Design and implementation of dual-mode configurable memory architecture for CNN accelerator

With the rapid development of deep learning algorithms,the computational complexity and functional diversity are increasing rapidly.However,the gap between high computational density and insufficient memory bandwidth under the traditional von Neumann architecture is getting worse.Analyzing the algorithmic characteristics of convolutional neural network(CNN),it is found that the access characteristics of convolution(CONV)and fully connected(FC)operations are very different.Based on this feature,a dual-mode reronfigurable distributed memory architecture for CNN accelerator is designed.It can be configured in Bank mode or first input first output(FIFO)mode to accommodate the access needs of different operations.At the same time,a programmable memory control unit is designed,which can effectively control the dual-mode configurable distributed memory architecture by using customized special accessing instructions and reduce the data accessing delay.The proposed architecture is verified and tested by parallel implementation of some CNN algorithms.The experimental results show that the peak bandwidth can reach 13.44 GB·s^(-1)at an operating frequency of 120 MHz.This work can achieve 1.40,1.12,2.80 and 4.70 times the peak bandwidth compared with the existing work.

Formation Mechanism of Curved Martensite Structures in Cu-based Shape Memory Alloys

The curved martensite structures have been observed in CuZnAI-based shape memory alloys by both transmission electron microscope and optical microscope. It was found that the curved martensite structures observed in as-solution treated, as-aged and as-trained alloys usually occurred around dislocation tangles or precipitate, at the plate boundary or grain boundary, and when the growing plates collided with each other or alternate mutually.

Memory Behaviors Based on ITO/Graphene Oxide/Al Structure

We investigate the memory properties of the ITO/graphene oxide/Al diodes. It is found that the devices show different memory behaviors with the diverse geometry and thickness of Al. When the thickness of the Al electrode is relatively thick, the device of the cross-point Al electrode shows a three-level memory effect, and the counterpart device of the cross-bar Al electrode exhibits a volatile static random access memory effect. When the thickness of the AI electrode is thinner, the above devices demonstrate a flash memory effect. The different memory behaviors of ITO/GO/AI diodes are ascribed to the mode and degree of reduction and oxidation of GO.

STRUCTURAL HEALTH MONITORING IN COMPOSITE MATERIALS USING EMBEDDED SHAPE MEMORY ALLOY(SMA) WIRE SENSORS

Shape memory alloy （SMA） materials possess completely superelasticity or pseudoelasticity above the austenite finish temperature and many unique mechanical, thermal, thermal-mechanical and electrical properties compared with other conventional materials. Many studies have reported that the superelastic and hysteresis properties of the SMA materials can absorb energies coming from external excitations or sudden impacts. In addition, due to the special electrical properties of NiTi superelastic wires, they can also be used as strain-sensing element to monitor structural health conditions. Composite laminated specimens embedded with SMA wire sensors are fabricated and detailed testing system is designed such as multi-parameters measuring for impact and weak signal processing for SMA sensor. Low velocity impact test shows that SMA wire sensors embedded in fiber-reinforced plastic （FRP） laminate can be well used to monitor impact responses, such as the location of impact damage, impact degree, and strain distribution. Experimental results and theoretical predictions reveal almost the same. Comparing with other method, a simple, economic and reliable technique method monitoring important engineering structures on line is provided.

Nonlinear Finite Element Method Considering Martensite Plasticity for Shape Memory Alloy Structure

This work presents a nonlinear finite element method to simulate the macroscopic mechanical responses and the effects of martensite plasticity in a shape memory alloy(SMA)structure.A linear relationship formulation is adopted to express the influence of martensite plasticity on the inverse martensitic phase transition of SMA material.Incorporating with a trigonometric-type phase transition evolution law and an exponential-type plastic flow evolution law,an incremental mechanical model with two internal variables is supposed based on the macroscopic experimental phenomena.A nonlinear finite element equation is formulated and solved by the principle of virtual displacement and Newton-Raphson method respectively.By employing the proposed nonlinear finite element method,the uniform tensile bar and three-point bending beam are simulated and analyzed.Results illustrate that the presented nonlinear finite element method is suitable to act as an effective computational tool for the wide applications based on the SMA material considering the effects of martensite plasticity because all material constants related to the method can be obtained from macroscopic experiments.

A Unified Buffering Management with Set Divisible Cache for PCM Main Memory

This research proposes a phase-change memory （PCM） based main memory system with an effective combi- nation of a superblock-based adaptive buffering structure and its associated set divisible last-level cache （LLC）. To achieve high performance similar to that of dynamic random-access memory （DRAM） based main memory, the superblock-based adaptive buffer （SABU） is comprised of dual DRAM buffers, i.e., an aggressive superblock-based pre-fetching buffer （SBPB） and an adaptive sub-block reusing buffer （SBRB）, and a set divisible LLC based on a cache space optimization scheme. According to our experiment, the longer PCM access latency can typically be hidden using our proposed SABU, which can significantly reduce the number of writes over the PCM main memory by 26.44%. The SABU approach can reduce PCM access latency up to 0.43 times, compared with conventional DRAM main memory. Meanwhile, the average memory energy consumption can be reduced by 19.7%.

A fuzzy PID-controlled SMA actuator for a two-DOF joint

Shape memory alloy （SMA） actuator is a potential advanced component for servo- systems of aerospace vehicles and aircraft. This paper presents a joint with two degrees of freedom （DOF） and a mobility range close to ±60° when driven by SMA triple wires. The fuzzy proportional-integral-derivative （PID）-controlled actuator drive was designed using antagonistic SMA triple wires, and the resistance feedback signal made a closed loop. Experiments showed that, with the driving responding frequency increasing, the overstress became harder to be avoided at the position under the maximum friction force. Furthermore, the hysteresis gap between the heating and cooling paths of the strain-to-resistance curve expanded under this condition. A fuzzy logic control was considered as a solution, and the curves of the wires were then modeled by fitting polynomials so that the measured resistance was used directly to determine the control signal. Accurate control was demonstrated through the step response, and the experimental results showed that under the fuzzy PID-control program, the mean absolute error （MAE） of the rotation angle was about 3.147°. In addition, the investigation of the external interference to the system proved the controllable maximum output.

4D Printing:History and Recent Progress

4D printing has attracted great interest since the concept was introduced in 2012. The past 5 years have witnessed rapid advances in both 4D printing processes and materials. Unlike 3D printing, 4D printing allows the printed part to change its shape and function with time in response to change in external conditions such as temperature, light, electricity, and water. In this review, we first overview the history of 4D printing and discuss its definition. We then summarize recent technological advances in 4D printing with focuses on methods, materials, and their intrinsic links. Finally, we discuss potential applications and offer perspectives for this exciting new field.

Can multi-modal neuroimaging evidence from hippocampus provide biomarkers for the progression of amnestic mild cognitive impairment?

Impaired structure and function of the hippocampus is a valuable predictor of progression from amnestic mild cognitive impairment（a MCI） to Alzheimer＇s disease（AD）. As a part of the medial temporal lobe memory system,the hippocampus is one of the brain regions affected earliest by AD neuropathology,and shows progressive degeneration as a MCI progresses to AD. Currently,no validated biomarkers can precisely predict the conversion from a MCI to AD. Therefore,there is a great need of sensitive tools for the early detection of AD progression. In this review,we summarize the specifi c structural and functional changes in the hippocampus from recent a MCI studies using neurophysiological and neuroimaging data. We suggest that a combination of advanced multi-modal neuroimaging measures in discovering biomarkers will provide more precise and sensitive measures of hippocampal changes than using only one of them. These will potentially affect early diagnosis and disease-modifying treatments. We propose a new sequential and progressive framework in which the impairment spreads from the integrity of fibers to volume and then to function in hippocampal subregions. Meanwhile,this is likely to be accompanied by progressive impairment of behavioral and neuropsychological performance in the progression of a MCI to AD.