Machine learning (ML) and artificial intelligence (AI) continue to capture media attention and business investments. IDC estimates machine learning and AI spending to increase from $19.1 billion in 2018 to 
$52.2 billion by 2021.


With billions of dollars invested in machine learning and AI, it’s no surprise that tech giants Google, Microsoft and Amazon are investing billions in cloud infrastructure and development tools to accelerate the delivery of custom machine learning applications. Case in point, machine learning was the third-highest category for the number of patents granted between 2013 and 2017.

So, what exactly is machine learning?

Machine learning is a form of artificial intelligence, with algorithms that automatically recognize patterns in big data sets to derive meaningful insight and take action. What’s more, these algorithms are self-learning – they improve over time as more data is fed through their algorithms – without the need for explicit programming.

You may be wondering:

“How is machine learning relevant to manufacturing, and quality assurance in particular?”

As the manufacturing sector embraces industry 4.0 technologies to optimize production, we’re already seeing large scale deployments of ML-powered systems that improve production throughput while maintaining adequate quality levels and reducing costs.

This is crazy:

It’s estimated that the global smart manufacturing market –  which combines IoT and AI technologies in manufacturing – will reach a staggering $395 billion by 2025.

In this article, we will elaborate on the leading use cases of machine learning in quality management, specifically in the domains of predictive quality assurance and control, and downtime prevention.


The role of machine learning in quality management

Depending on the industry, the manufacturing process can be complex and prone to errors. For instance, automotive manufacturing has a completely different set of quality assurance challenges and complexities than, say, printed circuit board (PCB) manufacturing.

Although manufacturing processes across industries can widely vary, manufacturers share common operational goals centered around overall equipment effectiveness (OEE) and overall line efficiency (OLE).

OEE is calculated as follows:

OEE = Availability x Performance x Quality

And OLE is the weighted average of OEE per machine in a production line.

The importance of quality as a fundamental operational metric is clear. Let’s take a look at the two leading use cases of machine learning in minimizing production faults.


Predictive Quality Control

Quality control is integral to manufacturing processes, where defective products are weeded out from the rest, as early on in the production process as possible.


While quality control identifies defects, and prevents faulty products from reaching the market, the process of identifying the root causes that lead to the production of the defective products is time-consuming, and often requires multiple disciplines to collaborate – process engineering, quality assurance, mechanical and electronic engineering, to name a few – to collaborate.

In other words, quality control and its root cause analysis is costly and lengthy.

Let’s take food manufacturing as an example:

Food manufacturing requires stringent quality control measures at every step in the manufacturing process to ensure food safety. Over the years, regulations and standards have been imposed on food manufacturers by governments worldwide, making food quality control critical, and costly.

Good Manufacturing Practices (GMP), Hazard Analysis Critical Control Point (HACCP), Hazard Analysis Risk-based Preventive Controls (HARPC), Codex Alimentarius, and ISO 22000 – are just a few of the regulations and standards.

Here’s where machine learning can provide amazing value to quality assurance:

By implementing machine learning, machinery and product data can be monitored throughout the manufacturing process to predict quality faults – before they arise. Quality and maintenance teams are alerted, together with the precise root causes of the anticipated faults. This is the essence of Quality 4.0.

Integrating machine learning into the quality management process – often referred to as predictive quality – reduces quality issues and waste, cuts manufacturing costs, and minimizes product recalls to protect brand reputation.

What’s the bottom line?

It has been reported that machine learning can increase fault detection rates by up to 90% while slashing the time to pinpoint the root cause of quality issues to minutes from days.


Predictive Maintenance

According to recent studies, unplanned downtime of industrial machines costs manufacturers an estimated $50 billion each year. And equipment failure accounts for a whopping 42% of this unplanned downtime.

Predictive maintenance is the condition monitoring of equipment and predicting when maintenance is required based on leading indicators of fault development.

Advanced machine learning techniques are increasingly applied to optimize predictive maintenance outcomes. Such machine learning techniques involve modeling the relevant manufacturing processes and assets in a production line,  and then applying the most appropriate machine learning algorithms in the context of the production process and the specific products being produced.

Here’s the best part:

With machine learning, predictive maintenance algorithms do not have to be fed with preassigned threshold values, rather they is “trained” to recognize data patterns and anomalies from standard behaviors and historical faults. While initially, the algorithms have to be trained, over time they self-optimize with minimal human intervention.

And the end result:

Manufacturers predict unplanned downtime events, and prevent them from occurring, by taking corrective and timely action.


Final words

Maintaining high quality in production is a strategic goal for manufacturers since it directly affects the top and bottom lines for a company.

Integrating machine learning technologies into the quality management process can minimize product faults and reduce production costs. What’s more, as machine learning is a self-learning system, it promises to continually improve results.