下载文档原格式
Robust ControlRobust control is a critical concept in the field of engineering and technology, particularly in the design and implementation of control systems for various applications. It refers to the ability of a control system to maintain stability and performance in the face of uncertainties and variations in the system and its environment. This is achieved through the use of advanced control techniques and algorithms that can adapt to changing conditions and disturbances, ensuring reliable and effective operation. From a technical perspective, robust control involves the use of mathematical models, analysis tools, and optimization methods to design control systems that can handle uncertainties and variations. This may include the use of robust control theory, which provides a framework for analyzing and designing control systems that are resilient to disturbances and uncertainties. It also involves the use of advanced control algorithms, such as robust controllers, adaptive controllers, and predictive controllers, which can adjust their parameters and behavior to maintain stability and performance in the presence of uncertainties. One of the key challenges in robust control is dealing with uncertainty in the system and its environment. This uncertainty may arise from various sources, such as variations in the system parameters, external disturbances, and measurement errors. Designing control systems that can effectively handle these uncertainties requires a deep understanding of the system dynamics, as well as the ability to model and analyze the impact of uncertainties on the system performance. Another important aspect of robust control is the trade-off between performance and robustness. In many cases, increasing the robustness of a control system may come at the cost of reduced performance, and vice versa. Balancing these trade-offs requires careful consideration of the specific application and the desired performance requirements, as well as the available resources and constraints. From a practical perspective, robust control has numerous real-world applications across various industries, including aerospace, automotive, manufacturing, and robotics. For example, in aerospace applications, robust control is essential for ensuring the stability and performance of aircraft and spacecraft in the presence of uncertainties and disturbances. Similarly, in automotive applications, robust control is used todesign advanced driver assistance systems and autonomous vehicles that can operate safely and effectively in dynamic and uncertain environments. In conclusion, robust control is a critical concept in the field of engineering and technology, with wide-ranging applications and implications. It involves the use of advanced control techniques and algorithms to design control systems that can maintain stability and performance in the face of uncertainties and variations. Addressing the challenges of uncertainty and trade-offs between performance and robustness is essential for the successful design and implementation of robust control systems in real-world applications.。
China's Ascent in Artificial IntelligenceIn the rapidly evolving landscape of technology, Artificial Intelligence (AI) has emerged as a pivotal force driving innovation and transformation across industries. Among the global players, China has emerged as a formidable contender, making significant strides in the AI domain, showcasing its potential to become a global leader. Advancements in Research and DevelopmentChina's commitment to AI research and development is evident from the substantial investments made by both the government and private enterprises. The country has established numerous research centers, universities, and incubators dedicated to AI technologies, fostering an environment conducive to innovation. Initiatives like the "New Generation Artificial Intelligence Development Plan" by the Chinese government aim to accelerate the country's AI capabilities, targeting breakthroughs in key technologies such as machine learning, computer vision, natural language processing, and robotics. Industrial ApplicationsChina's AI advancements are not limited to theoretical research but have found widespread applications in various industries. In manufacturing, AI is revolutionizing production processes, enhancing efficiency, and reducing costs through smart factories and predictive maintenance. The retail sector is leveraging AI for personalized recommendations, inventory management, and customer service automation. Healthcare is another area where AI is making a significant impact, from precision medicine and disease prediction to robotic surgeries and telemedicine. Additionally, China's transportation system is being modernized with AI-powered autonomous vehicles, intelligent traffic management systems, and smart cities initiatives.Data AdvantageOne of the key factors contributing to China's success in AI is its vast trove of data. As the world's most populous country with a booming digital economy, China generates an enormous amount of data daily, providing a fertile ground for AI algorithms to learnand improve. The government's support for data sharing and the willingness of private companies to collaborate have further accelerated the pace of AI innovation. Challenges and OpportunitiesWhile China's progress in AI is impressive, it also faces several challenges. Concerns over data privacy and security have been raised, necessitating stricter regulations and ethical guidelines. The country also needs to address the skills gap in AI talent, ensuring a continuous supply of highly skilled professionals to sustain its growth trajectory.However, these challenges also present opportunities for China to emerge as a global standard-setter in AI governance and ethics. By developing robust regulatory frameworks and fostering international cooperation, China can demonstrate its commitment to responsible AI development, enhancing its reputation and influence in the global AI ecosystem.ConclusionIn conclusion, China's ascendance in Artificial Intelligence is a testament to its commitment to technological innovation and economic transformation. With significant investments, a robust research ecosystem, and a vast data resource, China is poised to play a leading role in shaping the future of AI. As it continues to navigate the challenges and seize the opportunities ahead, China's journey in AI will undoubtedly be one to watch closely.。
AI at Work: It’s Time to Embrace AI in Higher EducationHR has the opportunity to be the agentof change in the next wave of IT consumerization for higher ed.We’ve seen this movie before. Only a decade ago, colleges and universities were slowto perceive the value of mobile and social technologies and discouraged their use in the workplace. Over time, however, higher education realized that they couldn’t fight the BYOD trend and they moved to enable their employees with the same technology they were already using in their personal lives. Today we’re on a similar path with artificial intelligence (AI). HR has an opportunity to get ahead of the curve and be the agent of change in the next wave of IT consumerization.When speaking of AI, we’re talking about enabling machines and their software to sense, comprehend, act, and learn. AI is meant to enhance human cognitive performance and is creating entirely new job categories. Through conversational interfaces that include Siri, Alexa, and other personalized chatbots, AI makes it possible for humans to interact with software in a more human way. Predictive analytics fed by sizable pools of data play an important role in this. Underpinning chatbots and other software interfaces is a great deal of technology and data that has countless applications and implications for HR and higher ed.When speaking of AI, we’re talking aboutenabling machines and their software tosense, comprehend, act, and learn.The data already in your systems—on your schedules, your applicants, your candidates and employees—can be mined for insights and put to work automating many of HR’s repetitive and mundane duties. AI processes massive volumes of data much faster and more accurately than humans, and is continuously improving. Automating your rote tasks—think application filtering, payroll and benefits queries, onboarding, compliance, and performance monitoring—can lead to significant increases in productivity. It can also position HR for bigger things such as career development, succession management, and workforce planning.Formulating a successful talent strategy in today’s competitive education market is made easier when AI provides HR with valuable insights for common challenges like managing the balance between adjunct and full-time faculty, succession planning as university leaders retire, and finding new ways for faculty and staff to support student success. Applying financial data on top of the HR data ensures that managers and leaders are supporting and pursuing the overall institutional strategy and objectives.Not embracing AI now will result in job loss, irrelevance, and loss of competitive advantage:• Respondents identified reduced productivity, skillset obsolescence, and job loss as the top three consequences of failing to embrace AI in the workforce• Respondents believe embracing AI will have the most positive impact on directors and leaders• Failing to empower leadership teams with AI could make higher education institutions less competitive AI as a new source of growth —including job growth.Jobs will change, new roles will be created, and continuous learning will become even more of a priority. Each of these findings is included in the research, along with the findings that AI will lead to better analysis and faster decision making. Employees believe this to be true from their exposure to AI in their personal lives. One headline to emerge from the study is that 93 percent of people would take directions from a robot at work. People have come to trust the quality of information AI delivers and to value the time savings it offers. AI’s ability to tailor itself to personal preferences and past behaviors is another plus.Of the threat that AI presumably poses to future employment, Erik Brynjolfsson and Andrew McAfee write in the Harvard Business Review that “Over the next decade, AI won’t replace managers, but managers who use AI will replace those who don’t.” Couple this with a belief on the part of many experts that AI will lead not to job loss but to job growth instead, and you begin to see both the urgency and the opportunity.The wrench in the works is the fact that only 6 percent of HR leaders are actively deploying AI in their departments. This supports a general view that HR is not only underinvesting in AI, but in people systems overall.This is a missed opportunity according to a Forbes article that posits, “While AI should allow colleges and universities to become more efficient and effective in supplying higher education, it will also shift the demand curve for postsecondary education to the right. With greater demand for cognitive and technical skills, colleges and universities will have a golden opportunity to reassert their preeminence in human capital development.”HR leaders have an opportunity to differentiate themselves and their organizations from others by integrating AI technology into their operations and training their workforce to use it. It may even create more demand for higher education. Bridging the gap between employee expectations and current AI reality is among HR’s highest priorities. HR can’t afford to be caught flat-footed again.70% of employees use AI in theirpersonal lives—for entertainment,ridesharing, personal finance, andpersonal relationships—yet only 24%use AI at work.In Forbes in March 2018, Oracle’s head of HCM product strategy, Gretchen Alacorn says that “In terms of the time you spend recruiting, the time you spend trying to find ‘the right fit,’ the amount of time you spend trying to rely on a gut feeling...anything we can do to help guide you, that’s huge value. What if such an [AI] application could help a 50,000-employee organization with an annual turnover rate of 4% reduce that turnover by just 1 percentage point? What else could you do if you didn’t have to find an additional 500 people?”HR bridges all aspects of the employee lifecycle,so success with AI in recruiting has implications for career learning and development, career pathing, payroll and benefits, and succession planning. Items such as workforce flight risk prediction become more important. Flight risk prediction draws on more than 140 different attributes and behaviors in order to formulate its conclusions. The attributes include employee sentiment, an employee’s mentors and influences, their number of years in a position, how long they’ve been reporting to their current manager, their potential career path, their salary history, and whether and when they last received a raise. These all factor into a predicted attrition rate and offer employers a number of useful cues and clues on howCopyright © 2018, Oracle and/or its affiliates. All rights reserved. This document is provided for information purposes only, and thecontents hereof are subject to change without notice. This document is not warranted to be error-free, nor subject to any other warranties or conditions, whether expressed orally or implied in law, including implied warranties and conditions of merchantability or fitness for a particular purpose. We specifically disclaim any liability with respect to this document, and no contractual obligations are formed either directly or indirectly by this document. This document may not be reproduced or transmitted in any form or by any means, electronic or mechanical, for any purpose, without our prior written permission.Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners.Intel and Intel Xeon are trademarks or registered trademarks of Intel Corporation. All SPARC trademarks are used under licenseand are trademarks or registered trademarks of SPARC International, Inc. AMD, Opteron, the AMD logo, and the AMD Opteron logo are trademarks or registered trademarks of Advanced Micro Devices. UNIX is a registered trademark of The Open Group.ORACLE CORPORATIONWorldwide Headquarters500 Oracle Parkway, Redwood Shores, CA 94065 USAWorldwide InquiriesTELE+ 1.650.506.7000 + 1.800.ORACLE1FA X + 1.650.506.7200 /higher-educationCONNECT WITH US/oracleHCM/oracle/company/oracle/oracleHCM。
人工智能发展现状和趋势英语作文The Current State and Future Trends of Artificial Intelligence.Artificial intelligence (AI) has come a long way since its inception in the mid-20th century. It has progressed from being a mere concept to a technology that is now integral to our daily lives, shaping various industries and sectors. The rapid advancements in computing power, data availability, and algorithms have fueled the growth of AI, making it possible to create machines that can learn, reason, and make decisions independently.Current State of Artificial Intelligence.Currently, AI is being used in various fields, such as healthcare, finance, transportation, and entertainment. In healthcare, AI algorithms are being trained to diagnose diseases, predict patient outcomes, and assist in surgical procedures. In finance, AI is being used for frauddetection, credit scoring, and investment decisions. In transportation, AI-powered autonomous vehicles are being tested on roads, while in entertainment, AI is being used to create music, art, and movies.AI is also being used to improve the efficiency and productivity of businesses. Automation of repetitive tasks, prediction of customer behavior, and optimization of supply chains are some of the ways AI is revolutionizing the business world. Additionally, AI-powered chatbots are providing customer support, handling inquiries, and resolving issues efficiently.Challenges Facing Artificial Intelligence.Despite its widespread adoption and usage, AI faces several challenges. One of the major challenges is the lack of interpretability. Most AI models are complex anddifficult to understand, making it challenging to explain their decisions to non-experts. This lack of transparency can lead to mistrust and skepticism among users.Another challenge is the ethical implications of AI. As AI systems become more autonomous, there are concerns about their potential to make unfair decisions or perpetuate biases. There is a need for robust ethical frameworks and regulations to ensure AI systems are developed and used responsibly.Future Trends of Artificial Intelligence.The future of AI looks bright, with exciting new trends and developments emerging. Here are some of the key trends that are likely to shape the AI landscape in the coming years:1. Enhanced Interpretability.One of the key areas of focus for AI research is enhancing the interpretability of AI models. Scientists are working on developing new algorithms and techniques that can make AI models more transparent, enabling users to understand how they make decisions. This could lead to increased trust in AI systems and wider adoption acrossvarious industries.2. Autonomous AI Systems.Autonomous AI systems that can operate independently without human intervention are becoming a reality. These systems are being developed to handle complex tasks, suchas driving, flying, and managing supply chains. However,the development of such systems raises concerns about their safety and reliability, necessitating rigorous testing and validation before deployment.3. Hyper-personalization.With the increasing availability of data, AI systemsare becoming increasingly capable of understanding and predicting individual preferences and behaviors. This trend, known as hyper-personalization, is likely to reshapevarious industries, such as retail, media, and healthcare. By leveraging AI, businesses can offer personalized experiences to their customers, enhancing satisfaction and loyalty.4. AI-powered Environments.The concept of AI-powered environments, where AI systems are integrated into our physical surroundings, is becoming a reality. This trend is likely to be driven by the development of AI-enabled sensors, actuators, and other devices that can communicate with each other and make intelligent decisions. For instance, smart homes that can automatically adjust lighting, temperature, and security settings based on occupant preferences and behaviors are becoming increasingly common.5. Collaborative AI.Collaborative AI refers to the integration of AI systems with humans to enhance their capabilities and efficiency. This trend is likely to be driven by the development of AI systems that can understand and interpret human language, gestures, and emotions. By leveraging AI, humans can work more efficiently and productively, achieving better outcomes in various fields, such ashealthcare, education, and research.In conclusion, the future of AI looks bright, with exciting new trends and developments emerging. As AI systems become more autonomous, intelligent, and integrated into our daily lives, they are likely to shape various industries and sectors, leading to improved efficiency, productivity, and quality of life. However, it is crucial to address the challenges facing AI, such asinterpretability and ethics, to ensure its responsible and sustainable development.。
生物信息学,英文BioinformaticsBioinformatics is a rapidly growing field that combines biology, computer science, and information technology to analyze and interpret biological data. It has become an essential tool in modern scientific research, particularly in the fields of genomics, proteomics, and molecular biology. The advent of high-throughput sequencing technologies has led to an exponential increase in the amount of biological data available, and bioinformatics provides the means to manage, analyze, and interpret this vast amount of information.At its core, bioinformatics involves the use of computational methods and algorithms to store, retrieve, organize, analyze, and interpret biological data. This includes tasks such as DNA and protein sequence analysis, gene and protein structure prediction, and the identification of functional relationships between different biological molecules. Bioinformatics also plays a crucial role in the development of new drugs, the understanding of disease mechanisms, and the study of evolutionary processes.One of the primary applications of bioinformatics is in the field ofgenomics, where it is used to analyze and interpret DNA sequence data. Researchers can use bioinformatics tools to identify genes, predict their function, and study the genetic variation within and between species. This information is essential for understanding the genetic basis of diseases, developing personalized medicine, and exploring the evolutionary history of different organisms.In addition to genomics, bioinformatics is also widely used in proteomics, the study of the structure and function of proteins. Bioinformatics tools can be used to predict the three-dimensional structure of proteins, identify protein-protein interactions, and study the role of proteins in biological processes. This information is crucial for understanding the mechanisms of disease, developing new drugs, and engineering proteins for industrial and medical applications.Another important application of bioinformatics is in the field of systems biology, which aims to understand the complex interactions and relationships between different biological components within a living organism. By using computational models and simulations, bioinformaticians can study how these components work together to produce the observed behavior of a biological system. This knowledge can be used to develop new treatments for diseases, optimize agricultural practices, and gain a deeper understanding of the fundamental principles of life.Bioinformatics is also playing a crucial role in the field of personalized medicine, where the goal is to tailor medical treatments to the unique genetic and molecular profile of each individual patient. By using bioinformatics tools to analyze a patient's genetic and molecular data, healthcare providers can identify the most effective treatments and predict the likelihood of adverse reactions to certain drugs. This approach has the potential to improve patient outcomes, reduce healthcare costs, and pave the way for more targeted and effective therapies.In addition to its scientific applications, bioinformatics is also having a significant impact on various industries, such as agriculture, environmental science, and forensics. In agriculture, bioinformatics is used to improve crop yields, develop new pest-resistant varieties, and optimize the use of resources. In environmental science, bioinformatics is used to study the impact of human activities on ecosystems, identify new sources of renewable energy, and monitor the health of the planet. In forensics, bioinformatics is used to analyze DNA evidence and identify individuals involved in criminal activities.Despite the many benefits of bioinformatics, there are also significant challenges and ethical considerations that need to be addressed. As the amount of biological data continues to grow, there is an increasing need for efficient data storage, management, andsecurity systems. Additionally, the interpretation of biological data requires a deep understanding of both biological and computational principles, which can be a barrier for some researchers and healthcare providers.Furthermore, the use of bioinformatics in areas such as personalized medicine and forensics raises important ethical questions regarding privacy, informed consent, and the potential for discrimination based on genetic information. It is crucial that the development and application of bioinformatics technologies be guided by strong ethical principles and robust regulatory frameworks to ensure that they are used in a responsible and equitable manner.In conclusion, bioinformatics is a rapidly evolving field that is transforming the way we understand and interact with the biological world. By combining the power of computer science and information technology with the insights of biology, bioinformatics is enabling new discoveries, improving human health, and shaping the future of scientific research. As the field continues to grow and evolve, it will be important for researchers, policymakers, and the public to work together to ensure that the benefits of bioinformatics are realized in a responsible and ethical manner.。
随机森林构建方法英语作文Random Forest Construction Method。
Random Forest is a popular machine learning algorithm that is used to solve a wide range of problems, including classification and regression. It is a type of ensemble learning method that combines multiple decision trees to produce a more accurate and robust model. In this article, we will discuss the construction method of Random Forest.Step 1: Data Preparation。
The first step in building a Random Forest model is to prepare the data. This involves cleaning the data, removing any missing values, and transforming the data into a suitable format for the algorithm. The data should be split into a training set and a testing set, with the training set used to train the model and the testing set used to evaluate its performance.Step 2: Random Sampling。
Random Forest uses a technique called bagging, which involves randomly sampling the data with replacement to create multiple subsets of the data. Each subset is used to train a decision tree, and the results are combined to produce the final model. The number of subsets is determined by the user and is typically set to a value between 100 and 1000.Step 3: Decision Tree Construction。
2021年第5期（总第49卷第363期) No. 5 in 2021 (Total Vol. 49,No. 363 >建筑节能（中英文）Journal of BEE■暖通空调HV&ACdoi ： 10.3969/j. issn.2096-9422.2021.05.010基于APC智能控制技术在医院中央空调节能中的应用付磊、罗淋俊2，刘浩2，蔡跃峰1，粘培坤、褚丹雷2(1.厦门大学附属心血管病医院，厦门361006;2.厦门奥普拓自控科技有限公司，厦门361026)摘要：提出了基于高端控制优化算法的中央空调智能化整体解决方案，并详细介绍了方案的系统建 模、控制器设计过程，通过厦门大学附属心血管病医院为研究对象，介绍了该方案的实现与应用，通过验证证明该方案抗干扰性强、控制精度高、非线性处理能力强、鲁棒性能好，能够在提升中央空调系统整体智能化水平的同时实现节能减排。
该技术为中央空调智能化运营提供了一个新的途径。
随着工业互联网技术与数据算法的不断更新，基于A PC多变量控制优化技术的中央空调整体解决方案将得到更为广泛的工程应用。
关键词：先进过程控制；多变量控制；模型预测控制；智能解决方案；实时优化中图分类号：TU83 文献标志码：A文章编号：2096-9422(2021) 05 ■005447Intelligent Solution Scheme of Central Air-conditioning System Based on APCControl Technique in HospitalFU Lei' , LUO Lin-jun2, LIU Hao2, CM Yue-feng', NIAN Pei-kun, CHU Dan-lei2(1. Xiamen Cardiovascular Hospital of Xiamen University,Xiamen 361006, Fujian,China；2.Optimal Process Control Technologies Co.,Ltd.,Xiamen 361026, Fujian,China)Abstract ：This paper proposes an intelligent total solution scheme to central air-conditioning based on advanced process control optimization algorithms, introduces the system modelling of the scheme and the design process of controller in detail, and demonstrates the realization and application of the scheme via researching Xiamen Cardiovascular Hospital of Xiamen University. Through verification, the scheme achieves excellent anti-interference performance, high control accuracy, high non-linear processing capacity, and outstanding robust performance, capable of enhancing the overall intelligentization of central air-conditioning system while accomplishing energy conservation and emission reduction. This technology provides a new way for intelligent operation of central air conditioning. Intelligent solution based on APC control technique will be more widely used in engineering applications.Keywords ：advanced process control；multi-variable control technique；model predictive control；intelligence solution scheme；real-time optimization〇引言中央空调系统作为近代建筑物温湿度调节的核 心运维系统之一，被广泛应用于星级酒店、大型购物 中心、医院学校、博物馆、无尘车间等，随着人们对舒 适型要求的提高，特别是医院中央空调运行时间长，能耗较大，高峰时占医院总能耗的soy^eo%11~6]。
7种回归⽅法！请务必掌握！7 种回归⽅法！请务必掌握！线性回归和逻辑回归通常是⼈们学习预测模型的第⼀个算法。
由于这⼆者的知名度很⼤，许多分析⼈员以为它们就是回归的唯⼀形式了。
⽽了解更多的学者会知道它们是所有回归模型的主要两种形式。
事实是有很多种回归形式，每种回归都有其特定的适⽤场合。
在这篇⽂章中，我将以简单的形式介绍 7 中最常见的回归模型。
通过这篇⽂章，我希望能够帮助⼤家对回归有更⼴泛和全⾯的认识，⽽不是仅仅知道使⽤线性回归和逻辑回归来解决实际问题。
本⽂将主要介绍以下⼏个⽅⾯：1. 什么是回归分析？2. 为什么使⽤回归分析？3. 有哪些回归类型？线性回归（Linear Regression）逻辑回归（Logistic Regression）多项式回归（Polynomial Regression）逐步回归（Stepwise Regression）岭回归（Ridge Regression）套索回归（Lasso Regression）弹性回归（ElasticNet Regression）4. 如何选择合适的回归模型？1什么是回归分析？回归分析是⼀种预测建模技术的⽅法，研究因变量（⽬标）和⾃变量（预测器）之前的关系。
这⼀技术被⽤在预测、时间序列模型和寻找变量之间因果关系。
例如研究驾驶员鲁莽驾驶与交通事故发⽣频率之间的关系，就可以通过回归分析来解决。
回归分析是进⾏数据建模、分析的重要⼯具。
下⾯这张图反映的是使⽤⼀条曲线来拟合离散数据点。
其中，所有离散数据点与拟合曲线对应位置的差值之和是被最⼩化了的，更多细节我们会慢慢介绍。
2为什么使⽤回归分析？如上⾯所说，回归分析能估计两个或者多个变量之间的关系。
下⾯我们通过⼀个简单的例⼦来理解：⽐如说，你想根据当前的经济状况来估计⼀家公司的销售额增长。
你有最近的公司数据，数据表明销售增长⼤约是经济增长的 2.5 倍。
利⽤这种洞察⼒，我们就可以根据当前和过去的信息预测公司未来的销售情况。
1671 Application of robust prediction for a laser–GMA hybrid welding process and parameter optimization of6061-T6 aluminium alloyI Jang and K Y Lee*School of Mechanical Engineering,Yonsei University,Seoul,Republic of KoreaThe manuscript was received on15January2010and was accepted after revision for publication on3March2010. DOI:10.1243/09544054JEM1986Abstract:A methodology to predict and optimize the ultimate tensile strength of6061-T6aluminium(Al)alloy under a laser–GMA hybrid welding process is presented.Experimentswere performed by changing welding process parameters.For the evaluation of weldability,the ultimate tensile strength in the weld zone was considered to be the most important cri-terion.In order to determine the optimum welding parameters,a robust regression analysiswas performed to present a reliable prediction and a genetic algorithm was used.Based on theresults of robust regression analysis and the genetic algorithm,the optimal process parameterswere obtained.The ultimate tensile strength result obtained by the optimal parameters wasmuch improved,when compared with that specified by the American Welding Society.Keywords:Nd:YAG laser–GMA hybrid welding,robust regression analysis,genetic algorithm,least trimmed squares,Cook’s distance1INTRODUCTIONOwing to a low specific density and good corro-sion resistance to the atmosphere,water,oil,and many chemicals,aluminium has been increasingly employed in many important manufacturing areas, such as aeronautics,military,and transportation industries.6061-T6aluminium alloy is one of the most widely used structural alloys in the6000series for the transportation industries and is the most versatile of the heat-treatable alloys.Alloy6061-T6is easily welded and joined by various welding techniques.Hybrid welding attracts con-siderable attention because of its many advantages, such as increased penetration,improved root opening tolerance,misalignment tolerance,increased welding speed,enhanced process stability,and a fast travel-ling speed.Many researchers have investigated the possibilities relating to hybrid welding[1–6].Subse-quently,numerous research groups have started to *Corresponding author:School of Mechanical Engineering, Yonsei University,Stress Analysis and Failure Design Labora-tory,Seoul120-749,Republic of Korea.email:KYL2813@yonsei.ac.kr concentrate on further opportunities and limitations for the use of hybrid welding.In hybrid welding,the weld quality is influenced by laser power,laser focus,travelling speed,welding wire feed rate,root opening tolerance,and voltage[7].As these factors are related to each other in a complex manner,it is very difficult to obtain the optimal weld-ing conditions.A prediction to quantify weldability using robust regression analysis is performed in this study.Regression analysis is an important statistical tool that is routinely applied in most sciences.Out of many possible regression techniques,the least-squares(LS) method has been generally adopted because of tra-dition and ease of computation.However,there is presently a widespread awareness of the dangers posed by the occurrence of outliers.Outliers occur very frequently in real data,and they often go unno-ticed because nowadays the data are largely processed by computers,without careful inspection or screen-ing.To remedy this problem,robust regression anal-ysis has been adopted,which is not so easily affected by outliers.The main purpose of the current study is to present the prediction robustly and suggest the opti-mum process factors for hybrid welding.With this1672I Jang and K Y LeeTable1Chemical composition of6061-T6aluminium alloyChemical composition(wt%)Si Fe Cu Mn Mg Cr Zn Ti Al0.40–0.8 0.70.15–0.40 0.150.8–1.20.04–0.35 0.25 0.15RemainderTable2Mechanical and thermal properties of6061-T6aluminium alloyMechanical properties:Density(×1000kg/m3) 2.7Poisson ratio0.33Elastic nodulus(GPa)68.9Tensile strength(MPa)290Yield strength(MPa)240Elongation(%)16Hardness(Brinell)95Hardness(Rockwell A)40Hardness(Rockwell B)60Fatigue strength(MPa)96.5Thermal properties:Melting point(◦C)582Thermal conductivity(W/m K)167 perspective,in this investigation the possibility of using robust regression analysis and a genetic algo-rithm for neodymium-doped yttrium–aluminium–garnert(Nd:YAG)laser–gas metal arc(GMA)hybrid welding of6061-T6aluminium alloy is explored.2EXPERIMENTAL PROCEDUREThe base metal used for the square groove welding of a butt joint is6061-T6aluminium alloy sheet.The size of each plate is150×300×2mm.The chemical composi-tion of the alloy is given in Table1and the mechanical and thermal properties are listed in Table2.A characteristic feature of all known hybrid weld-ing techniques is the simultaneous application of a laser beam and an arc heat source during the weld operation.Generally,the term laser–arc hybrid weld-ing is specified when the primary laser heat source such as Nd:YAG enables welding processes in the deep penetration mode and the arc heat source is operat-ing between a mechanically supplied wire electrode and the workpiece.Such an integrated addition of filler material offers the possibility to influence the weld properties by the proper choice of the elec-trode alloy composition.The shape of the arc is strongly influenced by the shielding gas type such as argon(Ar)or helium(He).The combination of an Nd:YAG diode pumped laser beam(the maximum power P0=4.4kW)and an electric arc(the maxi-mum current I0=400A)within a common process zone was used and the general schematic representa-tion of a hybrid laser–arc welding process is shown inFig.1.Fig.1Schematic diagram of the hybrid welding process In selecting welding parameters,structural weld-ing codes of aluminium from the American Welding Society and manuals of both laser and GMA welding systems were followed.Five welding parameters were selected,namely the laser power,the travelling speed, the welding wire feed rate,the root opening tolerance, and the voltage.In the experiments,conditions of varying levels were used for each parameter:laser power,travelling speed,wire feed rate,root opening tolerance,and arc voltages.The laser powers were3,3.5,and4kW;the travelling speeds were60,80,100,120,and160mm/s; wire feed rates were4.5, 5.7,and6.5m/min;root opening tolerances were0,0.2,and0.5mm;arc volt-ages were10,12.6,15,and20V.Distance between levels was determined by initial experiments to pro-cure welding quality.Each experiment was repeated three times for the same conditions.The weld angle was30◦.The focal position(250mm)was fixed on the surface of the plate.The shielding gas was argon and the flowrate was20l/min on the up-side of the aluminium sheet and10l/min on the down-side of the aluminium sheet.Acetone was used to clean the surface of the aluminium before welding.In order to compare the weldability of each experimental group, ultimate tensile strength(UTS)was chosen,since an evaluation of strength on the weld zone was judged to be a more important criterion than other evalua-tion methods such as observation of joint penetration and bead width,the quality of the surface form,the absence of the pore in the joining parts,and so on[8].The UTS was obtained from two samples for each specimen.Therefore,six tensile test samplesApplication of robust prediction for a laser–GMA hybrid welding process1673for each condition were made and the mean value of these samples was considered to be the experimental result.3RESULTS AND DISCUSSION3.1Regression analysisWelding affects not only the quality but also the safety of a structure,so welding should be per-formed under optimal welding conditions.In select-ing the optimal welding conditions,the experiments should be repeated many times,as the welding quality may change according to the thickness and shape of the welded part.This requires an immense number of experiments.However,it is not easy to obtain an optimal condition using a trial-and-error method or to perform so many experiments.There-fore,a method which obtains an optimal condition economically should be considered.The statistical approach method is used for efficiency of the weld-ing database.Recently,research has been carried out which explains the correlation between welding pro-cess parameters and weldability[7].However,this research has not been sufficiently widely presented to be applied to the parameter prediction system in hybrid welding.In the current study,a robust regres-sion equation was developed using hybrid welding process parameters.In order to test the accuracy of the regression equation,the data obtained through the experiment were applied to the data for the analysis and verification levels.No extrapolation is possi-ble because regression analysis is an interpolation approach.3.2Robust predictionIn the present study,a regression was carried out to examine the relationship between these parameters. Step-wise multiple regression analysis,a kind of sta-tistical method,was used to consider the regression in UTS predictions.Parameters such as the laser power, the travelling speed,the welding wire feed rate,the root opening tolerance,and the voltage were used as independent parameters in the regression analy-sis.The adaptability of the regression was estimated through data for the analysis of several conditions.The two statistical software packages used for these analy-ses were SAS9.1[9]and SPSS17[10].The analysis and verification data sets are shown in Table3.The regres-sion was obtained based on parameter ranges from 3to4kW for the laser power,from60to160mm/s for the travelling speed,from4.5to6.5m/min for the wire feed rate,from0to0.5mm for the root opening tolerance,and from10to20V for the arc voltages. The least-trimmed-squares(LTS)method and Cook’s distance value were used to build the robust Table3Ultimate tensile strength of hybrid welding joint Data UTS for UTS forindex analysis(MPa)verification(MPa) 1231.0217.02215.0145.53231.0235.54212.0177.85231.0201.76218.0234.97182.0235.58241.7226.19241.8178.810241.8243.911231.8241.212237.7186.913241.8184.814243.2236.815241.0256.916229.7237.017230.6227.318221.5214.519221.520234.521236.222221.523227.224223.025231.026215.027231.028212.029231.030218.0prediction.The LS algorithm is a general regression algorithm in statistics and has been used extensively in engineering applications.It can be summarized by MinimizeθNi=1r2(1)whereθis the regression parameter to be estimated,r2 is the squared residual associated with observation i, and N is the number of observations.However,the LS analysis can be severely compromised by a sin-gle outlier in the data set.The LTS estimator achieves robustness by trimming away observations with large residuals.This algorithm can be summarized byMinimizeθhi=1(r2)i:N(2)where(r2)1:N (r2)2:N ··· (r2)N:N.The LTS estimator tries to eliminate contaminated data and to estimate the desired parameterθusing the remaining data.When N/2+1 h (3N+P+1)/4,the esti-mator can obtain a good estimate ofθfrom a noisy data set with contamination[11].In this study,the number of observations(N)was30and the number of independent variables(P)was5.1674I Jang and K Y LeeCook’s distance is commonly used for estimating the influence of a data point when doing LS regres-sion.Cook’s distance measures the effect of deleting a given observation.Data points with large value of Cook’s distance may distort the outcome and accu-racy of a regression.The equation of Cook’s distance to deletion (C i )is expressed in the formC i =nj =1(ˆY j −ˆY j (i ))2pMSE(3)which was developed by Cook and Weisberg [12].ˆYj is the prediction from the full regression for observa-tion and ˆYj (i )is the prediction for observation j from a refitted regression in which observation i has been omitted.Kim and Storer [13]suggested a criterion C i 3.67/(N −P )for detecting outliers.When con-servative robust regression criterion was applied,the number of detected outliers was 6,as shown in Table 4and Fig.2.Recently,research has been conducted on the relation between outliers detected by parameter esti-mation and experimental defects [14].Figure 3shows the microstructures of a weld outlier detected in Fig.2using the robust regression criterion.It can be observed that there was a defect owing to porosity in the heat-affected zone(HAZ).Fig.2Cook’s distance plot with detected outliersTable 4Criterion for robust regressionCriterion for LTS,h Criterion for Cook’s distance,C i 16 h 240.1468 C iRobust prediction with deletion of outliers byusing the statistical program R [15]was analysed.The following is the equation of UTS in terms of parametersˆy(UTS )=−216.98+3534.84X 2−4196.33X 5+441.84X 21−1191.66X 22+197.06X 23−238.42X 24+1590.60X 25−2104.00X 1X 2+1053.07X 1X 3+70.22X 1X 4+1637.71X 1X 5−2140.49X 2X 3−1683.09X 2X 4+157.26X 2X 5−1631.51X 3X 4+2182.30X 3X 5+2339.10X 4X 5(4)where X i are defined in Table 5.Tables 6to 9and Figs 4to 7show the error rates using classical and robust regression analysis.There is a difference between error rates for the exper-imental and estimated values.The maximum and average error rates of the analysis data for the clas-sical method were 12.71per cent and 4.30per cent,respectively,and the error rates of the verification data for the classical method were 20.33per centandFig.3Microstructure of the weld zone with defect Table 5Expression of the welding parametersX 1X 2X 3X 4X 5GapLaser powerArc voltageWelding speedFeedrateApplication of robust prediction for a laser–GMA hybrid welding process1675Table6Error rates of analysis by the classical regression UTS Estimation Error Abs.error 156.0171.8−10.1610.16 228.0211.97.057.05 231.0218.8 5.26 5.26 215.0217.4−1.14 1.14 231.0218.8 5.26 5.26 166.0187.1−12.7112.71 212.0215.0−1.42 1.42 231.0218.8 5.26 5.26 218.0219.8−0.820.82 182.0190.3−4.57 4.57 234.4217.17.387.38 241.7245.7−1.65 1.65 241.8245.0−1.31 1.31 212.1234.2−10.3910.39 241.8245.0−1.31 1.31 231.8222.8 3.91 3.91 237.7242.1−1.83 1.83 241.8245.0−1.31 1.31 243.2243.5−0.120.12 241.0227.1 5.76 5.76 229.7231.1−0.630.63 230.6242.6−5.23 5.23 221.5230.5−4.05 4.05 229.9205.410.6510.65 221.5230.5−4.05 4.05 234.5222.4 5.14 5.14 236.2228.9 3.12 3.12 221.5230.5−4.05 4.05 227.2225.10.920.92 223.0228.5−2.49 2.49Mean error 4.30Max.error12.71 Table7Error rates of verification by the classical regres-sionUTS Estimation Error Abs.error 217210.0 3.23 3.23 145.5128.511.6811.68 235.5215.58.498.49 177.8165.8 6.75 6.75 201.7203.7−0.990.99 234.9225.9 3.83 3.83 235.5244.5−3.82 3.82 226.1217.1 3.98 3.98 178.8194.8−8.958.95 243.9255.9−4.92 4.92 241.2248.2−2.9 2.9 186.9148.920.3320.33 184.8212.8−15.1515.15 236.8222.8 5.91 5.91 256.9281.9−9.739.73 237229.0 3.38 3.38 227.3254.3−11.8811.88 214.5215.5−0.470.47Mean error7.02Max.error20.33 7.02per cent,respectively.In the case of the robust regression method,the error rates of the analysis data were2.23per cent and1.37per cent,respectively, and those of the verification data were8.86per centTable8Error rates of analysis by the robust regression UTS Estimation Error Abs.error 231.0232.0−0.440.44 215.0219.8−2.22 2.22 231.0234.1−1.33 1.33 212.0215.1−1.45 1.45 231.0236.1−2.21 2.21 218.0222.1−1.88 1.88 182.0178.3 2.06 2.06 241.7238.0 1.55 1.55 241.8243.9−0.850.85 241.8244.2−0.990.99 231.8234.2−1.03 1.03 237.7242.8−2.15 2.15 241.8246.6−1.97 1.97 243.2242.50.280.28 241.0235.9 2.12 2.12 229.7224.6 2.23 2.23 230.6231.3−0.30.3 221.5224.6−1.39 1.39 221.5224.2−1.23 1.23 234.5238.9−1.89 1.89 236.2237.2−0.430.43 221.5224.2−1.23 1.23 227.2229.6−1.05 1.05 223.0224.4−0.610.61Mean error 1.37Max.error 2.23 Table9Error rates of verification by the robust regressionUTS Estimation Error Abs.error 217.0232.3−7.077.07 145.5157.1−7.977.97 235.5232.1 1.45 1.45 177.8170.6 4.03 4.03 201.7216.7−7.447.44 234.9242.4−3.19 3.19 235.5221.2 6.08 6.08 226.1215.5 4.68 4.68 178.8170.6 4.58 4.58 243.9246.3−0.980.98 241.2252.8−4.81 4.81 186.9191.4−2.37 2.37 184.8201.1−8.868.86 236.8233.4 1.44 1.44 256.9274.7−6.9 6.9 237.0243.8−2.88 2.88 227.3235.8−3.75 3.75 214.5228.5−6.52 6.52Mean error 4.72Max.error8.86 and4.72per cent,respectively.In this interpretation, using the robust regression method is more reliable. An analysis of variation(ANOVA)was used to deter-mine the statistical effects of the parameters on this robust approach.Table10shows the analysis of vari-ance.Degrees of freedom are denoted DOF,SS is the sum of squares,and MS is the mean square.F0is an index that shows the ratio of the sum of mean squares1676I Jang and K YLeeFig.4Classical regression analysis prediction of analysisdataFig.5Classical regression analysis prediction of verifica-tion datato the sum of residual mean squares.F (0.05)means 95per cent of the standard deviation.If F 0 F (0.05),this means that the regression equation is well fitted to the experimental data.Table 10shows that the robust regression equation depicts a statistical relationship well between the input and output variables,because F 0=31.53is sufficiently greater than F (0.05)=1.93.3.3Determination of the optimum weldingparameters A genetic algorithm,which is based on natural selec-tion and genetics,overcomes the many problems of conventional local search algorithms.A genetic algorithm was used to find the near-optimal welding conditions;it has the followingcharacteristics.Fig.6Robust regression analysis prediction of analysisdataFig.7Robust regression analysis prediction of verificationdataTable 10ANOVA for robust predictionFactor DOF SS MS F 0F (0.05)Regression 54434.96886.9931.531.93Residual 4224451.38582.17Total4728886.35First,the genetic algorithm uses a specific length of binary strings,composed of 0and 1instead of input variables.Second,the genetic algorithm searches a set of possible solutions.Therefore,it is effective in finding a global optimal point by preventing it from converging to a local optimal point.Third,because the genetic algorithm uses only the fitness function value,the fitness function does not have to be either contin-uous or differentiable.Finally,the transition rule usedApplication of robust prediction for a laser–GMA hybrid welding process1677in the genetic algorithm is probabilistic rather than deterministic.In the current study,the genetic algorithm mod-ule for MATLAB was used for solving the optimization problem.The objective was to obtain the weld-ing parameters that give the maximum UTS,and equation(4)was used as the objective function. The search and optimizing ranges for the welding parameters were the same as those of the regres-sion analysis.The regression analysis was developed based on interpolation,so the result could not be guaranteed when the search and optimization ranges were wider than those of the regression analysis. The number of individuals in the population was20. Among the genetic operators,the crossover rate was set at0.8and the mutation rate was0.01.Each parameter value was determined based on previous studies[16].Figures8and9and Table11show the optimized welding parameters and estimation of UTS by the genetic algorithm.Within the investigated range of parameters,the optimized welding parameters are as follows:the laser power is3.34kW,the welding speed is93.3mm/s,the feed rate is6.5m/min,the gap is0.315mm,and voltage is14.52V.For these conditions,the estimated UTS from the robust pre-diction is268.61MPa and the experimentally mea-sured UTS is259.32MPa.This optimal point is found from the genetic algorithm including consideration of weldability and robustness,so it can be consid-ered as the optimal hybrid welding conditions with high UTS.4CONCLUSIONSA Nd:YAG laser–GMA hybrid welding experiment was performed.In order to evaluate the weldability,the UTS was measured.A robust regression analysis and genetic algorithm were adopted to build the prediction for hybrid welding and to optimize the welding parameters.The LTS method and Cook’s distance were used for detecting outliers.The result of the robust regression analysis shows a signifi-cant improvement in prediction over the classical regression analysis because the outlier associated with a weld defect was detected and deleted.Through the genetic algorithm,the optimum hybrid weld-ing parameters were determined as a laser power of3.34kW,a welding speed of93.3mm/s,a feed rate of6.5m/min,a gap of0.315mm,and avoltage Fig.8Genetic optimization(50thgeneration) Fig.9Genetic optimization(200th generation)Table11Results of the genetic algorithm for optimizationGeneration Gap(mm)Laser power(kW)Arc voltage(V)Weld speed(mm/s)Feed rate(m/min)UTS(MPa)500.30 3.3514.490.9 6.5268.4 1000.31 3.3514.692.6 6.5268.6 2000.32 3.3414.593.3 6.5268.6 4000.32 3.3414.594.0 6.5268.61678I Jang and K Y Leeof14.52V.The estimated UTS(268.61MPa)and the experimentally measured UTS(259.32MPa)based on the optimal conditions were much higher than the strength range(165MPa)which is specified by the American Welding Society.©Authors2010REFERENCES1Magee,K.H.,Merchant,V. E.,and Hyatt, C.V.Laser assisted gas metal arc weld characteristics.Proc.ICALEO’90,1990,71,382–399.2Beyer, E.,Imhoff,R.,Maier, C.,Neuenhahn,J., Behler,K.,and Dilthey,U.New aspects in laser weld-ing with an increased efficiency,1994(SPIE,Bellingham, WA).3Dilthey,U.and Wieschemann,A.Prospects by combin-ing and coupling laser beam and arc welding processes.Rivista Italiana della Saldatura(Italy),2000,52(6), 749–759.4Takashi,I.,Shuho,T.,and Masao,test MIG,TIG arc–YAG laser hybrid welding systems for various weld-ing products.In1st Int Symp High-Power Laser Macro, edited by I.Miyamoto(SPIE,Osaka),2002,pp.347–352. 5Bagger, C.and Olsen, F.O.Review of laser hybrid ser Applications,2005,17,2.6Mahrle,A.and Beyer,E.Hybrid laser beam welding–classification,characteristics,and ser Applic.,2006,18,169.7Kim,H.R.and Lee,ing the orthogonal array with grey relational analysis to optimize the laser hybrid welding of a6061-T6Al alloy sheet.Proc.IMechE,Part B: J.Engineering Manufacture,2008,222(8),981–987.DOI:10.1243/09544054JEM1070.8Park,Y.W.and Rhee,S.Process modeling and param-eter optimization using neural network and genetic algorithms for aluminium laser welding automation.Int.J.Advd Mfg Technol.,2008,37(9),1014–1021.9SAS Institute.SAS software,Version9.1of the SAS System for Windows.(SAS Institute Inc.,Cary,NC).10SPSS Inc.SPSS for Windows,Rel.17.0.1.,2001(SPSS Inc., Chicago,IL).11Rousseeuw,P.J.and Leroy,A.M.Robust regression and outlier detection,1987(Wiley,Chichester).12Cook,R. D.and Weisberg,S.Characterizations of an empirical influence function for detecting influen-tial cases in regression.Technometrics,1980,22(4), 495–508.13Kim,C.and Storer,B.E.Reference values for Cook’s muns Statists-Simulation Comput.,1996, 25(3),691–708.14Stache,N.C.and Zimmer,H.Robust circle fitting in industrial vision for process control of laser weld-ing.In Proceedings of the11th International Student Conference on Electrical engineering,Czech Technical University,Prague,2007,pp.1–7.15R Development Core Team.R:A language and envi-ronment for statistical computing.In R foundation for statistical computing,2009(R Development Core Team).16Schaffer,J. D.,Caruana,R. A.,Eshelman,L.J., and Das,R.A study of control parameters affecting online performance of genetic algorithms for function optimization.In Proceedings of the3rd International Conference on Genetic algorithms and their appli-cations,(Morgan Kaufmann Publishers,San Mateo, California),1989,pp.51–60.。
