Automated Optimisation of Production Lines Scheduling Problems

Research Services
in Manufacturing
Improvements in manufacturing
processes through advanced ICT
Key Experts:
Dr Ernest Cachia
Dr John Abela
Dr Ing Saviour Zammit
Mr Joseph Bonello
Mr Armand Sciberras
Process 1 – Automated
Optimisation of
Production Lines
Scheduling Problems
• Scheduling problems are proven to be a class of
very difficult problems in computing
• The project’s aim was to create a bridge between
academia and industry
– By finding common ground with respect to
– By providing a tool to
automate understanding
of scheduling problems
using already available
Scheduling Framework
• Extendible Framework for Scheduling
• Modelling Production Lines using an
easy-to-use interface
• Detection of hard sub-problems from
the given model
• Feasible schedule generation with a
number of algorithms and heuristics
Modelling and Detection
• Sometimes industry and academia refer to
the same concepts differently
– A Unified approach was developed for dealing
with scheduling problems using available data
– Our approach is based on a graphical
• Modelling allows us to
find out
‒ which problems
require academic
‒ help us find existing
• Providing a generic optimisation tool for diverse
production line is very difficult
– The focus is on scheduling orders, not planning
– Tools must be adapted for individual production
• Suitable for fast scheduling and “what-if”
‒ Example in case of
machine breakdowns or
for configuration changes
• Tools Provided
– Optimisation of Single Machine problems
• Very common for SME manufacturers
– Optimisation of Job Shop Problems
• Very common with Multi-Stage Production Lines
• Handling of Parallel Sequences
‒ When some sub tasks are done in
parallel to the main task to speed
up process (e.g. preparation of
basic components)
• Handling to setup time between
different jobs
‒ Machine Reconfiguration
• Support for release dates and
Process 2 – Automated QA
for Print Output
using Neural Networks
Automated Optical Inspection Motivation
• Quality is a key factor in the selling price of
manufactured goods
• Quality inspection is traditionally carried out by
human beings either:
– At the end of the production line on each and every
item removing defective products (slow and
– Or, inspecting samples from a produced batch and
assessing a batch quality (defective products are left
in batch)
• Although some automated off-the-shelf tools exist,
these are usually too general and fail to identify
certain defects
Missing or incorrectly placed components
Scratches and cracks
Print offset, missing/extra transfer of ink
Smudges and misalignment defects
caused by multiple printing on location
An Automated Inspection System
• Reliable 24x7
• Accurate
• Fast
• Consistent
• Cheap
Automated Optical Inspection
Image correction,
removal, alignment,
Output from
Data Input to
Classifier and other
QA tests
Further image
processing and
feature extraction
Setting up the system
• The system can be configured
to perform a number of QA
tests depending on the product
and type of defects
• A product representation is built by the system
based on one or more of the following
• A single product and some parameters
• A number of defect free samples
• Both defect free and defective samples
The two steps of AOI
• System is split up into two main modules
– Image Processing
• Alignment
• Background Removal
• Segmentation
– Inspection
• Feature Selection
• Colour Inclusion
• Label location
• Neural Networks (WISARDs)
Practical use of ICT in
Manufacturing – A Case Study
Mr Olaf Zahra
Toly Products Malta
• Toly Products is cosmetic packaging solution provider for the
beauty industry.
• Toly Products’ has been manufacturing in Malta for the past
40 years. Today the majority of corporate function are also
located her.
• Our factory here (and in other countries) operate on lean
manufacturing principles. We manufacture only to order on
the date the product is required
• We produce between 1.5 to 2 million units per week
• We have over 2,500 different live products
Process 1 – Automated
Optimisation of
Production Lines
Research Benefit
• Scheduling is a critical function:
– Impacts manufacturing efficiencies.
– Impacts delivery and customer satisfaction
– Impacts reputation if deadlines are missed
• Good scheduling requires
– Skilled operators who know the processes well
– Involvement of several key departments and functions/
– Good regular feedback from the production floor.
• Since 2008 this criticality has increased exponentially. In a
downturn, organisations try and release cash and in our industry
this was done by reducing stocks. As the market recovered the
new limits set on stockholding have been maintained. For an
upstream manufacturer this means shorter runs and more
frequent changeovers.
Major Challenges
• A lean manufacturing set up requires components to be
continuously in motion from the beginning to end of the process to
avoid bottlenecks and the build up of stocks. It also requires
materials to be delivered to the production floor when and as
• A large variety of products for different clients
– A bewildering combination of mould tools and moulding
machines that need to be configured for different jobs
– Finding a good balance between technical and human
constraints and meeting client demand
– Delivering to the customer On Time and In Full.
• The aim is to
– Minimisation of downtime
– Minimisation of work in progress
– Maximising resource utilisation
Exploring Possibilities
• Achieving better results while maintaining
a flexible manufacturing environment
– Explore different possibilities of
allocating resources to manufacturing
– Gain future insight to help sales and
marketing personnel plan orders
– Work around scenario changes in cases
of schedule disruptions (e.g. urgent
Achieving the target
• Having several scheduling possibilities allows for
more accurate planning.
• Functions have better visibility of the tasks they
are required to perform to service the plan.
• Minimising job changeovers reduces delays and
maximises overall plant efficiency
• Some of the benefits offered by the
– Effective planning tool
– New approach to scheduling
– Cost Effective
– Fast Results
– Configurable
– Low training requirements
– Uses information we already have and know
Process 2 – Automated QA
for Print Output
using Neural Networks
Visual Inspection at Toly
• Production of luxury cosmetic packaging for the
beauty industry.
• Visual quality is of utter importance
• Even using state of the art machinery there are a
multitude of parameters therefore variation is a fact of
life. Inspection needs to identify this variation before it
exceeds the permissible tolerance so that defectives
are avoided and the process can quickly be brought
back in line.
• Visual defects consist of
– Bad Printing (location, missing/extra ink)
– Incorrect assembly.
– Scratches, cracks or smudges
Challenges in the inspection process
Currently visual inspection is carried our in one of two ways
• Manual Inspection
– Advantages
• The inspector can identify and classify small variations from product to product
and take a call; i.e. is flexible
– Drawbacks
• Current (at Toly) camera based automated inspection technology
– Advantages
• Accurate
• Consistent
• Fast
– Drawbacks
• Limited application.
• Inflexible; it works with strictly defined parameters therefore typically will reject
numerous good parts with the bad parts
Using the system
• The system is configured with few
parameters and provided with a number of
defect free products.
• System then trains and builds a multilayered model for that particular product.
• The model is saved and used every time
that the corresponding product is
manufactured avoiding the need of
Benefits of the system?
• The system is able to identify defects on our
products from several good and several bad
images. Therefore it builds up a memory of
good parts which allow it to widen it’s
tolerances over traditional systems.
• It is applicable to a wide range of products
and also a wide range of defect possibilities.
• Easy to set up and easy to update once a
new defect has been encountered.
Summing it all up
• The system is preferred over manual
– Reliable
– Avoids manual contact of the product
– More accurate
– Faster
– Consistent
Thank you for

similar documents