1、 分类号 学号密级I201022002学校代码 1 0 4 8 7博士学位论文基于智能优化方法的预热加工切削热和表面粗糙度研究学位申请人：Asaad Abdulsayed Abdullah学科专业：机械工程指导教师：熊蔡华教授答辩日期：2014.01.06A Dissertation Submitted in Partial Fulfillment of the Requirementsfor the Degree of Doctor of Philosophy in EngineeringCutting Temperature, Surface Roughness Modeling and
2、Optimization of Machining Parameters Using Intelligent Approach to Study Preheat Effect on Machining ProcessPh.D Candidate: Asaad Abdulsayed AbdullahMajor : Mechanical Engineering: Prof. Caihua XiongSupervisorHuazhong University of Science it is very difficult to establish a mathematicalmodel. The p
3、rediction of surface roughness and cutting temperature in machining areessential to strengthen the NC codes and then contribute to improving the reliability,accuracy and productivity in CNC machining. In fact, it may give information about3华中科技大学博士学位论文surface integrity; surface roughness and cutting
4、 temperature can be optimized in theselecting appropriate cutting conditions. Machining process prediction models represent amapping of cutting speed, feed rate and depth of cut and output variables using responsesurface methodology has been widely reported in literature.The theoretical analysis is
5、hard to perform because there are problems in theestablishment of mathematical models and dealing with these models. Such problems arelargely avoided in experimental studiesIn recent years, the multiple regression technique 12,13, mathematical modeling basedon the physics of the process 14,15, the f
6、uzzy-set-based techniques and network-based(employing neural networks or similar techniques) modeling 16-19, are widely used topredict surface roughness and cutting temperature. At this point, the ANN can beemployed to predict the surface roughness and cutting temperature in the machiningoperations
7、because they are suitable for model complex physical phenomena. Concerningthe problem, Artificial neural networks do not need an accurate formulate of physicalrelations. In other words, they only need experimental results for training and test network.ANNs is employed for prediction of the machine o
8、peration factors. The intelligentmethods were developed with consideration of a single objective, such as minimization ofthe surface roughness and cutting temperature.1.2.4 Determine the optimum combination of cutting parameters to access objectiveCutting variables such as cutting speed, feed rate a
9、nd cutting depth play vital role inmachining process to give the product-desired shape. These have a great influence on thequantity, quality and cost of production, cutting temperature and tool life; hence, theirjudicious selection assumes significance. The selected machining variables must achievet
10、he desired goal of machining process with the benefit of machining resources fullypossible, consistent with the restrictions on these resources. Traditionally the selection ofmachining parameters is carried out manually depending on the experience of themachinist or the planner and referring the ava
11、ilable catalogues and handbooks.4华中科技大学博士学位论文In recent, optimization algorithms are becoming increasingly popular in engineeringfields, primarily because of the availability and affordability of high-speed computers.They are extensively used in those engineering problems where the emphasis is onmaxi
12、mizing or minimizing of a certain goal. Production workers are interested indesigning optimum planning of various machining operations to improve productionquality and minimize the machining time and the machining cost. Machining engineers areinterested in designing planning production to achieve th
13、e best quality and maximumproduction and tools life.Extensive works have thus been reported over the last two decades in optimizing thecutting variable of machining process for a given optimum combination of machiningparameters and tool conditions. In fact, it may give information about surface inte
14、grity;surface roughness and cutting temperature can be optimized in the selecting appropriatecutting conditions. The intelligent methods were developed with consideration of a singleobjective, such as minimization of the surface roughness and cutting temperature. In theprocess of the single objectiv
15、e optimization, many different techniques were proposed,such as the Taguchi and response surface techniques 3,5,20-23, and genetic algorithm andparticle swarm18, 24-29.1.2.5 Study the effect of the pre-heating process on the performance of the cuttingprocess and adaptive control of machining process
16、esSuper material alloys are known widely used in important industries such as turbine,aerospace and aircraft industry because of their superior chemical, mechanical and hightemperature properties. These alloys are limited machinability by intensive tool wear thatindirectly represents a significant p
17、ortion of the machining. There are several techniquestowards improving the machinability of super material alloys and other materials. Thepreheat operation is one of the techniques which is utilized to improve the machinabilityof super material alloys 30,31.5华中科技大学博士学位论文Apply preheat of workpiece as
18、 a way to improve machining process that has beenunder consideration since the late 19th century. In the preheat operation, an external heatsource is utilized to soften the workpiece surface layer in order to decrease its tensilestrength and strain hardening 30,32.1.3 Research ObjectivesOur objectiv
19、es in this research can be illustrated by answering the followingquestions.Study the effect of machining parameters on the surface roughness and cuttingtemperature.In this research, Taguchi experimental design method is utilized to extract and designcutting process. As we mentioned, the surface roug
20、hness and the cutting temperature arecomplex functions require further attention in the study and analysis. A combinedtechnique utilized a Taguchis orthogonal array, analysis of variance (ANOVA) and meaneffect plots are employed to investigate the contribution and effects of machiningparameters on t
21、he surface roughness and cutting temperature.How to develop cutting temperature and surface roughness prediction models?As we know, the prediction of the cutting temperature and surface roughness duringmaterial removal processes is important for engineers. The cutting temperature and surfaceroughnes
22、s are influenced by several cutting parameters and the uncertainty inherent in themachining process. An estimation the cutting temperature and surface roughness are achallengeable problem. In the last decades, many researchers have studied this problemand the multiple regression analysis, mathematic
23、al and intelligent techniques are utilizedfor modeling and predicting cutting temperature and surface roughness. Therefore, wepropose to investigate the possibility and effectiveness of predicting cutting temperatureand surface roughness with an artificial intelligence methods. In order to improve t
24、heprediction accuracy in machining process, improved techniques we are suggested to model6华中科技大学博士学位论文cutting temperature and surface roughness with artificial intelligence techniques andtaguchi approach.How to improve the intelligence optimization methods that are used to determineminimum surface r
25、oughness and cutting temperature?Increasing production requires the involvement of all production processes, technicalpossibility for full utilize or activating all available processing facilities. In order to exploitand selection, combination of optimum machining conditions must be considered. With
26、time, as complexity in dynamics of cutting processes increased substantially, researchersand practitioners have focused on mathematical modeling techniques to determine optimalcutting conditions with respect to various objective criteria 33,34. Several optimizationtools and techniques proposed are a
27、lso based on Taguchi method 35,36, response surfacemethodology 37, genetic algorithm 38,39, particle swarm optimization 40,41, and artificialbee colony 42. Therefore, we suggest an optimization approach depending on combinedthe adaptive neuro-fuzzy with particle swarm and genetic algorithm technique
28、s.What is the effect of the preheat process on the performance of machining process?To investigate the effect of the preheat temperature on the cutting temperature, wewill apply preheat of the workpiece at different temperature. This will be investigated onthe workpiece at different temperature to s
29、tudy the effect of the preheating process at eachof the tool wear and cutting temperature.How can develop the adaptive control system of CNC machine?In the last decades, researchers proposed many methods to develop adaptive controlsystems. However, prediction and adaptive control for tool wear, in a
30、ddition to othercontrollable features, seems to be remains great in the research and development stage. Inthe present study, we proposed the development of an in the process too wear adaptivecontrol (IVBAC) system in the milling process The control strategy of the system is basedupon control of preh
31、eating work-piece temperature by employing Adaptive neuro-fuzzyinterference system (ANFIS).7华中科技大学博士学位论文1.4 Research Methodology1.4.1 Plan of Experiments1.4.1.1 Taguchi Experimental DesignTaguchi methods are statistical methods developed by Genichi Taguchi to improvethe quality of manufactured goods
32、, and more recently applied to engineering,biotechnology, marketing and advertisingThe Taguchi approach utilizes a special designof orthogonal arrays to study the entire parameter space with only few of experiments andovercome the problem in the traditional experimental design methods 43. The mostim
33、portant features of this method is to reduce the experimental time, providing of an effortin conducting experiments, reducing the cost, and to identify the factors affecting quickly.1.4.2 Artificial intelligence technique1.4.2.1 Artificial neural network (ANN)Neural networks are composed of simple e
34、lements operating in parallel. Biologicalnervous systems inspire these elements. An artificial neural network (ANN) is a layerednetwork of ANs. An ANN is consisted of an input layer, hidden layers and an output layer.It consists of cells of artificial neurons, which profit a connectionist approach t
35、o processthe information. The different layers are fully interconnected such that each neuron in onelayer is connected to all neurons in the next layer. Moreover, connections between theneurons in the same layer and feedback connections are not allowed. Each of its neuronshas only one input, and it
36、simply transmits the value at its input to its output. Actualinformation processing is performed by the neurons in the hidden and output layers.Information is stored in the inter-neuron connections. Learning consists of adapting thestrengths (or weights) of the connections so that the network produc
37、es desired outputpatterns corresponding to given input patterns. In other words, we can train a neuralnetwork to perform a particular function by adjusting the values of the connections(weights) between neurons. As each input is applied to the network, the network output is8华中科技大学博士学位论文compared to t
38、he target. Therefore, artificial neural networks are utilized to establish therelationship between input and output data. Researchers started to use neural network inthe late eighties as a new approach to solve real-world problems.1.4.2.2 Adaptive-Network-Based Fuzzy Inference System (ANFIS)Fuzzy in
39、ference systems (FIS) are applied in the several areas of life and engineeringfor the purposes of functional modeling, control, and classification, among others. Whenmodeling functions, an FIS can approximate a highly nonlinear system via a group of rules.Originally, experts tuned the FIS by manuall
40、y adjusting the parameters, which was anexpensive effort.In 1993, Roger Jang suggested a new approach for automated FIS tuning called theadaptive-network-based fuzzy Inference System (ANFIS) of a combination of neuralnetwork and fuzzy inference system 44. ANFIS is similar to a neural network, wherei
41、n theantecedent parameters in the FIS are trained via gradient descent, and the consequentparameters are trained via least-squares estimation. Consisting of architecture ANFIS ofthe five successive layers which a membership function layer, release the power layer,layer and the normalization of the p
42、ower of fire, the consequent layer parameter, and thecombination or the output layer. The objective function for ANFIS is simply the trainingerror, which is to be minimized.1.4.2.3 Genetic Algorithm (GA) & Particle Swarm Optimization (PSO)Genetic algorithms (GAs) represent a powerful optimization ap
43、proach in which thecomputational process mimics the theory of biological evolution 45,46. Genetic algorithmswere applied successfully in many vital areas such as production planning, solvingoptimization problems in engineering, optimal cutting lumber, and optimization operations47,48. In this approa
44、ch, new child chromosomes are created by crossover and/or mutationprocesses. Crossover occurs as a probabilistic exchange of genes between two or morechromosomes. Then all chromosomes are evaluated according to objective function, withthe fittest survive to the next generation 49. The result is the
45、genes that develop over time9华中科技大学博士学位论文to produce better and better solutions to a problem.Meanwhile, the particle swarm optimization (PSO) is a population based stochasticoptimization approach, inspired by the flocking behavior of birds 50-52. The population ofthe potential solutions is called a
46、swarm and each individual solution within the swarm, iscalled a particle. This approach has good performance, low computational cost and easyimplementation. Due to these advantages, PSO has attracted attentions from researchersaround the world since its introduction in 1995 51,53.1.4.2.4 Dissertatio
47、n OrganizationThis dissertation can be arranged into many parts: background, surface roughness,investigation the effect machining conditions on cutting temperature, optimization ofmachining parameters and the developed of an in-process cutting tool life adaptive controlsystem based on preheat proces
48、s. The following Fig. 1-1 shows the relationship betweenthese parts. The background part is explained in chapter 2. In this chapter, we narrative aliterature reviews that presents the method and technical approach, which is utilized as animportant technique to estimate the surface roughness and cutt
49、ing temperature. Thischapter also is explain the most important optimization methods and approaches employedto determine optimal machining parameters, the preheating operations and adaptivecontrol system in machining process.Chapter 3 represents a developed prediction model of surface roughness, byemploying the Taguchi method and artificial neural network. In this chapter, the Taguchimethod is applied to determine the minimum surface roughness. Moreover,
