The Six-Sigma approach is a business-management strategy that has been widely adopted in industry to improve the quality of products. It allows engineers to minimise variability in manufacturing processes and to identify and remove the causes of component defects. It aims to reduce manufacturing defects to 3.4 per million of any products manufactured.
But, despite rigorous adherence to such quality-control methods, systems built from such component parts can still fail under warranty once they are in the field and, when they do, customers return the products to a service centre or distributor where details of the fault are logged and service personnel attempt to discover the root cause of the failure.
However, in many cases, the fault experienced by the customer cannot be replicated at the service centre and the product that has been returned operates properly when tested. Such problems are labelled ’no defect found’ problems and, because they result in a new replacement product being sent to the customer under warranty, they can seriously erode the profit margins of manufacturers.
Clearly, however, if such ’no defect found’ problems could be reduced by even a half, OEMs could dramatically reduce their internal processing costs and use the money saved more productively in research and development or reducing the cost of their products.
Isolating the root cause of such faults, however, is not trivial. Such quality problems are significantly more difficult to identify than issues incurred in the manufacturing process.
If ’no defect found’ problems could be reduced by even a half, OEMS could dramatically cut their costs
For each type of fault that arises in a given manufacturing process, engineers can statistically analyse large sets of data sampled from the process to identify the root cause of a fault. But with ’no defect found’ failures, they are faced with analysing a much larger number of variables or parameters and a very small sample set for each type of failure that occurs.
Clearly though, to keep warranty costs under control, manufacturers would still prefer to be able to trace every single warranty claim back to its root cause. However, because of the diverse number of problems that need to be addressed, dealing with each one individually would require a large resource of engineers to investigate the causes of each one.
Instead of performing such individual root-cause analyses, many manufacturers analyse and prioritise the reports of the different types of failures that are returned from the service centres. After this, in-house design teams attempt to address the failure issues by developing and performing tests to reproduce and analyse the failure scenarios; though this is also a time-consuming and costly business.
To address the issue, Prof Darek Ceglarek from Warwick Manufacturing Group’s (WMG’s) Digilab has created a suite of software tools that can help manufacturers to define the best means to address the root causes of ’no defect found’ problems, going beyond identifying Six-Sigma problems in manufacturing to determine the reason that such faults are occurring.
The software Prof Ceglarek developed is now being commercialised by Warwick Analytics, a recent university spin-out he formed with industry veteran and managing director Tim Blee.
’The computational-simulation software allows manufacturers to study the most unlikely combinations that could result in service failure of product due to issues encountered in a different environments of product-lifecycle management, such as design, manufacturing, service or a combination of these. A root-cause analysis can be carried out using the software - using data received from service centres and taken from the individual manufacturing processes - and faults can be traced to individual components within complex systems,’ said Prof Ceglarek.
He also said that the software itself performs two essential functions: first, it allows manufacturers to determine the root-cause analysis of product service failure and, second, it performs a data-mining operation on the data supplied from the service centres to uncover hidden patterns from it to pinpoint the key parameters of the faulty product.
“The software has the potential to save automotive firms hundreds of millions of pounds”
TIM BLEE, WARWICK ANALYTICS
Then, a process-mining operation maps these parameters into a multi-dimensional set of data acquired from many manufacturing processes to uncover the critical relationships or associations between the two sets of data that can narrow the cause of a faulty product down to specific manufacturing processes, explaining the failure.
According to Prof Ceglarek, the data mining is performed by a generic set of algorithms that can be applied to any manufacturing environment, while the process-mining algorithms must be tailored to work with data acquired from individual companies’ manufacturing processes. As such, the software is not a generic ’out of the box’ solution to ’no defect found’ issues.
Nevertheless, according to Tim Blee, there is no lack of data from manufacturers that can be analysed by the software. Most Fortune 100 manufacturers collect and store large amounts of service data in enormous databases, as well as amassing large amounts of data from theirs and their third-party suppliers’ manufacturing processes.
Aside from allowing manufacturers to trace warranty claims back to their root cause, the software can be used to trace problems arising from recall issues in the automotive business.
’While automotive assemblies perform to specification 99.9 per cent of the time, in the field there is a possibility that the marginal performance of a part in an assembly could become unsafe under certain conditions, leading to a recall to isolate and replace defective items. In such a scenario, performing a root-cause analysis is the only option open to a manufacturer and, here, our software can help to unearth the combinations of production processes that could be the cause of such a fault,’ said Blee.
The software has the potential to save automotive firms hundreds and thousands of investigative man hours and hundreds of millions of pounds, Blee added, by avoiding costly and embarrassing product recalls such as the ones that afflicted Toyota last year.
Prof Ceglarek believes that, in the longer term, the software might serve as a useful tool for design engineers too, by allowing them to play ’what if’ scenarios while developing a product - simulating potential failures that might arise even before a product is built and providing them with a reliable method to improve their designs.
Hopefully, the use of Warwick Analytics’ technology in the design and manufacturing environment will bring the day closer when warranty claims for faulty products and automotive recalls are a truly a thing of the past.
Glasgow trial explores AR cues for autonomous road safety
They've ploughed into a few vulnerable road users in the past. Making that less likely will make it spectacularly easy to stop the traffic for...