MENU

Big data and the human side of manufacturing

Feature articles |
By Rich Pell


Physical equipment can now have a ‘Digital Twin’ – a virtual representation of itself – which is able to inform predictions and subsequent processes. Automation is at an all-time high in terms of decision making and process control.

Subsequently, we have more data than ever with which to inform business decisions. From machines at the heart of the manufacturing process to incidental mechanisms like the supply chain or transportation, big data is providing the basis for better and quicker strategic decisions.

The potential that big data has to make operations more cost-effective is obvious. A 2017 survey from management consultancy McKinsey&Company suggested that the implementation of Big Data in manufacturing could boost pre-tax margins by 4-10%, enhancing everything from machine life to increased output.

The seemingly obvious conclusion — which is that you should push the benefits of big data to the maximum, quantifying and automating as much as you can — is not the case. The most effective enterprises will recognize the limitations of Industry 4.0 and continue to value the expert on the manufacturing floor, marrying individual intuition with automation.


Man vs machine

The cliché that automation can lead to the total removal of the engineer from the manufacturing floor is a pipedream, at least for our lifetime – we would need far more sophisticated AI mechanisms to make this a reality. The most effective digitalisation that we can implement right now remains at least partially reliant on the boots on the ground. No matter how many metrics at your disposal, there are always insights that human experience, expertise and intuition can offer that won’t be picked up by digital measurements.

For example, in virtually every line of manufacturing, the machines are unique; built to the same specifications but with tiny individual differences. Parts will have unique wear and tear, produce different sounds due to being in different areas of the factory floor, and so on. Big data drawn from these machines is not going to recognise these differences, which can lead to inexplicable differences in the data results.

If an operator has been working with an individual machine for long enough, he can feel whether or not a machine is working through vibrations, noises, appearance, etc. Data isn’t capable of replicating this or providing the context for it, and in many cases an operator’s annotations of a data sheet may offer greater insight than further digital analysis.

Utilizing dark data

It’s also true that, even with modern data analysis techniques, the sheer volume of data that a manufacturer produces is too much to use. Dark data — data that you record and don’t use, or that isn’t recorded at all — can’t contribute to the insights that an analyst is trying to glean.

Many companies aren’t even aware of the dark data they store, whereas others simply log it and forget it until a point at which they can make use of it. Given that IBM estimates that 90% of the data generated by sensors and analog-to-digital converters is never used, and that most companies only analyse 1% of the data, huge opportunities for further insight are being passed up by failing to utilise this resource.


Again, this is where human interpretation and intuition is capable of making the difference. Data scientists can offer an entirely new perspective, bringing light to dark data by reframing it in more accessible formats. They can ask the right questions, align the data of interest, and clean the results to make them more useful to decision makers; without human inputs to define the right context, you’re not going to maximise the utility of your data.

Finally, the unpredictability of human interference can also be difficult for data analytics alone to diagnose. The parameters of data analysis are limited to things directly related to a machine. They won’t, for example, explain how other human processes may disrupt things like performance, or even the analytics process itself – you’ll need to work that out for yourself.

For example, we have previously worked with an automotive manufacturer that found the wireless system used to underpin IoT communications on the manufacturing floor would regularly drop out during the same period every morning. The data showed the loss of connectivity, but it took human intervention to identify the problem; the network was disrupted every time an employee used the microwave to heat their breakfast!

All of these examples demonstrate the importance of the individual engineer, and the impact that they can have on the overall profitability of a manufacturing business. A talented individual is capable of filling in the gaps in our current data analysis; can make the most of the data that we fail to understand or use at all; and is capable of understanding the behaviour of his/her co-workers more than any machine. The manufacturers who want to run the most insightful and cost-effective operations cannot underestimate the influence that the individual can have on both profit margins and internal processes.

About the author:

Heinz Linsmaier is CEO at camLine – www.camline.com


Share:

Linked Articles
10s