- 上一篇：工业机器人的结构优化设计英文文献和中文翻译
- 下一篇：煤粉燃烧和NO形成的数值模拟英文文献和中文翻译
Quality control process becomes complicated for thematerials having narrow processing window, which was notthe case in our experiments with Zytel nylon material.The widely used quality control methods in injection moldingare visual control, mechanical control, weight control,dimensional stability, and viscosity control (GE Plasticsw,2008).In visual control, there is mostly a direct correlationbetween appearance and properties of the molding, butdecisions are influenced by molder or inspector’s experience.Mechanical control methods demonstrate whether materialquality and processing conditions affect the mechanicalproperties of the part. A common procedure is the fallingdart test, which assess the component’s ductility. In our case study, for the fixed combination of polymer material andmold, this method could rather expensive. In stress controlmethod, producing stress free components will be the majorfocus, a simple and effective way of checking the stress inmolded parts is observing parts after 24 h immersion in 808Csolution of 100 g dish washing powder with 10 g gloss agent in 1 l of water (GE Plasticsw, 2008). However, this method oftesting gives qualitative information, and it is more effective incase of transparent plastic molding, which was not the casewith Zytel used in this study. Viscosity comparison before andafter molding is one of the way to assess the degradation ofplastic inside the mold cavity, which was not practicable atinvestigators lab as it needs special instruments.On the other hand, weight control and dimensional stabilitytest methods are economical, fast and easy techniques that canbe carried out at molding shop environment. Often, weightcontrol method is preferred over dimensional methods, sincethe weight variations will be more readily noticeable than thoseof dimensions. Wide variation in weights of the molded partscan indicate insufficient production and/ormachine tolerances.Stabilizing partweight in general indicates stabilized processingconditions (GE Plasticsw, 2008; Shaharuddin et al., 2006). Itcan also assist in checking bubbles and voids, or otherdeviations from filling rate of the cavity in the tool. Hence, thepart weight control can be correlated to strength of themoldings produced. The shot weight (part weight) alsorepresents the filling and cooling rates influenced by heattransfer across the mold wall (Ong et al., 2001). Dimensionalstability of the parts largely influenced by shrinkage, which is afunction of thermal properties of mold, polymerization ofpolymer, injection and holding pressure (Vlachopoulis andStrutt, 2003). Maintaining dimensional stability is one of thebasic requirements for the products in the view of theirassembly and interchangeability.Therefore, in this studyweightof the part and shrinkage were selected as desired qualitycontrol parameters. Weight of the molding part was measuredjust after its ejection from the cavity and dimensionalmeasurements were performed after 24 h. In shrinkagemeasurements, to maintain the uniformity, and to avoid theinfluence ofRTmold error, the ratio between dimensional errorand the corresponding RT mold dimension has beenconsidered for further analysis. These calculations areexplained in equations (1) and (2):dlinear ¼ LRT 2 LPartLRT: ð1ÞInjection molding error:d ¼PN1 dlinearN ð2Þwhere LRT, RT dimension; LPart, part dimension (measuredfrom the current experiments); N, number of dimensionsconsidered ¼ 5.The experiment was conducted on DGM-150microprocessor controlled injection molding machine. Ininjection molding, the process variables need some time toreach stable conditions; hence, each experiment wasperformed separately. While doing so, the process variablesas per the experimental sets were used and material wasinjected outside (air shot) over 20 shots to ensure that themachine operating parameters (process variables) havereached the steady state. After reaching steady state, theparts were molded. The parts produced in 5th, 10th and 15thshots were considered for weight and dimensional (shrinkage)measurements. These experimental results are summarized inTable IV. The effects of process variables on selected partquality factors were estimated using “grey relational analysis”described in the next section. 4. Optimization using grey relational analysisIn grey relational analysis, experimental results (dimensionalerror and weight difference) are first normalized in the rangebetween zero and one, which is also called as grey relationgeneration. The grey relational coefficients are calculated fromthe normalized experimental data, to express the relationshipbetween the desired and actual experimental data. The greyrelational grade is computed by averaging the grey relationalcoefficient corresponding to each response characteristics(shrinkage and weight). The overall evaluation of moldingparameters for multiple responses (shrinkage and weight) isbased on the grey relational grade. The optimal level of theprocess parameters is the level with highest grey relational grade.Depending on the characteristics of a data sequence, thereare various methodologies of data pre-processing(normalizing) available for the grey relational analysis.These include “higher is better”, “lower is better” and “thetarget value is definite”.In this study, the desired shrinkage value of Nylon 66 ¼ 1.5percent, hence, d ¼ 0.015 is used as targeted shrinkage of themoldings produced. Accordingly, the correspondingexperimental measurements (shrinkage) were normalizedusing:x*iðkÞ¼ 1 2 x0iðkÞ 2 x0 max x0iðkÞ 2 x0: ð3ÞSimilarly, higher is better characteristic was used for weightbecause higher the weight, stronger the molded part. Thecorresponding experimental measurements (weight) werenormalized using:x*iðkÞ¼ x0iðkÞ 2 min x0iðkÞmax x0iðkÞ 2 min x0iðkÞ: ð4ÞThe normalized experimental data using equations (3) and (4),corresponding to shrinkage error and weight error aresummarized in Table V.In grey relational analysis, the measure of relevancy betweentwo sequences is defined as the grey relational grade. Areference (ideal) experimental set such that normalizedmeasurement values equal to one was assumed forcomparison.