Characterization Of Key Process Parameters In Blow Molding .

Characterization of key process parameters in BlowMolding using Artificial Neural NetworksPredictive Process AnalyticsPRAVEEN BHAT & CHANDRASEKHAR ARCOT JPredictive Engineering, Axiom Consulting Pvt. Ltd.Bangalore, Karnataka, INDIA

PURPOSE Overall Objective of this study is to characterize the key process variables such asambient, operating temperatures, process line operators ,resin types, resin colors andoperating lines associated with the polymer bottle manufacturing. Identify definitive relationship between the selected process parameters for theproduction machine and ensure product quality Evaluate the performance characteristics of Post Consumer Resin (PCR) with that ofVirgin material and identify diverse variations with in them.2

INTRODUCTION Blow molding is one of the well- known manufacturingtechniques that is used to manufacture bottle ofcomplex shapes. Final optimal process parameters are one of the keydrivers in the blow molding process that improves thequality of the molded parts. Combined effects of geometry, part, materialcharacteristics, mold design and processing conditionson the part manufacturing is challenging to analysethrough analytical/mathematical model because of thecomplexity in the process as well as multiplicity of theparameters and its interactive effects on one another

PROCESS HIGHLIGHTSPre-forminjection moldedover a mandreland transferredto the blowingdiePolymer meltsupplied to moldhalves frominjection moldedmachinePreform carriedover on core pinThe blow tubetogether withthe parison isremoved fromthe injectionmold andtransferred to ablow mold.In Injection BlowMolding method aparison is producedby injecting apolymer into a hotinjection moldaround a blow tubeor core rod.AirBlow moldInjection moldOpenClosePreform expanded into finalbottle shape inside the moldInjection BlowMolding is moreaccurate andcontrollableprocess ascompared to theExtrusion BlowMolding.Air is injectedunder pressurethrough themandrel blowingthe polymer againstthe mold wallswhere it cools andfreezes as withextrusion blowmolding

PROCESS OPTIMIZATION & NEURALNETWORKSExisting Process studyNeural NetworkOptimized ParametersConditions

NEURAL NETWORK MODELDEVELOPMENTTo ProcessOptimization CycleModel Inputparameters-Resin typeResin ColorGradesPreform temperatureSet temperatureManufacturing lineSample spaceOperator efficiencyMATLAB’s NeuralNetwork ModelingEnvironmentDefects Identified% AcceptanceModel OutputparametersValidationTraining DataData from existingProcess line studyYES

NEURAL NETWORK MODELDEVELOPMENT

NEURAL NETWORK PREDICTION25PCR Virgin ExperimentalDefectspredictedfordifferent types of resins without any colorants added.PCR Virgin NNPrediction20% Defects15The defects predicted is forboth Virgin and PCR Resinmaterials from differentsource of suppliers.10500-520406080Trials100120140

NEURAL NETWORK PREDICTION40PCR GREEN Experimental35Defectspredictedfordifferent types of resins withwith colorants added.Virgin Green ExperimentalVirgin Blue Experimental30PCR GREEN NNPrediction% Defects25Virgin Green NNPredictionVirgin Blue NNPrediction20The defects predicted is forboth Virgin and PCR Resinmaterials with green and bluecolors.1510500-512345Trials678910

NEURAL NETWORK PREDICTIONBottle sample acceptance in % predicted for PCR Green Resin at different setpoint and Preform temperature. The acceptance rate prediction is also based onthe operation conditions, manufacturing line, different resin supplier etc.

NEURAL NETWORK PREDICTIONBottle sample acceptance in % predicted for Virgin Green Resin at different setpoint and Preform temperature. The acceptance rate prediction is also based onthe operation conditions, manufacturing line, different resin supplier etc.

NEURAL NETWORK PREDICTIONBottle sample acceptance in % predicted for Virgin Blue Resin at different setpoint and Preform temperature. The acceptance rate prediction is also based onthe operation conditions, manufacturing line, different resin supplier etc.

NEURAL NETWORK PREDICTIONBottle sample acceptance in % predicted for PCR Clear Resin at different setpoint and Preform temperature. The acceptance rate prediction is also based onthe operation conditions, manufacturing line, different resin supplier etc.

NEURAL NETWORK PREDICTIONDefect rates by operators predicted at different iterations along with cumulativepercentage defects. The defect rate prediction is also based on the operationconditions, Set point & preform temperatures, manufacturing line, differentresin supplier etc.

NEURAL NETWORK PREDICTIONDefects predicted at different manufacturing lines. The defects predicted isfor both Virgin and PCR Resin materials from different source of supplierswith different colors and manufactured under different conditions.

CONCLUSION The Neural network platform developed in MATLAB was used––––– In evaluating the variability in virgin grade of resin when compared to PCR. Variability in this context is the increased defectswith in PCR when compared to virgin material.Providing the correlation between the preform temperature and product quality. The exercise showcased no significant impactmade by the preform temperatures on the Injection blow molding process.Providing the correlation between the set point temperature and product quality. The exercise showcased no significant impactmade by the set point temperatures , where it shows minimal defects with accepted production level.Identifying product quality on different manufacturing lines with different operating conditionsProviding quantitative difference with respect to the variability's observed in Colored resins when compared to the clear resin.Process operators can employ range of suitable temperatures to reduce the amount ofdefects. The resin type, operating lines and resin colour have found to have significantimpact on the variability of the product manufactured using Injection blow moldingprocess.

ABOUT AUTHORSPraveen Bhat, Team lead, Predictive Engineering, Axiom Consulting, Bangalore, holds a Masters degree inMechanical Engineering (Design and Analysis) from Manipal University and has over Twelve plus years experiencein the field of Finite Element Methods (FEM), Advanced Computational Fluid Dynamics (CFD), ComputationalTribology, Multiphysics Engineering Design, Paper & Package Development, Predictive Engineering tooldevelopment. During his career, he has been involved in several Multi-field innovation and developmentalprojects. His expertise includes structural and thermal modeling in consumer electronics & packaging, Predictivetool and methods development, Optics, Acoustics & vibration. He has been working in different domains thatincludes Automotive & Aerospace, Consumer Package goods, Engineered Products, consumer electronics,Healthcare and medical devices. He has 4 Patents filed with 20 International Conference & Journal papersChandrasekhar A J, Director, for Predictive Engineering, Axiom Consulting, holds a Master’sdegree in Heat transfer from National Institute of Technology, Karnataka. He has hisengineering roots in the automotive and aerospace industries with specialty in applyingPredictive Engineering methods and tools (CAE) to solve engineering problems. He has spentthe past 20 years working on various product development programs, evolving methodologiesfor complex families of problems and developing “libraries” of applications that can be usedacross verticals to assist OEM’s and their suppliers with creative solutions. He also hasextensive experience with Mechatronics and Instrumentation for data capture and it’sapplications in building correlated Predictive Engineering models. In his current role, he drivesthe vision and charter for the Predictive Engineering Group within Axiom.

THANK YOU

Process operators can employ range of suitable temperatures to reduce the amount of defects. The resin type, operating lines and resin colour have found to have significant impact on the variability of the product manufactured using Injection blow molding process.