Analysis of surface roughness of rock dust reinforced AA6061-Mg matrix composite in turning

R.Blchndhr,R.Blsundrm,M.Rvichndrn

a Department of Mechanical Engineering,Pavender Barathidasan College of Engineering and Technology,Tiruchirappalli,620024,Tamil Nadu,India

b Department of Mechanical Engineering,K.Ramakrishnan College of Engineering,Tiruchirappalli,621112,Tamil Nadu,India

Abstract This work,examines the Surface Roughness(SR)of composite consisting Aluminium alloy(AA6061),Magnesium and Rock dust during turning process.To study the performance,three different test specimens with different constituents of Al 6061-T6,AZ31 and Rock dust were prepared by stir casting method.Turning experiments were carried out using MTAB Siemens-CNC lathe.The input parameters for machining are speed,depth of cut & feed and output response is surface roughness For each test specimen,there are 15 turning operations were performed using Box-Ben hen Design approach.To analyze the process parameters for SR,the models of ANOVA and Decision Tree(DT)algorithms were performed.Both algorithms are confirme that,speed is the most significan factor for SR.The addition of AZ 31 with 1% and rock dust of 2% in AA6061 produced better surface finish Regression models of linear regression,multilayer perception and support vector regression from data science were formulated to fin the relationship between variables.Among these models multi layer perception produced minimum root mean square error.? 2021 Chongqing University.Publishing services provided by Elsevier B.V.on behalf of KeAi Communications Co.Ltd.This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/)Peer review under responsibility of Chongqing University

Keywords:Composite;Turning;Surface roughness;Anova;Decision tree;Regression.

1.Introduction

Aluminum composites holds of good attractiveness in a wide variety of applications such as constructions,aerospace,aerospace for exterior & interior due to its strength to weight ratio,little density,good corrosion resistance,etc.[1,2,3].In order to enhance their properties,various reinforcing materials were used.There are quite lot of Al-based MMC’s were available in the literature which are able to produce better properties and used in wide range applications[4].In recent years,researchers are using various natural wastes such as fl ash,rock dust etc.,which are abundantly available and unaltered degradable among the reinforcements[5].These materials are low cost and able to improve wear and hardness properties.Anilkumar et al.[6]showed that addition of 10 % weight fractions of fl ash increases hardness from 48 to 62 BHN ie.,29.16 % improvement in Al 6061 alloy.Soorya Prakash et al.[7]studied addition of rock dust along with Al 6061 for its wear characteristics.The addition of rock dust as reinforcement enhanced the wear resistance of Al6061.The AZ31 is largely popular amongst aluminum and magnesium and alloys due to its lesser density and elevated mechanical characteristics.It is widely used for making ribs,brakes in aircraft industry[8].Maleque et al.[9]examined the fracture mechanism of Sic reinforced aluminum matrix composite.For optimum fatigue life strength,20 % of Sic by weight produced better results.Vadivel et al.[10]successfully synthesized the composites containing Al6061 with different proportions of Tic Particulates and Mg.The test result showed that by increasing reinforcement,the tensile strength also increases but there is decrease in ductility.Valerio Oddone et al.[11]proposed spark plasma sintering method for metal-graphite composites out of magnesium and aluminum alloys.The proposed composite achieved four times better specifi thermal conductivity than that of copper.Francis Xavier & PramasivamSuresh[12]used wet grinder stone as reinforcement along with aluminum and results indicated that by increasing reinforcement,its hardness & wear resistance increased but it also reduces the ductility.Ratna Sunil et al.[13]studied drilling characteristics of Al MMC’s with AZ31 and AZ91 and test results indicated that,the cutting forces are affected by presence of second phase-Mg17Al12.Devganiya & Patel[14]prepared composite consisting of Al 6061 with Sic of 5% and 10% weight fraction proportion.Hardness,Density and yield strength,were increased with the increase in reinforcement particles of silicon carbide.Jufu Jiang et al.[15]reinforced Al 7075 aluminum matrix composite with nano-sized Sic particles.The true stress-strain curve at 600°C is analogous to that at 500°C.Jayalakshmi Subramanian et al.[16]studied properties of composite consisting of Al with Mg & amorphous glass.The proposed composites showed better properties than that of conventional method.Ramesh et al.[17]proposed quarry dust as reinforcement material for A356 MMC.The wear resistance of the proposed composite drastically increases by increasing reinforcement.The mechanical characteristics of pure Mg reinforced with Al and Ti were studied by Muhammad Rashed et al.[18].The addition of Mg improved the yield strength,elastic modulus and failure strain.Abhijit dey and Murari[19]studied microstructure and various mechanical properties of Mg with various alloys.The Mg with Sic has superior wear resistance.Pitchayyapillai[20]studied wear characteristics of Al 6061 reinforce with various percentage of ceramic alumina and soft solid lubricant of MoS2.The incorporation of MoS2and alumina increases the wear,friction and mechanical characteristics.

Balachandar et al.[21,22]investigated dry sliding wear characteristics of Al6061 reinforced with AZ31 and rock dust with 1-2% variation.The results showed that 2 % of AZ 31 and 1% of rock dust produced minimum wear.The authors also analyzed cut quality characteristics of above composite using abrasive water jet machining.The ultimate tensile strength improved by addition of AZ31.The adding of rock dust improved the wear resistance of proposed the composite.The rock dust and Mg,were varied from 1 %-2 %with Al 6061 T-6 and its composition is shown in Table 1.The test specimens were prepared using stir casting machine.The AZ31 and Al6061-T6 were taken as solid metal and rock dust as 30μm fin power form.Temperature of 500-800°C maintained inside the crucible and constantly stirred.The molten metal is transformed into work piece of having length of 120mm and 20mm diameter.The Fig.1 shows the fabricated test specimen along with stir casting equipment.

Table 1Proposed composite proportions.

Table 2MTAB CNC lathe specification

Table 3Parameters for ANOVA.

Table 4Machining parameter and Surface roughness.

2.Experimentation

To analyze the surface roughness characteristics,turning operations were performed on Siemens CNC lathe with CCNG cutting tool.The typical CNC machine and work plan are shown in Fig.1.Machining operations were performed for 100mm length and surface roughness is measured using surface roughness tester TR110 Fig.1.In order to minimize the number of experiments,time,cost and efforts,Design of Experiments(DoE)is used.It is statistical tools for proficien design and analysis of experimental data.The RSM based Box-Ben hen Design is used for to conduct number of experiments.For each specimen 15 experiments were performed.The Tables 2-4 shows the specificatio of CNC machine,Process parameters and the coded value along with experimentation.

Table 5Specimen-2 turning parameters for construction of Decision Tree algorithm.

From the experiments it is clearly shown that the test specimen?2 having Al 6061-97 %,AZ 31-1% and rock dust-2 % produced better surface finish Hence,the analysis of surface roughness is limited to specimen-2 alone.

Fig.1.Stir casting machine and work plan.

Fig.2.Various interaction effects of process parameters for specimen-2.

3.Analysis of surface roughness using anova

The Analysis of Variance table consist of sum of square,source of variance,degrees of freedom & probability associated with it.The Fig.2 shows the interaction plot of various parameters and where A is speed in rpm,B is feed in mm/min and C is depth of cut in mm.

Fig.3.Interaction plot of Speed Vs Feed.

The interaction effect of feed and speed is shown in Fig.3.It is observed that the effect on surface roughness by feed is minimum as compared to speed.For low speed,surface roughness is high and as the speed increases,it is significantl reduced.The interaction effect of speed and Doc is shown in Fig.4.From the figure it is observed for the lowest speed the variation of depth of cut has significan effect on surface roughness but its effects are minimum for maximum speed.The interaction effect of feed and Doc is shown in Fig.5.From the figur it is observed that the variation of feed on effect of surface roughness is minimum as compared to Depth of cut.

Fig.4.Interaction plot of Speed Vs DoC.

Fig.5.Interaction plot of FeedVsDoC.

Fig.6.Decision Tree for the specimen-2 turning dataset.

4.Analysis of surface roughness using data science

Data science is an interdisciplinary fiel responsible for the researchers to visualize and get insight into the data.It is the study of probability& statistics can provide right data model to manufactures about key performance indicators about the process.Now a day’s manufacturing researchers are using data science tools to analyze the data generated during production process to know minimize the risk,increase the profi and factors affecting the productivity.Harding et al.[23]reviews application of data mining in manufacturing engineering in various areas of production process,maintenance,product quality improvement,fault detection etc.,The task of data mining are i)classificatio(classify the give data based on its attributes),ii)prediction(predicts the possible value)iii)Clustering(grouping the data)iv)Association(relationships between items that are co-occurring are expressed as association rules.Since data science is dominating most of the manufacturing industries.This work proposes the classificatio and regression analysis are applied to analyze the surface roughness characteristics.

4.1.Application of decision tree algorithm to analyze surface roughness

The application of Decision Tree algorithm is belongs to classificatio problem.It is supervised machine learning method for prediction models from the data.The advantage of DT is,the rules generated from data set are simple to understand.Quinlan[24]proposed algorithm for creation of DT called Iterative Dichotomiser 3(ID3)and enhanced version is C 4.5.Balasundaram et al.[25]Proposed step by step procedure for applying C 4.5 alogorithm for fl w shop scheduling data set for total fl w time criterion.The C 4.5 algorithm is applied for the specimen?2 turning data set.The algorithm used both top down and bottom up approach and searches through the attributes and extract the best splitting attributes that divides.To apply C 4.5 algorithm the output response(class attribute)is in terms of categorical one.To covert numerical value in to caterigorica lone,the average valueμm is taken as base.Below average value are taken as‘Low’and above average value are considered as‘High’.The reconstructed table is shown in Table 5.

Average value-1.0797,minimum value-0.89,Maximum value-1.3.

After applying C 4.5 algorithm,the fina decision tree is shown in Fig.6.

Fig.7.A simple linear regression model.

The Rules from the DT are:

?Speed≤1000,SR is High Rule—(1)(1,2,3,4).

?Speed≥1000 and DoC≤0.2mm,SR is High else Low Rule—(2).

The rule(1)correctly classifie the trial runs of 1,2,3,4 and remaining runs are correctly classifie by the rule(2).

From DT,it is understood that speed foremost significan parameter followed with Doc for machining proposed composite.These parameters are well validated with ANOVA.

4.2.Application of regression models for surface roughness

It is machine methods to predict the output variable from set input variables.ie.,it builds the mathematical model that define output variable as functions of input variables.It determines the best fi line,which is a line that passes through all the points in such a way that closeness of the line from each data point is minimum.There are different types of regression models are used.They are i)linear regression ii)Bayes regression,iii)support vector regression iv)multi layer perception v)Logistic regression etc.In this work,the post popular regression models of such as linear,multi layer perception and support vector regression models are used to predict the surface roughness.In linear regression model,both input and output variables are linearly related to each other.If input data involves more than one variable,then linear regression is called multiple linear regression models.A linear regression model is shown in Fig.7.

The fiel of multilayer perceptron is most powerful neural network.It can be used to solve difficul computational task to investigate how simple models of biological brains are used.It develops robust algorithms that can be used to model difficul problems.The data structure can pick out features at different scales or resolutions and combine them into higher-order features.A simple perception model is shown in Fig.8.

Fig.8.Multilayer perception model.

Support Vector Machine(SVM)is another straightforward algorithm that every data science expert should have in his/her arsenal.It produces significan accuracy with minimum computational effort.It can be used both classificatio and regression models.The purpose of SVM is to fin the best hyper plane in a data space that classifie data points.The Fig.9 shows simple two classes of data sets are classifie by support vectors.

Support vectors are data points that are closer to the hyper plane and influenc the position and orientation of the hyper plane.The support vector regression is to fin the best function that approximates mapping from an input to real numbers.The main aim is to decide a decision boundary at‘a(chǎn)’distance from the original hyper plane such that data points closest to the hyper plane or the support vectors are within that boundary line.These three algorithms are applied through WEKA.It is open source software that proves tools for data preprocessing,accomplishment of several Machine Learning algorithms to real-world data mining problems.The actual value algorithm with various regression model values is tabulated as shown in Table 6.

The regression models from various algorithms are:

?Linear Regression:

-0.0003 x Speed-0.317 x DoC+1.6468.

?Multi LayerPerception:

-2.5095 x(Threshold weight)+2.3682 x Speed-0.1781 x Feed-1.9873 x DoC.

?The support vectors from Support vector regression:

-0.4878 x(normalized speed)-0.4878 x(normalized Doc)+0.9746.

From regression equations,and DT it is well understood that feed has no effect on surface roughness for the proposed composite.

The root-mean-square error(RMSE)is a term mostly used to measure of the differences between values predicted by a model.It shows how close a regression line is to given data points.The value of RMSE zero(almost never achieved in practice)would indicate a perfect fi to the data.From the Table 6,it is shown that multi layer perception produced lowest root mean square error & mean absolute error.The correlation coefficien indicates data points clustering around a line.It can be between 1 and-1,with values closer to 1 indicates a stronger relationship.The three regression models produced more than 0.8 correlations co-efficient that indicates that there is strong relationship between input and output variables.

Fig.9.Support Vectors model for data sets.

Table 6Various regression model values along with actual value of surface roughness for specimen-2.

4.3.MATLAB tool box genetic algorithm(GA)

GA is one type of non traditional optimization method based on philosophy of natural genetics.It is used to fin optimal or near-optimal solutions to solve constrained and nonconstrained problems.The regression model from ANOVA is given to input of Mat Lab GA to fin the best operating conditions.There are 20 trials are carried out and best train is shown in Fig.10.The Table 7 shows the best value obtained from GA.

Table 7Best value of surface roughness for composites.

5.Conclusion

In this paper,AZ 31 and rock dust were reinforced Al 6061-T6 with different constitutions and prepared using stir casting.Box-Ben hen Design is used to select machining parameters and turning operations were carried out in Siemens CNC Lathe.From the experiments,composite consisting of 97% of Al6061,2% of rock dust and 1% of AZ 31 produced better surface finish To analyze the input parameters ANOVA and DT(C4.5)algorithms were applied.Both of them confirme that speed is foremost significan parameter followed with DoC and Feed has less significan for surface roughness of the proposed composite.To fin the correlation between input variables,the regression models of linear,multi layer perception and support vector regression were applied.All the algorithms show that there is strong relationship between input variables of Speed and DoC.Among three regression models,multi layer perception is best suited for the proposed composite,since it has less root mean square error and its correlation co-efficien is 0.98(achieving closer to 1).To fin the optimal machining parameters for the proposed composite Mat Lab GA is used.In future,ensemble methods(combination of more algorithms)from data science can be used to analyze the process parameters.

Fig.10.Best trial run of for test specimens 2.

Acknowledgement

The authors are thankful to the reviewers for their comments and suggestions to improve the earlier version of the work.