Automated systems reduce production waste

9th March 2018

The implementation at one of the largest foundries in the South Africa of supervised and unsupervised machine learning capabilities produced a 0% external scrap rate, saving the manufacturer as much as R10-million a month, explains business management consultant Dataprophet cofounder Daniel Schwartzkopff.

He explains that not a single defective casting was shipped from the foundry. This has enabled the foundry to be able to reinvest savings into boosting productivity further, while also enabling the executive team to bid for projects against leading manufacturers globally.

“These are the building blocks of creating a globally competitive manufacturing industry. However, all of this requires that the human element interfacing with the technology is sufficiently skilled to understand the insights gained from the data, make the necessary adjustments, and track improvements and changes in production to optimise yield,” he says.

He highlights that, in South Africa, a lack of such skills is undermining industrywide efforts to revitalise the local manufacturing sector, while global manufacturing leaders are powering ahead with a lean but highly skilled workforce.

He states that machine learning is a transformative technology that can bring the South African manufacturing sector up to speed with the rest of the world. However, without the requisite skills, manufacturers will struggle to realise its full potential.

“The sector should prioritise skills development over mass job creation, and work with experienced technology partners who can deliver rapid productivity gains while helping build a roadmap to survive the Fourth Industrial Revolution,” he says.

Schwartzkopff highlights that to reignite the manufacturing sector, industry bodies have set out ambitious growth targets. Industry representative body the Manufacturing Circle has launched its ‘Map to a Million New Jobs in a Decade’ plan. It aims to expand manufacturing’s contribution to the country’s growth domestic product (GDP) from 13% to 30% and create between 800 000 and 1.1-million direct jobs, as well as five to eight times as many indirect jobs in the process.

“However, aside from a single line about focusing on education as an essential service and promoting high standards instead of high pass rates, there seems to be little to no acknowledgement that for South Africa to become a globally competitive manufacturing hub, it needs skills that can interface with the technologies driving global manufacturing progress,” he stresses.

Mass Employer
He explains that modern manufacturing is capital intensive, with a focus on using technology to improve production. New, more efficient production technology requires fewer, but much more highly skilled workers to operate.

“The Minister of Trade and Industry, who, until recently, was touting manufacturing as a key job growth enabler, has acknowledged the disruptive changes in the nature of production due to technology. At the recent ANC policy conference, the Minister stated there will be data management and digital firms that will be at the apex of value chains,” he explains.

He goes on to advise that countries need high-quality public education for a well-functioning manufacturing sector, highlighting how, over the past 250 years, the global increases in our standard of living can mostly be ascribed to improvements in learning.

“Here, government should focus on the so-called science, technology, engineering and maths fields to equip workers with the skills they need to successfully interface with the broad range of exponential technologies entering our world,” he says.

Modern manufacturers are all focused on three key priorities: speeding up production, improving the quality of manufacturing output and reducing cost. Further, he states, machine learning and artificial intelligence are emerging as the go-to technologies for driving enhanced process efficiencies while lowering operational costs and maintaining global quality standards.

“Quality engineering methodologies such as Six Sigma programmes are valuable, but the increasing complexity of quality process monitoring requires a level of sophistication that these techniques cannot match,” he explains.

Schwartzkopff mentions that machine learning can serve as a powerful tool in large, sophisticated manufacturing operations. Its power lies in the ability to examine process data and extract patterns and relationships without imposing any adjustments to the production process, the outcomes of which can augment production processes to achieve higher yields.

He highlights that, while most manufacturing operations are different, all have rudimentary data infrastructure that can be processed and mined for value by skilled data scientists. A single view of all operational data helps to reduce the complexity of managing multiple types of data and assists in the development of a set of prescriptive models that can help manufacturing processes move toward deeper forms of automation.

Schwartzkopff concludes that the manufacturing industry’s role in the South African economy has seen a long and sustained decline from contributing nearly a quarter of GDP in the 1980s, the sector added a mere 13% to the country’s GDP in 2016. Job creation has also not materialised: after 400 000 jobs were lost following the 2008 global financial crisis, the sector failed to recover and today employs 300 000 fewer people than in 2008.