Machine Learning Techniques for Smart Manufacturing: Applications and Challenges in Industry 4.0
Industry 4.0 is now underway, changing the traditional manufacturing processes into smart manufacturing and creating new opportunities where machines can understand the involved processes, interact with the environment, and adapt to their behavior.
Artificial Intelligence and Big Data are making machines in the manufacturing industry smarter than before by addressing how to build computers that enhance automatically with experience. As a subfield of AI, Machine Learning is the primary driver of such innovations in the manufacturing sector.
The advancement in technology through machine learning has brought the opportunity to accelerate discovery processes and improving decision making.
Also Read: The Future of the manufacturing industry: Technology trends for 2019 & Beyond
In recent years, ML has become more prevalent in the building and assembling sectors by using advanced technology to reduce the cost and time involved in the production. In Smart assembly manufacturing robots can put items together with surgical precision as technology adjusts errors in real-time to reduce wastage.
Employing ML in businesses allows the monitoring of quality as well as optimizing operations. According to TrendForce, Smart manufacturing is expected to grow rapidly in the next few years. The company also predicts that smart manufacturing will be worth more than $200 billion by the end of 2019 and to grow by $320 billion by 2020.
The following are Machine Learning Techniques for Smart Manufacturing.
Quality control and OEE:
Machine learning plays a critical role in enhancing Overall Equipment Effectiveness (OEE). The metric measures performance, availability, and the quality of assembly equipment, which are all enhanced with the integration of deep learning neural networks. It quickly learns the weaknesses of such machines and helps to minimize the weaknesses.
Optimized semiconductor manufacturing:
According to McKinsey & Company, there is great value in using ML to improve semiconductor manufacturing yields up to 30%. The company believes this is achievable by reducing scrap rates and optimizing ML operations.
The technology uses root-cause analysis to reduce testing costs through streamlining manufacturing workflows. Also, manufacturing equipment that runs on ML technology is expected to be 10% cheaper in annual maintenance expenses with a reduced 20% downtime and a reduced inspection cost of 25%.
Perfecting the supply chain
ML plays a vital role in improving an organization’s value by maximizing its logistical solutions such as asset management, inventory management system, and supply chain management. The combination of IoT and Artificial Intelligence (AI) is crucial for a modern company to realize the optimal operation of its supply chain.
A study conducted by A.T. Kearny and the World Economic Forum established that manufacturers are looking on how to combine emerging technologies such as IoT, ML, and AI to improve asset tracking, supply chain visibility and optimizing inventory.
PWC predicts that more manufacturers will use machine learning and its analytics to enhance predictive maintenance slated to grow by 38% in the next five years. Process automation and visualization are expected to grow by 34% over five years. The integration of APIs, analytics, and big data will grow the connected factories by 31%.
Use Cases of Machine Learning Techniques applications in Smart Manufacturing:
Machine learning techniques have made tremendous improvements in the manufacturing industry. The application of ML is constantly increasing over the last decade. This growing implementation of ML has led to the availability of big data with interesting patterns, database technologies, and the usability of ML techniques.
Renowned companies such as Siemens, GE, Funac, NVIDIA, KUKA, Bosch, and Microsoft are implementing ML-powered approaches to improve their manufacturing processes. They use the technology to reduce the cost of production, reduce the number of defect products, shorten unplanned downtimes, increase the speed of production, and improve transition times.
The international federation of robotics estimated that the number of industrial robots in operation in factories would grow to 2.6 million in 2019 from a low 1.6 million in 2015.
Most of the firms using ML for their manufacturing processes are using the same tools in their manufacturing before releasing the technology to the rest of the market. This approach makes them the developers, the test, and the initial consumers of many of these advances.
General Electronics (GE)
GE is the 31st largest company in the world by revenue. It is also among the largest and most diverse manufactures making everything ranging from home appliances to industrial equipment. The company has started to transform its branches into smart facilities.
In 2015, GE launched its brilliant manufacturing suit for its customers, a product it had been testing in its factories. The brilliant manufacturing system assumes a holistic approach to tracking and processing the entire manufacturing procedure to identify possible problems and inefficiencies before they spread.
The first brilliant factory was made in 2015 in Pune India by investing $200 million. After the installation of the system, equipment effectiveness was increased by 18%. The primary goal of GE’s smart manufacturing system is to connect design, engineering, supply chain, manufacturing, and service distribution into a globally scalable and intelligent system.
GE’s brilliant system is powered by Predix, which is its industrial IoT platform. Predix uses sensors to automatically capture every manufacturing step and track each piece of complex industrial equipment. The IoT platform uses the acquired data to identify potential problems and devise possible solutions. General Electronics spent about $1 billion in developing the system and expects it to process 1 terabyte of data in a day by 2020.
The Predix system is now running in seven GE factories serving as test cases. The system record gradual improvement with GE stating a 5% increase in productivity for their Vietnam wind generator factory that is powered by Predix.
Siemens
Siemens, a German conglomerate, has been using neural networks for decades in its firm to enhance efficiencies. The firm claims that this practical experience has aided it in developing AI for manufacturing and industrial applications.
In 2016, the company launched Mindsphere, which is the main competitor to GE’s Predix. Mindsphere, as described by Siemens, is a smart cloud that can be used by industrial manufacturers to track machine fleets for service purposes throughout the world. In 2016, Siemens integrated IBM’s Watson analytics in the tools provided by their service.
The primary aim of Siemens is to monitor, record, and analyze the entire manufacturing process from design to the finished product. This close tracking helps in identifying problems and solutions that people may not know of their existence.
The successful implementation of Siemen’s ML technology has facilitated the prevention of specific gas turbines emissions more than any human could do. Dr. Nobert Gaus from Research in Digitization and automation in Siemens says even after experts had done their best to enhance the turbines emission of nitrous oxide, the AI system was able to reduce emissions
by 15%. gas turbines emissions more than any human could do.
Dr. Nobert Gaus from Research in Digitization and automation in Siemens says even after experts had done their best to enhance the turbines emission of nitrous oxide, the AI system was able to reduce emissions
by 15%. Siemens latest gas turbines have more than 500 sensors that constantly monitor temperature, stress, pressure, and other vital variables. The gathered information is fed to the neural network-based AI.
According to Siemens, the network continues to learn on how to adjust fuel valves to come up with the best conditions for combustion based on the current state of equipment and specific weather conditions.
The company envisions the technology to be used with a product called Click2Make, which is a product-as-a-service technology. If a company has a complete understanding of the resources available and a highly adaptable robot, the end goal is to make manufacturers have optimal mass customization. For the technology to work, if a company decided they would like to produce a specific object, it would submit its design and the system would automatically initiate a bidding process between facilities with equipment and time to process the order. This system would enable manufacturers to automatically derive production plans and
provide them in real time to potential buyers.
KUKA
This is a Chinese owned German company and a leading manufacturer of industrial robots. KUKA is heavily investing in robot-human collaboration through machine learning. Previously, industrial robots were strong and non-intelligent meaning it was dangerous to work alongside humans. KUKA has developed an LBR iiwa robot that uses intelligent control technology
to collaborate with human workers safely.
The company used LBR robots in their manufacturing plant. If robots can work safely with humans, it means they will be deployed in areas and functions they haven’t been deployed before, like positioning manufacturing components with human workers. The robots can also be reassigned new tasks as the need arises.
Fanuc
Fanuc is a Japanese company specializing in industrial robots. Recently, the company made a strong push for greater connectivity and the use of AI in their equipment. In 2016 Fanuc announced its collaboration with Rockwell Automation and Cisco to develop and launch FIELD (Fanuc Intelligent Edge Link and Drive), an industrial IoT manufacturing platform.
After performing the same task repeatedly, Fanuc robots learn to achieve a high rate of accuracy. The company is also partnering with NVIDIA with a goal of allowing multiple robots to learn together. What would normally take one robot to learn in four hours would now take four robots to learn in one hour. Faster learning ensures less downtime and handling varied items simultaneously in a factory. In the future, robots could transfer their skills and learn together.
Challenges
Smart manufacturing enabled by machine learning is still a young scientific sector which is growing rapidly. Despite the enormous benefits it has brought in the manufacturing sector, it is still faced with various challenges.
Security
The interconnection of manufacturing components poses a great risk to the security of the entire processing plant. If part of the network is compromised, through an attack by malicious people, the production process could be tampered with. In cases where robots are working alongside human beings, it could result in exposing them to danger if the robots are compromised.
Complexity of system
Today, the manufacturing industry is facing an increment of challenges related to complexity and dynamic behaviors while adding that the manufacturing is affected by uncertainty. The constant enlargement of big data coupled with its availability poses a great challenge to the manufacturing environment since the knowledge cannot be extracted.
Misuse of data
There is a big concern related to the collecting of big data in its privacy, economic value, and security since many organizations store the data in virtual cloud platforms. The misuse of data in the manufacturing sector is on the increase because many devices involved in the process of collecting and examining data are controlled remotely. Therefore, there is a need to do additional research to improve the storage and security of the data.
Integration
The manufacturing industry is majorly characterized by a culture of repairing or replacing the equipment once they are broken. Many machines are used beyond a point where getting their parts becomes difficult. Manufacturers continue to use them because replacing them would be costly, and expense small industrial manufacturers are unwilling to meet when the existing machinery is working perfectly.
If the old machines continue to be in use, it becomes hard to optimize IoT on all manufacturing equipment. Therefore, companies continue to operate in the old era characterized by improper decision making, high costs of production, prolonged downtimes, and low accuracy.
Insufficient Skills in companies
Bring the technology of smart manufacturing in a firm is of much importance as possessing the skills to run the technology. If a company is planning to implement smart manufacturing, it must also have the expertise needed to maintain the equipment involved in the process. Since this is a new technology, many manufacturers are faced with the challenge of recruiting new staff with the right knowledge or training the existing staff on the smart manufacturing environment.
Endnote:
For many years, robots, automation, and complex analytics have been used in the manufacturing industry.
However, institutions have also been looking at ways to reduce waste and improve efficiency. The evolution of this industry has led to smart manufacturing. The use of intelligent robots, advanced analytics, and sensors is expected to bring tremendous improvements
in the manufacturing sector.
The new advancements in industrial manufacturing will reduce equipment failure, improve production
time, and improved accuracy. When then changes are added together and spread over a large sector, a company can significantly save on cost and increase returns.