In recent years, leading industrial companies have made solid progress in improving productivity across the entire manufacturing value chain. But the traditional levers that have driven these advances, such as lean operations, Six Sigma, and total quality management, are beginning to lose steam and the incremental benefits they offer are diminishing.

As a result, leading companies are now seeking disruptive technologies to improve performance . Many are beginning to experiment with software development solutions such as machine-to-machine digital connectivity (Industrial Internet of Things or IIoT ), artificial intelligence (AI), machine learning, advanced automation, robotics and additive manufacturing (AM ) . All of them are part of what is commonly known as the fourth industrial revolution or industry 4.0.

In this article we will talk about what a company should do to achieve flexible automation using artificial intelligence.

Flexible automation

Before we get into it, you should keep in mind that flexible automation can magnify existing inefficiencies if it is introduced into a poorly designed factory layout and production line configuration . This argument fits the second rule of custom enterprise software development  stated by Bill Gates:

It seems obvious, but it's important to remember that automation doesn't eliminate a task, it eliminates the human element needed to complete it . Thus, if the process is characterized by unnecessary movements, defects and rework or overprocessing, automation will only magnify that waste. Therefore, it is of utmost importance that the technology we are going to implement solves a problem that is well defined before being implemented. Furthermore, widespread adoption of advanced robotics, cobots and AM (additive manufacturing) will not eliminate process variation and waste in the production flow, but will, at best, reduce it.

Although, such a perspective describes automation as potentially harmful if applied to inefficient processes; Automation can still increase efficiency by decreasing or eliminating waste. However, the risk of magnifying different types of waste as a result of inefficient automation and “wasteful processes” is imminent. For example, implementing expensive and complex advanced robotic systems in an inefficient assembly for the purpose of decreasing the number of manual steps, which could otherwise have been reduced or eliminated, will only cause those "useless" steps to be completed faster .

Therefore, it is important to remember that  AI should not be used as a means to optimize processes that have been being automated for decades. The real potential is to do something completely new with software development services. Tasks that were previously performed by humans or machines can now be carried out by software controlled by AIs that power robots. This increases the robots' flexibility and traceability and, in many cases, reliability.

However, there are two obstacles to take into account:

• A limited number of specialists
• Lack of transparency of the technology itself

Limited number of specialists

We have already commented on this point on numerous occasions, especially in those articles in which we made reference to the study published by the European Commission where it gave specific figures on this.

The lack of professionals is one of the 3 main barriers to AI adoption by companies.

Generate trust in new technologies

The second obstacle is the technology itself, which initially seems inscrutable to many. In the context of AI, many people are asking the question, can we trust AI as a technology? So, compared to other custom development services, it is crucial to understand that AI (and specifically, for example, machine learning), does not always behave exactly the way it is intended, it can make mistakes or it can behave in ways " unethical . Training non-technical personnel is the key to solving this obstacle.

As on numerous occasions, knowledge helps to overcome obstacles. It is not about everyone in your company becoming data scientists, but yes, from the CEO to the one who tightens a screw, it is important that they have a minimum knowledge of what AI is and how it affects them, so so that any reluctance to implement this technology can be eliminated.

Some examples of application of AI in manufacturing

Below we will see a few selected examples to show possible practical applications of artificial intelligence and machine learning in industry 4.0. But the possible applications are almost infinite, almost as many as there are industrial companies.

Use of autonomous mobile robots to improve processes in the chain

Manufacturing flexibility improves a company's ability to react in a timely manner to customer demands and to increase the productivity of the production system without incurring excessive costs or expending excessive amounts of resources.

Autonomous mobile robots (AMR ) offer a suitable alternative to decentralize the flow of materials due to their strong integrated computational power. Decentralizing material flow provides more flexibility to production systems and improves process productivity.

AI-based solutions will be used for the positioning and navigation of intelligent vehicles, such as AMRs. It is very important to note that the greatest flexibility can be achieved with the help of AMR without the need for a complete redesign of production lines. Thanks to the use of AI, a route (a model) is designed between the workstations and the interoperable buffers that allows congestion to be avoided by using multiple crossings and analyzing both the flow and the loading/unloading phases.

The cost of AMRs and the number of shifts are the key factors to improve flexibility and productivity.

Industrial organization engineers can use the data and results obtained and, thanks to the application of machine learning models, determine the optimal configurations and improve their decision-making process, so that the production lines gradually evolve towards autonomous production networks.

Flexible production networks based on AMR are more advantageous compared to traditional production lines.

With all this information, analyzes can also be carried out to understand the impact on the associated performance of AMR-based production networks compared to traditional balanced production lines.

Automatic part sorting under adverse and unpredictable circumstances

An automotive industry supplier has a simple automation solution for sorting metal parts. The problem is that correct classification is very difficult because the lighting conditions in the facility are unpredictable, the parts often receive direct sunlight and, in addition, the parts are metallic and highly reflective. The possible appearance of instant rust on them also had to be taken into account, which makes their classification even more difficult.

A machine learning system that is based on artificial vision and data is capable of managing various variables: position, lighting conditions, color, obstructions due to remains of materials... and autonomously learning to classify the pieces, regardless of the time of day, sunlight intensity, surface condition and packaging matches.

Not only that, but the model continues to learn over time and feedback, so if the conditions or particularities of the parts change, it automatically adapts and is still able to classify them correctly regardless.

To finish, in the last three examples we will see what three large, world-known companies are doing .

GE (General Electric)

General Electric one of the largest and most diverse manufacturers on the planet, making everything from large industrial equipment to household appliances. It has more than 500 factories around the world and is transforming them into smart facilities.

In 2015, GE launched its Brilliant Manufacturing Suite for customers , which it had been testing in its own factories. The system takes a holistic approach to monitoring and processing the entire manufacturing process to find potential problems before they arise and detect inefficiencies. The goal of GE's Brilliant Manufacturing Suite is to link design, engineering, manufacturing, supply chain, distribution and services into one intelligent, globally scalable system. It is powered by Predix , its industrial Internet of Things platform. In the manufacturing space, Predix can use sensors to automatically capture every step of the process and monitor every piece of complex equipment.

SIEMENS

German conglomerate Siemens has been using neural networks to monitor its steel plants and improve efficiency for decades. The company claims that this hands-on experience has given it an edge in developing AI for manufacturing and industrial applications .

Siemens aims to monitor, record and analyze everything in manufacturing , from design to delivery, to find problems and solutions that people don't even know exist.

A notable example of success in the case of Siemens is how it has improved emissions from gas turbines. The head of Digitalization and Automation Research at Siemens Corporate Technology commented that "even after experts did everything possible to optimize the turbine's nitrous oxide emissions, our AI system was able to reduce emissions by 10 at 15% additional"

Siemens' latest gas turbines have more than 500 sensors that continuously measure temperature, pressure, stress and other variables. All this information feeds its AI based on neural networks. Siemens says its system is learning how to continually adjust fuel valves to create optimal conditions for combustion based on specific weather conditions and the current state of the equipment.

FANUC

While GE and Siemens are very focused on applying AI across the entire manufacturing process, other companies that specialize in industrial robotics are focusing on making robots smarter.

This is the case of FANUC, which is using deep reinforcement learning to help some of its industrial robots train. Robots perform the same task over and over, learning with each iteration until they reach sufficient precision. FANUC has partnered with NVIDIA with the goal of allowing multiple robots to learn together. The idea is that what one robot can take eight hours to learn, eight robots will be able to learn in one hour. Rapid learning means less downtime and the ability to handle more varied products in the same factory.