检索结果-南通市图书馆

Effect of Temperature and Ram Speed on Isothermal Extrusion for Large-size Tube with Piece-wing

Journal of Materials Science & Technology 2005年第4期21卷 499-504页

作者： He YANG Jun ZhanG Yangmin HE bingtao han College of Materials Science and Engineering Northwestern Polytechnical University Xi'an 710072 China Xi'an Heavy Machinery Research Institute Xi'an 710032 China

Heat energy change during the extrusion of 7075 aluminium alloy large-size tube with piece-wing in a container was analyzed. Extrusion load vs ram displacement diagrams and exit temperature vs ram displacement diagrams at various speeds were obtained by 3D FEM simulation. Results show that the exit temperature becomes higher as the ram speed and displacement increase. For large-size tube with piece-wing, there is certainly a curve of ram speed decreasing with increasing ram displacement, which enables isothermal extrusion to be achieved. Therefore,an attempt was made to divide the working stroke into five different zones. Each of them has a preset speed that decreases from the ram displacement beginning to the ending. And then, new exit temperature vs ram displacement diagram was obtained by 3D FEM simulation for the five different speeds. It is shown that the variation of exit temperature is very small. Through the above research, a basic method for realizing isothermal extrusion of 7075 large-size tube with piece-wing was obtained, that is, the working stroke was divided into several different zones with a decreasing speed during extrusion, each zonest speed was real-time adjusted on the feedback signal of exit temperature by proportional hydraulic valve through closed-loop control. The engineering experiment verification was carried out on 100 MN aluminium extrusion press with oil-driven double action. The experimental results of the exit temperature agrees with the simulation ones. The achievements of this study may serve as a significant guide to the practice of the relevant processes, particularly for isothermal extrusion. The verified method has been used in the design and manufacture of 125 MN aluminium extrusion press with oil-driven double action.

关键词： Extrusion Large-size aluminium profile FEM Temperature Unequal plastic deformation

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

KEY TECHNIQUES IN R&D OF OUT-SIZE EXTRUSION PRESS WITH OIL-DRIVEN DOUBLE ACTION

引用

Chinese Journal of Mechanical Engineering 2007年第2期20卷 36-40页

作者： ZhanG Jun YANG He XIE Donggang han bingtao College of Materials Science and Engineering Northwestern Polytechnical University Xi＇ an 710072 China China Heavy Machinery Research Institute Xi＇ an 710032 China College of Materials Science and Engineering Northwestern PolytechnJcal University Xi＇ an 710072 China

A modem design method, in which traditional design formulas are conjoined with numerical simulation and optimization, is successfully used to design the out-size extrusion flame precisely so that the press cost can be saved. A new technology used for decompressing by a multi-steps dynamical mode is put forward, which makes it possible to decompress the large flow-volume high-pressure oil in the main cylinders. In addition, a method for realizing the fixed mandrel process by hydraulic support is proposed and its control equation is established. Pre-tightening frame tests are carried out by over-operating pressure on 100 MN aluminium extrusion press with oil-driven double action, which is developed based on the above key techniques and is the largest press so far in the world, and the results show that the frame structure designed is reasonable and reliable, and the modem design method used is an useful tool for designing large and out-size heavy plastic forming machinery. The results of decompressing curve in main cylinder and noise inspection indicate that multi-steps dynamical mode for decompressing the large flow-volume high pressure oil is valid and reliable. Meanwhile, the fixed mandrel process is well realized based on the control equation. These key techniques have been used in the development of 125 MN aluminium extrusion press with oil-driven double action.

关键词： Extrusion Out-size press Aluminum profile FEM numerical simulation

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

A Unified Deep Learning Method for CSI Feedback in Massive MIMO Systems

引用

ZTE Communications 2022年第4期20卷 110-115页

作者： GAO Zhengguang LI Lun WU Hao TU Xuezhen han bingtao State Key Laboratory of Mobile Network and Mobile Multimedia Technology Shenzhen 518055China The College of Computer Science and Technology Nanjing University of Aeronautics and AstronauticsNanjing 210016China

A unified deep learning(DL)based algorithm is proposed for channel state information(CSI)compression in massive multipleinput multipleoutput(MIMO)*** importantly,the element filling strategy is investigated to address the problem of model redesign⁃ing and retraining for different antenna typologies in practical *** results show that the proposed DL-based algorithm achieves better performance than the enhanced TypeⅡalgorithm in Release 16 of *** proposed element filling strategy enables onetime training of a unified model to compress and reconstruct different channel state matrices in a practical MIMO system.

关键词： deep learning channel state information element filling strategy

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

Neursafe-FL:A Reliable,Efficient,Easy-to-Use Federated Learning Framework

引用

ZTE Communications 2022年第3期20卷 43-53页

作者： TANG Bo ZhanG Chengming WANG Kewen GAO Zhengguang han bingtao ZTE Corporation Shenzhen 518057China The State Key Laboratory of Mobile Network and Mobile Multimedia Technology Shenzhen 518055China

Federated learning(FL) has developed rapidly in recent years as a privacy-preserving machine learning method,and it has been gradually applied to key areas involving privacy and security such as finance,medical care,and government ***,the current solutions to FL rarely consider the problem of migration from centralized learning to federated learning,resulting in a high practical threshold for federated learning and low ***,we introduce a reliable,efficient,and easy-to-use federated learning framework named *** on the unified application program interface(API),the framework is not only compatible with mainstream machine learning frameworks,such as Tensorflow and Pytorch,but also supports further extensions,which can preserve the programming style of the original framework to lower the threshold of *** the same time,the design of componentization,modularization,and standardized interface makes the framework highly extensible,which meets the needs of customized requirements and FL evolution in the ***-FL is already on Github as an open-source project^(1).

关键词： federated learning privacy-preserving Neursafe-FL

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

大模型关键技术与应用

引用

中兴通讯技术 2024年第2期30卷 76-88页

作者：韩炳涛刘涛中兴通讯股份有限公司深圳518057 移动网络和移动多媒体技术国家重点实验室深圳518055

介绍了自ChatGPT发布以来,大模型关键技术和应用的主要进展。在大模型设计方面,模型规模不断增加,但已有放缓趋势,更长的上下文以及多模态已经成为主流,计算效率明显提升;在模型训练方面,从单纯追求数据数量逐渐转变为关注数据的多样性... 详细信息

介绍了自ChatGPT发布以来,大模型关键技术和应用的主要进展。在大模型设计方面,模型规模不断增加,但已有放缓趋势,更长的上下文以及多模态已经成为主流,计算效率明显提升;在模型训练方面,从单纯追求数据数量逐渐转变为关注数据的多样性和质量,特别是如何使用合成数据训练大模型成为主流探索方向,这是迈向通用人工智能(AGI)的关键;在模型推理方面,模型量化和推理引擎优化极大降低了模型使用成本,诸如投机采样等新兴算法逐渐成熟。在应用层,Agent技术获得了重大进展,在克服大模型固有缺陷方面发挥了不可替代的作用。越来越多的企业开始规划、研发以及使用大模型,企业级大模型应用架构日益成熟完善,并以场景、技术、算法三要素为抓手加速大模型商业价值闭环。

关键词：大模型模型训练推理加速大模型安全智能体

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

现代挤压装备的发展

引用

中国材料进展 2013年第5期32卷 264-268,282页

作者：谢东钢陈蕴博郑文达韩炳涛金属挤压与锻造装备技术国家重点实验室陕西西安710032 机械科学研究总院先进制造技术研究中心先进成形技术与装备国家重点实验室北京100083

简要介绍了挤压装备的发展历程、挤压产品的特点及其在国防工业和国民经济各行各业中的应用;阐述了我国挤压装备迅速发展的现状。尽管我国近年来挤压装备发展迅速,已经成为世界挤压大国,挤压装备数量和挤压力位列世界前茅,然而我们还不... 详细信息

简要介绍了挤压装备的发展历程、挤压产品的特点及其在国防工业和国民经济各行各业中的应用;阐述了我国挤压装备迅速发展的现状。尽管我国近年来挤压装备发展迅速,已经成为世界挤压大国,挤压装备数量和挤压力位列世界前茅,然而我们还不是挤压强国。与发达国家的挤压装备技术比较,挤压装备的前沿技术、挤压工艺、模具设计理论和方法、自动控制、模具结构创新和现代化精密加工、产品质量乃至挤压装备的数量以及规格和类型的分布诸方面依然落后。随着科学技术的发展与进步,人类生产活动的不断延伸,对于挤压制品的要求越来越高,因此挤压机的大型化、智能化、可靠性、设备生产能力极限化以及超高压油泵直接控制是未来挤压装备的发展趋势。

关键词：挤压装备智能化可靠性超高压油泵直接控制

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

深度学习的10年回顾与展望

引用

中兴通讯技术 2022年第6期28卷 75-84页

作者：韩炳涛刘涛唐波中兴通讯股份有限公司数据智能平台中心中国深圳518057 中兴通讯股份有限公司中国深圳518057 移动网络和移动多媒体技术国家重点实验室中国深圳518055

过去10年深度学习在算法、算力、数据方面获得了长足发展,使人工智能(AI)技术突破商用限制,行业应用场景日益广泛,产业规模持续扩大。在基础模型方面出现了卷积、注意力机制等关键突破;在学习方法方面,强化学习、自监督学习、大模型并... 详细信息

过去10年深度学习在算法、算力、数据方面获得了长足发展,使人工智能(AI)技术突破商用限制,行业应用场景日益广泛,产业规模持续扩大。在基础模型方面出现了卷积、注意力机制等关键突破;在学习方法方面,强化学习、自监督学习、大模型并行训练等使模型学习能力大大加强。新型AI计算芯片不断涌现,使计算能效提升百倍。未来10年,深度学习若要保持可持续的指数增长态势,绿色、高效、安全将成为新的核心要素。空间计算、近似计算等技术有望使AI芯片效能继续获得百倍提升。一系列生态融合工具的出现将解决目前日趋严峻的生态碎片化问题。AI安全、可信将成为AI技术应用的基本要求。

关键词：深度学习 AI芯片推理加速可信AI 开源

来源：

维普期刊数据库

同方期刊数据库评论