thus clear that high temperature die-attachment materials are in demand. The challenge is that typical bonding processes require the bonding temperature to be higher than the melting temperature. This temperature is often too high for the chips and other packaging materials to survive. In this research, we looked into the Ag-In system and developed a bonding process that needs only 190°C but produces joints that have melting temperature higher than 600°C at least. The process is entirely fluxless. To demonstrate this new process, silicon chips were bonded to aluminum (Al) substrates. Al was chosen as demonstration substrate because it has the highest CTE, 23×10-6/°C, among all substrate materials. The Si-Al pair provides the largest CTE mismatch among all electronic bonding structures. This large CTE mismatch allows us to show the strength and durability of the resulting Ag-In joints. Any weak joints would have broken as the samples cooled to room temperature as a result of the severe CTE mismatch. Many samples were produced. Despite the large CTE mismatch, no joint breaks. To evaluate the joint quality, composition and microstructure, scanning electron microscopy (SEM) with energy dispersive X-ray spectroscopy (EDX) is used. Nearly perfect joints are produced. The joint consists of Ag, solid solution (Ag), and Ag2In in stack of Ag/(Ag)/Ag2In/(Ag)/Ag. The melting temperature is higher than 695°C. To determine the strength of the joints, six samples went through shear test. They all pass the 5kg requirement specified in MIL-STD-883H standard. The fracture force ranges from 17kg (6.7MPa) to 94kg (36.9MPa). There are two fracture types. One is fracture within Ag-In joint where it breaks along the Ag2In region in brittle mode. The other type is Si chip breakage. It is worthwhile to mention that Al is not solderable. A new process was developed to make Al bondable that results in high bonding strength as evidenced by the high shear strength measured.

A novel compact slot patch antenna with two spirals on single thin substrate is introduced, which is 66% smaller than the conventional half-wave patch antenna. The four parallel slots on surface can enhance the length of current path and reduce the size of patch, the two spirals used can decrease the height of antenna at the same tme. Good agreement between the theoretical and experimental results is obtained. The measured return loss curve and radiation patterns are all presented.

Since Ag and In react to form compound at room temperature was investigated, but the chemical reactions during bonding are not well understood. The Ag-In phase diagram shows two dominating intermetallic compounds (IMC), AgIn 2 and Ag 2 In. Therefore, based on the following systematic experiments to study the chemical reactions of AgIn system and compound formation. Copper (Cu) substrates were electroplated with 40 μm thick Ag layer, followed by indium layer of 1, 5, 10, and 15 μm, respectively. Thick Ag layer is chosen to prevent In reaction with underlying Cu substrate. The samples were annealed at 180°C in 100 millitorr for 5 minutes to emulate the bonding conditions. The microstructure and the composition on the surface of the samples were evaluated using scanning electron microscopy (SEM) with energy dispersive X-ray spectroscopy (EDX). X-ray diffraction (XRD) was also employed to identify the IMC on the surface. The samples were then cut in cross section, polished, and studied also by SEM and EDX. The results show that, for the sample with 1μm thick indium, the In layer converts to Ag 2 In entirely after annealing. For sample with In layer thickness of 5 μm, AgIn 2 , Ag 2 In, and silver solid solution (Ag) all exist after annealing. No indium was identified. For sample having 10, and 15μm thick indium, In was detected on the surface, but was not found on the cross section. The polishing process probably bevels the thin In layer on the surface, making it invisible. We next studied what would happen if the Ag was annealed first before plating In. After plating the Ag layer, samples were annealed at 450°C for 3 hours to grow Ag grains. This was followed by plating 10μm In and annealing at 180°C. The result shows AgIn 2 , Ag 2 In as well as In. Significant In remains on the surface. The thicknesses of AgIn 2 and Ag 2 In layers are thinner than the sample whose Ag is not annealed.

A novel design of monopulse microstrip radiator array with side feed network has been introduced in this paper. The simulated and measured radiation patterns at center frequency are both presented. The −18dB side-lobe level for the sum pattern and −28dB null depth for the difference pattern have achieved in the experiment at center frequency.

A new slotted rectangular waveguide array antenna is introduced, which is constructed on the conformal cylindrical surface. Unlike other slotted waveguide arrays, this array has a beam with a large tilt angle from the broadside of the linear array. The array consists of a waveguide with inclined slots cutting in the narrow wall and a matching load at one end for the traveling wave operation. The antenna is designed and built using the analytical procedure combining with the commercial software. The radiation patterns and gain of the antenna are measured, and the experimental data obtained are compared with the theoretical results.

A novel compact slot patch antenna loaded with a pair of spirals on thin substrate is introduced, which is 66% smaller than the conventional half-wave patch antenna. The effects of four parallel slots on surface current distribution and loading spirals on impedance matching are all investigated theoretically respectively. Good agreement between the theoretical and experimental results is obtained. The measured return loss curve and radiation patterns are all presented.

The successful design of complex cooperative planning process on daily dispatching plan allows multi dispatchers in railway bureaus to cooperatively make daily dispatching plans. This is helpful to improve the accuracy of daily dispatching plan and further to improve the overall efficiency of railway transportation in China. And building the cooperative planning process is also important for the configuration of the transporting dispatching management information system (TDMS 4.0) used in railway bureaus in China. However, designing such a process is very difficult, considering the complexity of the process. In this paper, a new design method based on roles, activities and time sequences is provided. Finally, the new cooperative planning process is constructed by re-arranging the time sequences relationships among roles, activities in the current process.

The structure of the elbow-bar mechanism of injection molding machine has a great influence on the axial dimension and efficiency of clamping unit. Based on the motion analysis of the toggle clamping system of the injection molding machine, a mathematical model was established for the optimization design of elbow-bar mechanism. With the least length of the elbow-bar mechanism used as the optimization objective, the design obtained by genetic algorithm was compared with the conventional method. The result shows that the genetic algorithm is much better than conventional design methods and has practical application value.

A time-frequency analysis technique, combined with kurtosis method and wavelet analysis, was present for the detection and diagnosis of the faults based on the unstable vibration signals from the rolling bearings. With this method, the signals were decomposed and reconstructed by the wavelet analysis, followed by the analysis of demodulation and spectral refining by using Hilbert transformation. The experiment results show that the fault information of the rolling bearings can be detected and diagnosed effectively, which favor the quick determination of the detailed faulty type within the bearings.

This letter outlines the design and manufacture of a UHF-band miniaturized band-pass filter realized by low-temperature cofired ceramic (LTCC) technology for super heterodyne microwave receiver applications. A distinct feature of this filter is the smaller size than conventional filter at such a low frequency and steady pass-band temperature electrical performance compared with surface acoustic wave filter. The band-pass filter with a central frequency of 750 MHz and a 40MHz pass-band is designed as a three-dimensional (3-D) structure based on distributed components. This filter has excellent out-of-band attenuation performance and it has 50.1 dB and 56.8 dB of attenuation at 550MHz and 950MHz respectively. The measured insertion loss is less than 2.7 dB at 750MHz, and the input/output VSWR is less than 2.0. The overall size of the miniaturized filter is only 4.5mm×3.2mm×1.5mm.

In recent years, many applications in precision engineering require a careful isolation of the instrument from the vibration sources by adopting active vibration isolation system to achieve a very low remaining vibration level especially for the very low frequency under 10Hz vibration signals. In this paper, based on the previous research experiences in the system- atical modeling and study of parallel robots, a hybrid robot is described and the vibration model is given by using Lagrange's Equations. Then the present study addresses the issues related to the active vibration control schemes for the MIMO system using LQR algorithm. Finally, numerical simulations on the effect of active vibration control are presented.

In recent years, many applications in precision engineering require a careful isolation of the instrument from the vibration sources by adopting active vibration isolation system to achieve a very low remaining vibration level especially for the very low frequency under 10Hz vibration signals. In this paper, based on our previous research experiences in the systematical modeling and study of parallel robots, a hybrid robot is described and the vibration model is given by using Lagrange's Equations. Then a dynamic analysis by using Kane's method is presented to verify the result. Finally, numerical simulations on the dynamic characteristics and natural frequency are presented.

This paper is concerned with the design and modeling of a 3-DOF dual parallel robot with flexure hinges. Based on our previous research experiences in the systematical modeling and study of parallel robots, a dynamic modeling has been performed by using Kane's method and Newton-Euler equation. A state-space framework is defined to facilitate the design of an optimal controller. Numerical simulations are presented to verify the validity of both models. The investigations of this paper may not only have significance in robotics research, but also make contributions to the research on dynamic characteristics based on novel parallel manipulators.

In recent years, nanotechnology has been developing rapidly due to its potential applications in various fields that new materials and products are produced. In this paper, a novel macro/micro 3-DOF parallel platform is proposed for micro positioning application. The kinematics model of the dual parallel mechanism system is established via the stiffness model of individual wide-range flexure hinge and vector-loop equation. The inverse solution and parasitic rotation of moving platform is analyzed and simulated on a parallel mechanism with real parameters. Whereafter, the reachable and usable workspace of the macro motion and micro motion of the mechanism are plotted and analyzed. Finally, based on the analysis of parasitic rotation, usable workspace of macro and micro motion, an optimization of the parallel manipulator is presented. The investigations of this paper will provide suggestions to improve the structure and control algorithm optimization for the dual parallel mechanism in order to achieve the features of both larger workspace and higher motion precision.

Since many applications in precision engineering will strongly require a careful isolation of the process from the vibration environment to provide a sufficiently quiescent gravity environment, the active isolation system which can achieve a very low remaining vibration level has become more and more important in recent years. In this paper, a general dynamic model of a kind of parallel platform for vibration control applications based on KANEpsilas method is presented. The analysis result is a state-space, analytical set of linearized equations of motion. At the end of this paper the control system architecture is introduced for the kind of parallel platform as a multiple-input/multiple-output (MIMO) problem.

This paper presents kinematic analysis of two kinds of large displacement parallel platforms for micro/nano positioning applications. The kinematics model of the dual parallel mechanism systems is established via the stiffness model of individual wide-range flexure hinge. The displacements of the end platform and the input parameters of prismatic actuators are discussed and the corrected values of input motions are proposed on some checking points in workspace referring to the real parameters of two kinds of In-Plane parallel mechanisms. The FEA models are established by ANSYS software, both the theoretical analysis and FEA simulation results are presented and compared. The investigations of this paper will provide suggestions to improve the structure optimization for a class of parallel mechanism in order to achieve such features as larger workspace and higher motion precision.

Nowadays many applications in precision engineering would be impossible without a careful isolation of the instrument from the vibration sources. To reduce the vibration on precision instruments, active isolation systems are urgently required, since these units can achieve a very low remaining vibration level. This paper is concerned with the development of a vibration isolation system for highly sensitive equipment applications. In this paper, a 3-DOF parallel robot with flexure hinges will be analyzed and a dynamic model will be presented intended to facilitate the design of an optimal controller. The chosen approach is the method of Thomas R and the analysis result is a state-space, analytical set of linearized equations of motion. The investigations of this paper may not only have significance in robotics research, but also make contributions to the research on active vibration isolation based on parallel manipulators.

This paper presents a novel macro/micro 3-DOF parallel platform for micro/nano positioning applications. The kinematics model of the dual parallel mechanism system is established via the stiffness model of individual wide-range flexure hinge and vector-loop equation. The inverse kinematics is analyzed and simulated on a parallel mechanism with real parameters. Whereafter, the reachable and usable workspace of the macro motion and micro motion of the mechanism are plotted. Finally, the influence of load acted on the moving platform is discussed. The investigations of this paper will provide suggestions to improve the structure and control algorithm optimization for the dual parallel mechanism in order to achieve the feature of both larger workspace and higher motion precision.

Kinematics analysis and dimensional synthesis are two important problems of a parallel manipulator. Dimensional synthesis is optimization of the kinematic parameters according to desired workspace or other design requirements. In this paper, the dimensional synthesis of a 3-DOF parallel manipulator, which mimics DELTA robots, is studied considering maximum inscribed workspace and reciprocal of the condition number on the workspace section based on the concept. The kinematic model of the 3-DOF parallel manipulator is given at first. The golden section search is used to search the workspace of the manipulator and mesh the boundary. Then the algorithm for calculating the inscribed workspace and dimensional synthesis considering the inscribed workspace are presented. Thirdly, the distribution of dexterity index on the workspace section and dimensional synthesis considering the reciprocal of condition number is studied. The two dimensional synthesis results are compared at the end of the paper. While the synthesis methodology of the manipulator is studied. It is helpful to improve design efficiency of the 3-DOF parallel manipulator. The quickly design for the manipulator can be performed through the proposed methods of dimensional synthesis.

The paper presents a dissymmetrical tricept-like parallel robot providing two rotational freedoms and one translational freedom. It has a fixed base and a platform connected by three legs. One leg connects the fixed base and the platform by a universal joint a prismatic joint respectively. The other two legs are UPS chains or SPS chains. All of the three prismatic joints in the robot are actuated joints. The robots are aimed to use in robotics or machining. The mobility of the proposed structure is firstly analyzed. Then inverse kinematics problems are described in closed forms and velocity equation of the new parallel robot is given. After workspace analysis and dexterity analysis, the kinematic characteristics of the robot and the tricept robot are compared.