EXTREME PROCESS MONITORING AND IN-LINE QUALITY ASSESMENT OF MICROMOULDINGS Polymer Process Engineering 2005 B R Whiteside, R Spares, M T Martyn, P D Coates, IRC in Polymer Science & Technology, Dept of Mechanical & Medical Engineering University of Bradford, Bradford BD7 1DP, UK
Contents Introduction Process monitoring research Experimental set up Results Product Inspection Theory Experimental Evaluation Introduction Process monitoring research Experimental set up Results Product Inspection Theory Experimental Evaluation
Micromoulding at Bradford Micromoulding research since 2001 Analysis of novel process dynamics Product property assessment 2005 – Centre for Micro and Nano Moulding MNM Lab - 7 micromoulding machines Metrology laboratory (AFM, SEM, Interferometry, Optical techniques) Micromoulding research since 2001 Analysis of novel process dynamics Product property assessment 2005 – Centre for Micro and Nano Moulding MNM Lab - 7 micromoulding machines Metrology laboratory (AFM, SEM, Interferometry, Optical techniques) Microsystem 50 Fanuc Roboshot 5t Metrology laboratory
Battenfeld Microsystem 50 Purpose built micro injection process Novel solution for injection/metering Servo-electric drives Automatic parts handling Clean room filtration Modular Purpose built micro injection process Novel solution for injection/metering Servo-electric drives Automatic parts handling Clean room filtration Modular
Battenfeld Microsystem 50
Process Characterisation Dynisco PCI 4011 Piezo load transducer Dynisco PCI 4006 piezo load transducer Temposonics R series displacement transducer J-type thermocouples A suite of sensors installed on the Microsystem to help determine process dynamics Plus machine encoder outputs
Typical Process Data Start inject Start inject End inject End inject Product solidification Product solidification Mould open Mould open
Why measure process dynamics? Pure research Highlight interesting/unexpected behaviour Validation of constitutive equations/computer based models Development for industry Identify processing problems Assist with process optimisation Pure research Highlight interesting/unexpected behaviour Validation of constitutive equations/computer based models Development for industry Identify processing problems Assist with process optimisation
Experimental details Evaluate the response of process measurement to forced changes in moulding conditions Melt temperature variation Mould temperature variation Determine which parameter is the statistically most sensitive to process variation Peak injection pressure Peak cavity pressure Injection pressure integral Cavity pressure integral Evaluate the response of process measurement to forced changes in moulding conditions Melt temperature variation Mould temperature variation Determine which parameter is the statistically most sensitive to process variation Peak injection pressure Peak cavity pressure Injection pressure integral Cavity pressure integral
Product details 0.34mg (HDPE), 0.49mg (POM) Large diameter = 1.0mm Small diameter = 0.5mm Gate dimension 0.1 x 0.2mm 0.34mg (HDPE), 0.49mg (POM) Large diameter = 1.0mm Small diameter = 0.5mm Gate dimension 0.1 x 0.2mm
Melt temperature variation Cavity pressure integral data appears to be the most sensitive indicator of change
Melt temperature variation Scatterplot matrix shows that integral measurements perform best and cavity pressure measurements are the most sensitive
Mould temperature variation Cavity pressure integral appears to show most sensitivity to process variation
Mould temperature variation Cavity pressure measurements most sensitive
Repeatability comparison 23.1mg (HDPE) Plaque dimensions 7.3 x 3 x 1mm Steps 1.0, 0.5, 0.25mm 23.1mg (HDPE) Plaque dimensions 7.3 x 3 x 1mm Steps 1.0, 0.5, 0.25mm 0.34mg (HDPE) Large diameter = 1.0mm; Small diameter = 0.5mm Gate dimension 0.1 x 0.2mm 0.34mg (HDPE) Large diameter = 1.0mm; Small diameter = 0.5mm Gate dimension 0.1 x 0.2mm
Coefficient of variation 23.1mg product 0.34mg product
What can we draw from this? Micromouldings form a small fraction of the total shot weight at the end of the flow path Small process variations have a large impact on moulding quality Cavity pressure sensors are required to monitor/maintain the process window Product imaging required where process yield <100% Micromouldings form a small fraction of the total shot weight at the end of the flow path Small process variations have a large impact on moulding quality Cavity pressure sensors are required to monitor/maintain the process window Product imaging required where process yield <100%
How do we relate process conditions to defective mouldings? Monitor the process conditions during a production run and subsequently measure product properties Atomic force microscopy Surface profilometry Nanoindenting Machine vision Use statistical methods to correlate defects with data acquisition results Monitor the process conditions during a production run and subsequently measure product properties Atomic force microscopy Surface profilometry Nanoindenting Machine vision Use statistical methods to correlate defects with data acquisition results Time consuming
Commercial vision systems Expensive Require multiple cameras for 3-d measurements Typically low resolution cameras Ideal system Microscope lenses Megapixel resolution or better 3-d measurements Fast acquisition speeds Reasonable price Expensive Require multiple cameras for 3-d measurements Typically low resolution cameras Ideal system Microscope lenses Megapixel resolution or better 3-d measurements Fast acquisition speeds Reasonable price
Image analysis of Micromouldings For many micromoulded products microscope lenses are required for accurate optical assessment Microscopes typically have very small depths of field so it is difficult to image a 3-dimensional surface Extended depth of field techniques have arisen to address this problem and these methods can also be used to generate 3-dimensional information For many micromoulded products microscope lenses are required for accurate optical assessment Microscopes typically have very small depths of field so it is difficult to image a 3-dimensional surface Extended depth of field techniques have arisen to address this problem and these methods can also be used to generate 3-dimensional information
Method CCD Camera Microscope Traverse the sample towards the microscope in 1um increments Capture each image to pc Process data to determine which frames are in focus for each pixel in the image Create 3-D dataset Traverse the sample towards the microscope in 1um increments Capture each image to pc Process data to determine which frames are in focus for each pixel in the image Create 3-D dataset Focal Plane Motorised Stage Image capture/stage controlling PC
Focus algorithms? Use convolution kernels to look for pixel regions with high local intensity gradients (contrast) Sobel filters: - Use convolution kernels to look for pixel regions with high local intensity gradients (contrast) Sobel filters: Gx Gy Reconstruct a full focus image from the pixels of best contrast in each of the image ‘slices’. The slice location provides height information for that pixel
Resultant images The capture system creates two datasets – the full focus image and the height data Height data well suited for standard machine vision analysis The capture system creates two datasets – the full focus image and the height data Height data well suited for standard machine vision analysis Extended depth of field Heightmap Coloured heightmap
Analysis procedure National Instruments Labview 7.1 / Vision 7.1
3-dimensional information Cursors allow calibrated dimension information to be read directly from the plot Results appear good Cursors allow calibrated dimension information to be read directly from the plot Results appear good
Short shot study Short shots produced on Battenfeld Microsystem Resin: BP Rigidex 5050 HDPE
Short shot component 2-d view is not easily able to spot incomplete filling. EDOF techniques can easily detect part filled components
3-D representation Microscope images 3-D image generated from heightmap and full focus data generated by EDOF system
Fracture/debris defect 2-dimensional data may miss surface defects such as this. EDOF technique clearly shows presence of undesirable surface properties 2-dimensional data may miss surface defects such as this. EDOF technique clearly shows presence of undesirable surface properties
Technique validation The process appears to provide reasonable results but comparison of results with other techniques gives confidence Products were imaged using a Wyko optical profiler and compared with EDOF data The process appears to provide reasonable results but comparison of results with other techniques gives confidence Products were imaged using a Wyko optical profiler and compared with EDOF data Wyko NT1100 uses white light Interference to generate high accuracy surface measurements Technique is slow and susceptible to mechanical and thermal instabilities making it unsuitable for at-process monitoring Wyko NT1100 uses white light Interference to generate high accuracy surface measurements Technique is slow and susceptible to mechanical and thermal instabilities making it unsuitable for at-process monitoring
Comparison of techniques EDOF technique WLI technique EDOF technique data shown above is unfiltered WLI system loses data where reflected light intensity is not sufficient for adequate interference fringes EDOF technique data shown above is unfiltered WLI system loses data where reflected light intensity is not sufficient for adequate interference fringes
Comparison of techniques Similar profile information, but EDOF technique shows errors at edges where peaks occur due to lighting
Comparison of techniques Good agreement between results with slight ‘tilt’ on WLI data Due to image flattening – different x-y planes Good agreement between results with slight ‘tilt’ on WLI data Due to image flattening – different x-y planes
Technique refinements For maximum accuracy within machine cycle time: 1-D high precision stepping stage Ring lighting/darkfield lighting High speed, high resolution camera/PCI express Rapid image processing Fast PC On-card processing For maximum accuracy within machine cycle time: 1-D high precision stepping stage Ring lighting/darkfield lighting High speed, high resolution camera/PCI express Rapid image processing Fast PC On-card processing
Vision system summary Single camera system capable of 3-D measurements Resolution ~ few µm Fast camera required to reduce acquisition times System allows 3-D manipulation of virtual product to verify moulding quality Single camera system capable of 3-D measurements Resolution ~ few µm Fast camera required to reduce acquisition times System allows 3-D manipulation of virtual product to verify moulding quality
The goal Data acquisition suite incorporating: Temperature measurement Piezo pressure measurement Ultrasonic measurements 3-d characterisation Allowing for evaluation of full process history of micromoulded products and enabling the determination of crucial parameters that influence micromoulding success Data acquisition suite incorporating: Temperature measurement Piezo pressure measurement Ultrasonic measurements 3-d characterisation Allowing for evaluation of full process history of micromoulded products and enabling the determination of crucial parameters that influence micromoulding success
Thank You Acknowledgements EPSRC Yorkshire Forward Members of the Micromoulding Interest Group Acknowledgements EPSRC Yorkshire Forward Members of the Micromoulding Interest Group
For more information and details about the Micromoulding Interest Group