3 Big Data Challenges for Manufacturers and How to Solve Them


Big data challenges are often reserved for Fortune 500 or brand new companies in the modern manufacturing world. They either have extensive resources or don’t have to worry about disrupting existing infrastructure and so can hire data scientists to track and garner insights from metrics. However, the vast majority of manufacturers are small to medium enterprises (SMEs) that have been around for decades and have to operate within tight budgets. Arjun Chandar, Founder and CEO of IndustrialML, shares strategies to solve manufacturers’ three most prevalent Big Data challenges.

Out of 600,000 SME manufacturers surveyed in the US, the average age of the machinery/assetsOpens a new window was often over ten years old. And the US Census Bureau from 2017 indicated that there were a little over 36,000 Food and Beverage factories in the country owned by just over 31,000 companies – so SMEs still dominate the manufacturing landscape.

When it comes to industrial IoT for manufacturing, the public, VCs, and even software providers often assume that the manufacturers themselves have a certain infrastructure already in place (connectivity to the cloud or IT infrastructure, the sensors on the plant floor that’s already creating the data, or even just knowledge of what they want to measure). However, most manufacturers need something out-of-the-box that improves their bottom line even in the absence of that infrastructure going in.

Challenges vs. Solutions

Let’s look at three key data challenges manufacturers face and how IoT solutions can be highly effective.

1. Computer Vision

Production errors are a big problem for any manufacturer. One of IndustrialML’s partners, Daiwa Steel Tube Industries, recently installed a laser marker and a “lot tracking system” that puts a barcode on the product and can use real-time computer vision to track tube quality. As a practical example: a piece of steel numbered 12345 is stamped with the barcode at 4:57:32 PM. Just two seconds later, at 4:57:34 PM, the computer vision algorithm detects an error. The IndustrialML system can then correlate the error to product 12345 and alert the factory operator to take immediate action to scrap the product. The cameras detecting these quality issues oversee the line and require no disruption to be installed.

Real-time computer vision is an example of how companies can use the help of IoT platforms to work from the small amount of data they have and even acquire new data in a way that doesn’t impact the manufacturing processes that they have successfully established.

2. Process Control

It’s straightforward for industrial IoT players to say – “we can do calculations in real-time.” But a manufacturer needs to take that extra step of precisely understanding their problems to be solved in real-time. This is where process control is crucial.

For example, a silicon wafer manufacturer might have robust automated systems for tracking their wafers’ dimensional accuracy and location. That is a helpful technology and the most expensive part of the process, but it can’t succeed on its own. Manufacturers also need to know the average, upper and lower control limits. They can then gain better insights from learning the number of wafers they produce outside of that limit.

On top of that, this information needs to be communicated appropriately with real-time alerts. For example, if the machinery produces five consecutive chips above the specified average diameter, the operator needs to know immediately. In many cases, for smaller manufacturers, this information never gets back to the operator, so it is too late to make the necessary changes before generating thousands of dollars of scrap.

Industrial IoT companies often rely on some measure of change in the operator’s process from their initiative. If you can give the operators specific information about their existing process in a way that they can act on, that’s far more useful.

3. Communication Is the Answer to Unharnessed Data

Very wealthy and large manufacturers often have data scientists who can process a lot of information to develop significant trends over time. But that’s not the same as getting the information to the people that are influencing the product as it’s happening. This creates a bit of a disconnect between data science and actual production operators. And one of the biggest challenges that industrial IoT needs to target is bringing either basic or advanced data science to the production floor right away.

One answer is to create a dialogue by capturing the audio from the production floor. For example, an alert might show that a 10-second clip was recorded from a particular headset, and the speech-to-text translation would then confirm that the operator spotted an error. Over time, the IndustrialML platform can correlate circumstances with the operators’ actions on the production floor and adjust their alert recommendations accordingly.

A Big Step Forward

In whichever area manufacturers face data challenges, be it communication, product monitoring, or processes, some solutions don’t require a complete overhaul. This should encourage SMEs, who previously could not solve these critical pain points. The future holds a lot more potential with the quick leaps in technology,