The manufacturing industry can be overly optimistic when it comes to new technologies — believing AI will fast-track development and IoT-empowered factories are bound to catch up with the German and Swiss in precision manufacturing.
While management teams and consultants may spout trendy tech terms such as “process”, “automated”, or “smart”, and seem keen to hand over precisely-timed, measured tasks to machines, the truth is that AI tech is not a panacea — because manufacturing is hard.
Should Manufacturers Buy Costly New Equipment?
Data guru Professor Zhou Tao believes there are two paths for industrial smart equipment upgrades — “intrusive” or “non-intrusive”.
Intrusive upgrades involve rebuilding the production line, scrapping old equipment and replacing it with modern machines. Says Zhou, “Decade-old equipment can be a real pain source because they may contribute to 20 percent of product defects.” The downside of this approach is that companies may have to spend millions on new machines, gadgets and technical staff.
As the name suggests, non-intrusive upgrades involve working with old equipment, but installing data-monitoring sensors on key links, or leveraging existing CNC machine ports and log information.
One example of this approach is “pre-arcing”. If it takes 20 processing steps to make an LCD screen, then by step 12 the real-time monitor should know if the product is defective. “The problem may be high currency power or a malfunctioning gripping tip, and the system will suggest abandoning the semi-finished product with a 97% chance of it being defective,” explains Zhou.
As CEO of iCloud Union, a company providing data insights to industrial applications, Zhou saw a trend several years ago wherein company management opted for solution one: purchasing state-of-art equipment and ditching their still functional but outdated machines in order to get a tech boost. Zhou believes determining how to simultaneously smarten up machines, lower production costs, and boost production quality is a true trilemma problem.
A high-upvote response on Chinese question-and-answer website Zhihu presents another point of view against manufacturers simply upgrading to the latest smart hardware: “After touring factories, I realized that equipment in leading Chinese firms is not backward compared to their Europe counterparts, sometimes even better. The question is, those factories simply haven’t configured equipment, workers, and other costs before making themselves smart.”
Is Big Data the Answer?
If hardware is not the solution, could it be data? Maybe. A current complaint about data in manufacturing however is that much of it is unhelpful. In surveys conducted by Zhou, many respondents describe their data as either “poor quality” or “useless.”
Poor quality data can be defined as “fragmented, isolated info bits.” Although many factories are willing to invest money to mine data, there are no standard and constructive collection methods. For example, vibration data from a production line may be inaccurate due to out-of-date sensors, other data collection, etc. “It’s like having important data assets and not knowing how they can be used. In the end, factories just use the data to analyze the most basic stuff or store it in the dark,” says Zhou.
Real world problems are also complex. For example a production line for knives monitors steel wear and broken blades by collecting electrical current data through a Hall current sensor. The knife-making process also generates data on vibration, sound, pressure, etc. Producing different knife products also requires different data models. While electric current data can be used for most standard knives, specialty products may also require torque and vibration data, or even data captured by high-speed cameras.
Such data issues have less impact on major manufacturers like Foxconn and SAIC, whose new digital factories are designed to deal with multi-dimensional data such as sound, noise, temperature, humidity, torque, pressure, etc.
Humans Remain the Boss on Factory Floors
In order to replace human workers with bots, a factory needs to hire technicians — a transitional step that contradicts the “robots are coming for the jobs” adage.
In China there’s a huge shortage of experienced technicians for precision manufacturing, and many labor-intensive high-tech industries are having a hard time recruiting tech-savvy youngsters.
Without the help of capable frontline technicians, bots and sensors will not be able to operate effectively on a production line or in quality control. Zhou says he has a lot of respect for technicians helping manufacturers with data sorting.
Says Zhou, “Let’s say there are a total of five million pieces up for visual inspection. If one percent are defects, then we are talking about 50,000 pieces. And if the defects are rare, it is very hard for the machine to decipher the problem due to small sample sizes. Only experienced technicians can say for example ‘This was damaged by etching, and there are only three machines that can do that.’”
Data analysts work on disassociated Excel sheets where they deal with heaps of codes, for example “ZT17” and “FYK03.” Zhou notes that although both these defect codes indicate “small holes” on a product, a data analyst may be confused, whereas a frontline technician will know that ZT17 isn’t as serious because the hole is on a less critical area.
Smart Manufacturing Is Not All About Robots
When it comes to manufacturing, many tend to think that minimizing human error is the key to improving production quality and efficiency.
However, experts like Zhou know that the beauty of a high-quality, precision-made Swiss knife results not from cutting-edge tech or big data, but from a well-managed combination of good technology, good machines, and good human craftsmanship and oversight. This something that today’s factory owners should stop to consider in their rush to automation.
Source: Synced China
Localization: Meghan Han | Editor: Michael Sarazen