Search

Yield improvement is a critical component of semiconductor manufacturing. It is done by collecting, analyzing, identifying the causes of defects, and then coming up with a practical solution to resolve the root causes. Semiconductor components such as Through Silicon Vias (TSVs) and other package interconnects are getting smaller and smaller with the ongoing miniaturization progress in the industry. Detecting defects in these buried interconnects is becoming both more difficult and more important. We collect both 2D and 3D X-Ray scans of defective TSVs containing defects such as voids. We label the data in 3D and perform registration between 2D and 3D scans. We use this registration information to locate the TSVs and void defects in these 2D X-ray scans which would be difficult to label manually as these voids look very fuzzy in 2D scans. Thereafter we use a state-of-the-art deep-learning segmentation network to train models to identify foreground (TSV, void defects) from the background. We show that our model can accurately identify the TSVs and their voids in images where it is impossible to locate the defects manually. We report a dice score of 0.87 for TSV segmentation and a dice score of 0.67 for void detection. The dice score for voids demonstrates the capability of our models to detect these difficult buried defects in 2D directly.

As a piece of key equipment in the power system, GIS will cause great losses when it fails. Partial discharge detection technology is an important technic aims to evaluate the degradation status of GIS equipment insulation. Among them, the UHF method has been widely used to detect GIS partial discharge due to its superior anti-interference performance, but the partial discharge mechanism is more complicated and is affected by many factors such as operating voltage and load. Sometimes, the fault may develop quickly. Operation and maintenance personnel cannot resolve the fault in time, and this situation can be effectively solved by predicting the monitoring data of partial discharge. Although it is difficult to predict the long-term slow variability of partial discharges of internal devices in GIS, the detection of short-term variability discharge status is relatively stable. In this paper, an auto-regressive and moving average (ARMA) model is established for the extracted feature quantities in partial discharges in GIS. It can well predict the discharge trend in the few minutes or hours before the failure, which has good application value.

In high voltage power system, with the gradual increment of SF 6 gas closed-end insulation composite appliance, the application of Gas insulated switchgear (GIS) becomes more and more widely accepted based on its high reliability, less maintenance, long maintenance cycle, and low failure rate (only 20% ∼40% of conventional equipment). However, the GIS has its inherent shortcomings, as a result of the SF6 gas leakage, external water infiltration, the conductive impurities as well as insulator aging factors, such as internal flashover. To this end, this paper designs a rail-type inspection robot for GIS substation. First, it analyzes the background of high voltage substation and GIS. Secondly, considering the distribution of GIS and the task of substation inspection, the paper gives the key components of rail-type inspection robot, such as straight walking components, turning and changing track component, partial discharge (PD) detection component. Thirdly, it proposes that the architecture of rail type inspection robot system is a network distributed, which includes three parts: the background layer, communication layer and terminal layer. In addition, it discusses in detail the motion control method of inspection, especially positioning control. Then it explains the software and hardware structure of the motion control system: Radio frequency identification (RFID) is used for motion positioning of inspection robots; the background layer controls the motion of the inspection robot and uses the background software to analyze the detection data. Finally, the paper proves that the inspection robot has good balance, flexible turning, and a large inspection range. The robot can completely inspect work well instead of staff.

With the development of the scale of the power grid, Relay protection has gradually become an important part of ensuring the smooth and safe operation of power system while Pilot protection has also gradually become an important development direction of high-voltage line protection in the process of Relay protection development. Based on the research and comparison to the existing methods of current comparison pilot protection, direction comparison pilot protection and distance comparison pilot protection. In this paper, a protection method based on the characteristic of fault component of current at both ends of the circuit is proposed to overcome the disadvantage of the current differential protection scheme which is easily affected by the load current and the transition resistance, and the protection method is adopted which based on the instantaneous value of current fault component. In order to verify the rationality of the method which is proposed, an appropriate simulation model is designed, a simulation system is established, and various possible situations in the power system are simulated by using the simulation system. Simulation shows that the judgment based on the value of the fault current as a protective action has the following advantages: Firstly, the calculation is simple which only compares and judges the size of the fault component without real mathematical operations; Secondly, this method has lower requirements for the communication channel, which only a very simple signal needs to be transmitted when a possible fault is found; In addition, the action is fast, and only the instantaneous value of the fault component of the current can determine whether the action is or not.; Finally, the action is reliable, the judgment of this method is very reliable and will not refuse to acting. At the same time, it can avoid the influence of harmonics and other factors by adjusting the action threshold, which ensures that the protection will not malfunction.

The 500kV overhead transmission line induction power supply technology can provide stable and reliable power for high-voltage line online equipment and inspection robots. Since the magnitude of the induced current induced in the ground wire is small, in order to make full use of the induced energy, the principle of impedance matching can be used according to the topology of the circuit to increase the power drawn from the induction. This article focuses on the operating environment of 500kV overhead transmission lines, and designs a typical induction power circuit structure consisting of a power coil, a rectifier filter circuit, and a two-level DC-DC converter circuit. Based on this simplified equivalent circuit model, this paper designs the controller of the two-level DC-DC converter circuit, and realizes the maximum power point tracking (MPPT) control of inductive power draw by changing its duty cycle. Finally, the simulation results on the Simulink platform verify the accuracy of the simplified circuit model and the effectiveness of the maximum power point tracking control method.

Angle-based positioning has been agreed by 3GPP to be one of NR positioning techniques, and is now in standardization in 3GPP Release 16. In this paper, the key techniques for angle-based positioning are addressed including the localization algorithm and angle estimation algorithms. Uplink angle-of-arrival (UL-AOA) and downlink angle-of-departure (DL-AOD) techniques for both FR1 (sub-6GHz) and FR2 (above-6GHz) bands are presented along with their impacts on the existing NR standards. Simulation results show that both UL-AOA and DL-AOD can reach meter level accuracy in indoor scenarios, and hence can well meet the positioning requirements of the commercial cases defined in 3GPP NR. Compared to traditional timing-based positioning, the angle-based positioning is robust to the synchronization error, which makes it competitive for NR positioning.