Egicon, an Italian electronics manufacturer, used emerging data analytics tools to fully automate its production starting in 2017. Along the way, it cut repair rates by 80 percent, eliminating scrap, improving warranty support, and reducing lead time for quality reporting to real time from one month.
Based in the Modena region of Italy, Egicon produces electronic control units, instrument clusters and human-machine interfaces for the automotive, agricultural, biomedical and aerospace sectors.
Egicon integrated Siemens’ Valor and Opcenter Execution Electronics IoT software into its production and quality systems, allowing continuous monitoring and the ability to provide customers with better warranty support and traceability data.
“We were able to reduce our repair rate from 30 parts per million to 6, and achieved a scrap rate of zero percent in 2019,” Egicon production manager Michele Magri said in a published case study. “Now I can get instantaneous updates on all our manufacturing processes without leaving my desks. I can spend my time on innovation and improvements.”
Other software manufacturers also report significant results.
GE Digital’s Proficy software has helped manufacturers across numerous sectors achieve a multitude of benefits, including 90 percent reduction in waste, $5 million in quality improvement savings, and 80 percent decrease in downtime, Cobus van Heerden, senior product manager for analytics and machine learning software for GE Digital, said. One company gained key insight within hours on how to control their de-watering chemicals to get the best quality.
FactoryTalk Innovation Suite, a joint offering from Rockwell Automation and PTC, helped Rockwell achieve a 33-percent increase in labor efficiency, a 70-percent increase in outputs, and a 50-percent reduction in training time, according to a published case study.
These and other emerging data analytics tools are overcoming the limitations and barriers of their predecessors.
Making analytics more accessible
One significant barrier in the past was that the tools offering potential benefit remained unused, said Izik Avidan, business unit manager, digital manufacturing analytics, Siemens Digital Industry Software.
More than 80 percent of advanced analytics projects fail, he said, a statement backed up by research from Gartner and others.
“The main issue with data analytics tools of the past—from the manufacturers’ perspective—was the fact that they remained tools,” Avidan said. “A lot of the platform and solution providers did not realize that the average manufacturer did not possess all the necessary skills to completely utilize these tools. The tool delivered the function it was designed to do but the overall project probably failed. You have to be able to bridge between the manufacturing language and all these new technologies. The manufacturing customer does not possess these skill sets.”
“Historically, you really needed a PhD in math or data science to get value from analytics,” van Heerden said. “You need to put analytics in the hands of their existing operational people. You can’t go to the manufacturing customer and say, ‘You need to retrain your people or employ new people to benefit from analytics.’ The key is to make analytics accessible to process engineers and line operators.”
“Tools were designed for the experts, to make a difficult challenge easier for the experts as opposed to simplifying their jobs,” Ed Cuoco, VP of strategy and solutions at PTC, said. Many tools also required an onsite data scientist. The end result: “These tools were not suited for big manufacturers, which typically do not have their own data scientists,” Cuoco said.
Nine additional barriers from the past
Other barriers, according to Avidan, Cuoco, and van Heerden have been:
Failure to understand and address the manufacturers’ pain points.
Tools that required manufacturers to replace expensive legacy equipment.
Lack of access to the necessary data needed to derive insight, often because that data was in siloed systems, sometimes called dark data. Between 60 percent (Forrester) and 97 percent (Gartner) of data collected remains unused.
Data that could not be easily combined with
Data that was difficult to clean, format and prepare.
Tools that assumed data met a high quality benchmark, requiring an expert to improve data quality in many cases.
Lack of analytics tools enabling managers
Analytics tools that were too difficult for the average operator to use.
Tools that could not be scaled beyond an initial pilot or demonstration.
New day dawning
Today’s tools offer fast-time-to-value, easier operation and scalability. Increasingly, manufacturing software providers understand that their customers need platforms that combine several tools and integrate well on the shop floor, Avidan said.
“We are now seeing more success in analytics tools overcoming those barriers,” van Heerden said. “The tools we are providing are showing evidence of rapid value.”
Software manufacturers are designing tools and platforms that will run in factories that use machines that are 40 years old, as well as two years old, Cuoco said. “These tools need to function in a real-world environment,” he said. “That’s the key to applicability in a factory. That allows a factory to leverage things within their power without asking them to be good at things that aren’t in their wheelhouse.”
Emerging tools offer the ability to access, store, and handle the data, the availability of a subject matter expert on site or available remotely, with a low cost of ownership that doesn’t require too many additional servers or cloud resources, that are easily configurable, customizable and able to provide some value immediately, Avidan said.
“Nowadays, most of the software companies understand that throwing some machine learning solutions on the shop floor won’t solve your quality problems,” he said. “Now, they’re providing full turnkey solutions, which is probably the game changer.”
“My profession lives inside that tension in trying to provide an out-of-the-box solution and also understanding that the solution needs to be customized, that flexibility to tailor the solution to optimize the manufacturer’s need,” Avidan added. “In the past five years, we’ve seen more and more hybrid projects, that are both platforms and tools, combined with software specifically tailored for the type of industry.”
The industry has not yet arrived at the point where tools work out of the box, similar to an iPhone, Cuoco said.
“Out of the box is the direction,” he said. “Out of the box is the goal. We’re starting to get mature enough in our solutions to see a point where that will come to be.”
These tools, sometimes combined with an experienced manufacturing engineer, can help manufacturers improve performance and predictive maintenance and integrate quality control into production, Avidan said.
Tools also are “moving beyond alerts” that leave it to a human to act to become more closed loop—where the tool itself can take real-time, safe control actions, allowing a plant to gain or sustain optimized productivity, van Heerden said.
Instead of offering analytics as merely a component of an offering, more manufacturing software providers are offering analytics in solutions that address specific use cases, Cuoco said.
“If you introduced an experienced manufacturing engineer to data patterns in a data set or a list of conclusions he has seen before, he will be able to easily convert that data to actions that the operator, line manager, or owner of the factory can take and dramatically improve results,” Avidan said. “With that turnkey solution we can tackle the majority of challenges within days or even hours.”
Even better will be a time when more expert domain knowledge can be added to these tools, Cuoco said. “More and more specific domain knowledge needs to be embedded,” he said. “How can we put the expert and the machine together and have them both understand the problem? The machine has to be able to say, ‘I am able to factor in parameters specific to this domain.’”
Analytics also are improving as they are applied throughout the supply chain from raw material suppliers, to shippers, to manufacturers to end customers, van Heerden said.
Some challenges remain
Still needed are refinements that make tools easier to build, as well as tools designed so that machines can do more of the work, Cuoco said.
Costs definitely need to fall further so that the technology can become accessible to small and medium-size manufacturers that face similar problems, Avidan said.
More standards are needed so that manufacturers can more easily integrate technology from different vendors, he said.
“We have to understand that there is already a lot of software on the shop floor,” Avidan said. “Any solution you would like to introduce to that ecosystem would have to integrate well and seamlessly into these IT solutions … so that if you have an action item that needs to push from one engineering system to another, it can be done within a single portfolio. That’s one of the most important things you can do.”
As standards and interfaces become more open, integration will become easier, van Heerden said.
To succeed, partner with a trusted industrial vendor providing a comprehensive product, van Heerden said.
“A lot of analytics vendors offer solutions that can solve part of the problem. Some can analyze data,” he said. “Some can make predictions. Some can run simulations. Another one can optimize a setting. Partner with trusted vendors who will not disappear tomorrow and who can offer all these capabilities in a single product.”
Manufacturers looking for immediate perfection and complete integration should scale back expectations in favor of a step-by-step approach. “Don’t take a ‘big bang’ approach to adopt technology until all systems are perfect,” he said. “I recommend a rapid, incremental approach. Equip operations people with easy-to-use tools so they can get rapid incremental value.”