Skip to main content Skip to navigation

Success Story Assetbank

Select tags to filter on

Machine learning project automates complex tooling job at Expert Technologies Group

A two-year Knowledge Transfer Partnership (KTP) between WMG, University of Warwick and expert industrial automation specialist Expert Technologies GroupLink opens in a new window has introduced machine learning into a range of engineering settings that were previously thought too complex to implement. KTP Associate Mateusz Ogrodnik reduced the time of designing new mechanical fixture plans* from weeks to minutes.

Challenge

Historically the job of creating mechanical fixture plans was deemed too complex to automate. Despite involving repetitive tasks, the processes were too nuanced for simple automations to handle, and tooling designers were still an essential component.

However Expert Technologies were keen to devise and implement a novel approach to this unsolved problem, so that designers’ time could be better utilised on higher value, strategic tasks in the business.

The key challenge was the sheer computational complexity of this process. A typical fixture plan has around 40 elements which would mean that around 250 parameters had to be implemented into an automated solution.

As part of this project, Expert wanted to create and embed a systematic approach to solving complex engineering problems for the future. With the use of machine learning on an upward curve, this project would pave the way for them to continue to use this kind of technology for other hugely complex problems.

Solution

With the support of WMG, they opted to apply for an Innovate UK funded Knowledge Transfer Partnership (KTP). KTPs link forward thinking businesses with a knowledge base to deliver on an innovative project led by inspired graduates.

Initially, Knowledge Transfer Associate (KTA), Cesar Reyes, was appointed. His work laid strong foundations and a year later, a second KTA, Mateusz Ogrodnik joined the team. His approach, thanks to experience in optimisation and software development, along with support from WMG's Professor Pasquale FranciosaLink opens in a new window, led to the development of a completely novel set of algorithms that reduced the computational complexity.

Mateusz arranged meetings with senior engineers at Expert to understand how an experienced designer would approach the problem which helped to speed up the time it took for the automated system to create a design. This was followed by building a prototype app, where multiple case studies could automatically be tested. As the number of bespoke algorithms grew, and time left to create a solution decreased, Mateusz spent time working directly with the engineers to undertake further beta testing to arrive at the final prototypes. Mateusz also attended the Association for the Advancement of Artificial Intelligence conference in Canada, where he was able to gather further insights for this project as well as other opportunities to take forward at Expert.

Impacts

· A final prototype tool was created which reduced the time of developing mechanical fixture plans from weeks to hours which will lead to significant cost savings for the business.

· Mateusz has been hired by Expert as a Senior Software Development Engineer.

· Expert has put new software development approaches into practice which will de-risk other software-based projects.

· The project opened up a number of ideas for other Expert employees to build, develop and share within the company.

· New software projects, including complex machine learning tasks have been brought forward and de-risked thanks to this project.

· Expert has now applied for a further Accelerated Knowledge Transfer grant to work on a similar project.

Ian Snape, Engineering Director at Expert Technologies Group said:

“To witness the ground-breaking final prototype demonstration was a realisation of an unrealistic idea four years ago. Not accepting the ‘norm’ and embracing development failure as an opportunity to dissect and optimise, is the heartbeat of this achievement. We celebrate the powerful new fusion of engineering and data science within this Expert / WMG collaboration and look forward to the next set of limitations we rip up.” 

WMG Reader and academic supervisor on this project Pasquale Franciosa said: 

“The work with Expert Technologies Group through the Knowledge Transfer Partnership root is great evidence of transferring academic research outputs into industrial operation. The project nicely bridged engineering design with digital technologies and proved that this is the way forward to support business growth, whilst releasing capacity and maximising productivity.”

For more information on working with WMG on KTP projects contact wmgsme@warwick.ac.uk

For more information on Expert Technologies Group visit:

Expert Technologies Group | Industrial Automation Technology

*A fixture is a device for locating, holding and supporting a workpiece during a manufacturing operation. Fixtures are essential elements of production processes as they are required in most automated manufacturing, inspection, and assembly operations.

Thu 20 Jun 2024, 21:21 | Tags: Digital SME SME-Digital SME-Growth SME-KTP Success Story

Digital solutions leave seat manufacturer sitting pretty

Car manufacturing in the UK is changing radically, as the electrification is forcing innovation and design change of almost all parts of cars. Industry leaders are challenged to change design radically, launch new products more quickly, while maintaining quality levels and costs. To keep up with this, many manufacturers in the supply chain are already embracing digital technologies to connect their factories, improve processes and reduce costs. By embracing digitalisation, the electrification of the automotive sector could unlock growth of up to £75bn by 2035, benefitting the entire supply chain.

Using digital to develop manufacturing

Lear Corporation are a leading supplier of automotive seats and electrical systems, and their customers include major OEMs such as Jaguar Land Rover. With a long history in luxury and performance seating, they are one of the world’s largest providers of premium automotive leather seats, operating over 250 facilities across 38 countries, including 3 Just in Time final assembly plants here in the UK.

Employee working on a car seat.

With safety and performance paramount, Lear must ensure that every seat is of the highest quality but working with comfort materials, such as foam, and natural materials, such as leather, brings with it a number of challenges. Not least, the natural defects that occur in these materials, such as wrinkles and discolouration, which must be rectified before products are ready for the customer.

Employee working on a car seat.

Back in 2017, Andrew Williams, Director, Process Innovation at Lear Corporation was looking for ways to deploy digital technology to improve their seat inspection process, reducing waste and rework. “Seat inspection takes place at the end of the production line, and it’s a manual process which relies on the judgement of experienced operatives,” Andrew said. “But ultimately, it is subjective, we set our standards very high to protect our customer but human judgement will always have an element of variation.”

Red leather car seat.

Lear wanted to innovate and deploy smart manufacturing solutions that would automate the inspection process, providing increased consistency and generating quantitative product data to the engineers in our relentless pursuit of perfect quality. Lear approached WMG as a market study showed a lack of readily available commercial and deployable solution that suited Lear’s needs at that time.

“We already had a relationship with WMG with apprentice, undergraduate and post graduate training programs. Lear’s philosophy is to constantly look for ways to use new technology to improve manufacturing processes and knowing WMG were all about industry-focused research, we knew they would be a good fit for this challenge.”

Andrew teamed up with our Automation Systems Group on the Lear Seat Manufacturing (LSM) project, supported by funding from WMG Centre HVM Catapult.

Exploring the possibilities

After scoping the project carefully, WMG set about building and testing machine learning-based object detection systems as an inspection tool; Using image capture systems mounted on a seat inspection station, the solution would be capable of detecting and classifying defects, and of assessing whether the seat would be acceptable against the customer standards, or whether additional rework is needed.

Developing the solution and training the AI, involved manually labelling a large data set. To facilitate this process, the team developed a “digital marker” that allows an in-line, in-process labelling workflow.

Dr Dan Vera, Associate Professor at WMG, said: “To train the AI inspection system, you need to tell it what defects are and what they look like. You have to label defects so it learns. Typically, labelling is done offline, after images are captured. With the labelling system developed at WMG, operators can digitally label defects on the real seat, and directly on the production line – we call it the magic wand!”

In order to test how the different elements would work together in a real manufacturing environment, we installed a prototype inspection station in the National Automotive Innovation Centre (NAIC) at WMG before taking the station into Lear’s plants for evaluation in production.

White leather car seat.

Looking to the future

The machine-learning seat inspection software was trialled at two of their factory sites, allowing it to collect data from different types of seat and improve its ability to recognise and categorise all types of defect. It is now on track to deploy across our plants worldwide.

Andrew said that without the help of the team at WMG, supported by funding from HVM Catapult and Innovate UK, they might have missed this opportunity. “This was unchartered territory,” he reflects. “Because this technology was so new, and there were no commercially available solutions at the time, we would not have had the bandwidth to develop this in house. The cumulative effect of these projects has the potential to significantly increase the consistency of our inspection processes and provide our customer with even higher quality products.”

Lear has showcased the test bed in NAIC to some of their key customers, who have been impressed with the outcomes of their work. There is a continued appetite within Lear to develop new ways to use digital technology to improve their manufacturing processes.

For more information about this project, or working with us, please email wmgbusiness@warwick.ac.uk.


New lab ‘must-have’ wireless sensor stirs up new opportunities for data collection

Challenge

Manufacturing novel advanced materials, polymers, and medicines push innovation forward and help to solve problems of health, pollution, security, and clean energy. The discovery of innovative products often requires a vision backed up with months and years of routine work. The human is crucial in generating the vision; but the human factor can cause errors, irreproducibility, and side-track the discovery in routine work. The past pandemic and lockdown have delayed innovation and laboratory research. A slowing economy decreases R&D investment, which further puts downward pressure on future economic growth and risks undermining the UK's leading position in materials discovery and innovation.

Monitoring of process parameters is a necessary part of modern production systems, mechanical engineering and science. With the increasing complexity of processes in today's industry, information gathering, and data acquisition can also become more complex using multiple instruments, increasing the project's overall cost. Data collection can be done in various ways, and WMG at the University of Warwick wanted to offer an innovative multi-sensor wireless data acquisition tool that can be deployed across industries to improve manufacturing accuracy and efficiency.

Solution

The researcher, Dr Dmitry Isakov, an Assistant Professor at WMG and his team developed and tested the device (called Smart Stirrer) that comprises a multi-functional sensor system, midrange Bluetooth module and System on Chip (SoC) and protecting housing manufactured in the form of a conventional magnetic stirrer bar. The team recognised that magnetic stirrers could be the best vehicle for a sensor that captures multiple parameters. Through a series of experiments, it has been demonstrated the performance of the Smart Stirrer capable of accurate in-situ monitoring of the physical properties of the chemical reaction, such as temperature, conductivity, visible-light spectrum, opaqueness, and viscosity. A prototype with components cost < £35 has demonstrated excellent work in laboratory conditions when conducting titration and oscillation reactions.

The functionality of the Smart Stirrer is implemented using open-source code in accordance with the user's needs. The team is currently working on the implementation of the Smart Stirrer into the IoT platform thus enabling even further to extend the application of the device.

Impact

Using an array of sensors combined with a low-cost wireless transmitter, all in a small chemically resistant package creates a powerful measuring device placed in-situ into the reactor and operating in conjunction with data storage on the cloud. The data processing by machine learning algorithms will facilitate the task of reproducibility of results and reduce the influence of the human factor. The feedback system will simplify this task using software algorithms and robotizing the production process.

Dmitry Isakov, WMG, commented, “The beauty of the Smart Stirrer is that it can be used everywhere, such as a sealed vessels thus minimising the contamination of the reactor. It may give a push to new discoveries as well. It is easy to integrate the stirrer into the labware family and make it “speak” to other lab equipment.”

Samuel Baldwin, from the Mathematics institute at the University of Warwick worked on the smart stirrer during his WMG summer internship, he added: “We have leveraged state-of-the-art technology to build a device with very low power consumption, a broad range of sensor capabilities, and high data-throughput over the Bluetooth Low Energy platform.

The laboratory of the future is that of automation, reproducibility, and safety; our all-in-one Smart Stirrer device eliminates the need for a vast array of individual wired sensors whilst maintaining the control and customisability that one would expect from any piece of advanced laboratory equipment. I look forward to seeing the Smart Stirrer solve laboratory problems and help us understand complex reactions.”

The Smart Stirrer can be an essential element in the future digital laboratory platform – a concept currently developed by the team. Digital laboratory would allow interaction with laboratory equipment (hot plates, switches, pumps) using single-board computers making such equipment “smart” by replacing the manual control with electronic (while using the same manual equipment). Such a platform will provide an unprecedented level of detail and enable monitoring of experimental tasks remotely, thus contributing to the continuation of progress in innovation and research.

The future of this project is to gain further insights into how this product can benefit end-users. The product is currently available for lease and will be available for purchase in the future. The team are keen to learn to more about improving the product for machine learning and to create the most useful device. For more information, email wmgbusiness@warwick.ac.uk.

Monitoring Chemistry In Situ with a Smart Stirrer: A Magnetic Stirrer Bar with an Integrated Process Monitoring System

Nikolay Cherkasov*, Samuel Baldwin, Gregory J. Gibbons, and Dmitry Isakov*