[InnoVEX FORUM 2022] The Future of Manufacturing Requires New Paradigm

 2022-07-29 By: InnoVEX Team

AI implementation in industry is one of the most reliable ways to improve production efficiency and capability. By using AI and machinery to take over jobs that are unsuitable or too dangerous for people, companies will be also able to improve plant performance and safety. However, AI implementation is not a simple matter and can be quite costly in addition to the high tech machinery that needs to be installed within the facility. In addition, there is always the risk of a failed improvement if the AI model is not supported with the correct data.

The 2022 USA-Taiwan International Conference on AI+ Smart Manufacturing forum focused on advocating technology excellence, encouraging potential commercial opportunities, and realizing bilateral co-operation in any possible form, between USA and Taiwan AI industries. The forum was opened with opening remarks by Audrey Tang, Minister without Portfolio of Executive Yuan; Dr. Ta-Ching Shih, Economic Specialist of AIT; and Pei-Wei Chiang (Leon), Chair of Taiwan Artificial Intelligence Association. The speakers in this event were Shu Huang, Division Director of Industrial Technology Research Institute; Johnny Chen, Chairman of Solomon Technology Corporation; and David Huang, Manager of Techman Robot Inc. representing Taiwan. Representing the USA were Kai Yang, VP of Product of Landing AI and Jack Wang, TCE Application Development Manager of Rockwell Automation.

Robotics as an Enabler of Digitization

Dr. Shu Huang, Division Director of Industrial Technology Research Institute (ITRI) shared in his keynote that as one of the newer fields, AI robots aim to give robots the capabilities to perform human skills with manufacturing domain knowledge. To achieve this, factories can digitize their human workers’ expertise through sensors that will capture and generate data that can be analyzed through machine learning and AI models. For verification, users can also test the model in a digital twin or cyber space to make sure the desired movements are executed perfectly. This method is beneficial in 2 different ways: enabling full skill transfer & result replication while reducing variations due to human intervention.

As AI robots rely on their controllers to function, it is important to use one that suits the users’ needs; especially as different factories might practice different manufacturing methodologies as well as require different functions. Smart manufacturing will also benefit from digital twins that can visualize and simulate results of the new changes before they are implemented in the facility. This will also prevent unnecessary or even harmful changes from taking place and when the upgrade is successfully implemented, it will also allow workers to move away from the more dangerous works and enable further optimization when necessary.

Ubiquitous AI Increases Quality, Speed, and Consistency

Mr. Johnny Chen, Chairman of Solomon Technology Corporation stated that there are plenty of things we can do with AI in factory floors and warehouses. Robots have been in use as helpers in factories and warehouses for a long time, but often they are programmed only to follow certain routes or simple actions as is the goal of process automation. An AI powered vision system will be able to perceive complex patterns as well as enable cooperation between robots working on the same cell while preventing collisions or accidents. This means factories will be able to increase processing speed by using multiple robots for material handling in the same location even if the products were randomly placed.

AI in manufacturing and logistics are not limited to the production process as AI can be used for high-speed and high accuracy inspection or quality control. In this application, the AI will need to discern which products are within acceptable parameters as well as log the data properly. In addition, AI can also be implemented in wearable devices such as AR glasses to help employees in their work such as by displaying information or giving feedback on quality of work for SOP validation and quality control.

Proper Training is Needed to Develop a Good AI

Mr. Kai Yang, VP of Product of Landing AI shared that the main application of AI in manufacturing focuses on visual inspection by combining a camera and AI model to free up the humans for more valuable work. While this sounds simple, in reality only less than 25% of AI projects are live. The 2 main causes for this low adoption rate are that the implemented AI systems are not accurate enough and that AI needs much more than just codes, but also significant software & hardware infrastructure.

AI models that do not meet the users’ accuracy requirement are often not trained enough; meaning that they are not supplied with sufficient data or high quality data. This might be caused by user error during training such as by mislabeling data or inconsistent labeling. When combined with low data quantity, the AI model will not have adequate level of accuracy needed for industrial applications. There is more to AI than just code as various supporting hardware and software can enhance its operation to a respectable degree. Users who wish to implement AI should start quickly and not be afraid to start small; however they need to approach AI from a data-centric perspective to properly train the model. It is also important for users to use an end-to-end platform that supports the AI project lifecycle.

AI + Robots = Brain and Hand

David Huang, Manager of Techman Robot Inc. focused his speech on cobots (cooperative robots) with AI vision. He shared that robotics in manufacturing are mostly used for jobs that are ideally not done by humans; the so-called 4D tasks: Dangerous, Dull, Dirty, and Dumb; essentially tasks that while necessary are repetitive, unsafe, harmful, or simple. However, a majority of industrial robots today are separated from humans for safety reasons despite them taking over the dangerous tasks and becoming the source of danger themselves. Cobots overcome this problem by making human-robot collaboration in close proximity safer.

With AI vision, robots can have a degree of versatility as they can be trained to follow and duplicate patterns or instructions without needing to repeat the training process for each individual robot. AI vision is most often used for quality control and defect checking, but depending on the production system, combining AI vision with robotics might be able to reduce processing, testing, and calibrating time exponentially. For example, robots might not be able to do too much with low volume high variety production systems, but when AI vision is introduced, the robots will be able to do quality inspection in various settings and for multiple products while logging the data as needed.

Digital Transformation Needs Proper Strategies & People

Mr. Jack Wang, TCE Application Development Manager of Rockwell Automation stated that there are a few key drivers that push local manufacturers to implement digital transformation. One of the main factors is the aging workforce that often means a lack of possible skill or knowledge transfer. IT-OT convergence often implements new systems without concern on the various systems’ compatibility. Factories today also employ aging equipment that might not need to be phased out yet, but need to be maintained and upgraded to make sure it still generates value for the users.

Factories aiming to implement AI in their process must keep a clear strategy and keep in mind that while data is important for AIs, without connectivity; ie. access to data, data integration, etc; the AI will not be useful. It is also important to note that a majority of people in factories might not be data scientists and without the proper expertise, it is impossible to transform data into useful information. AI will require 3 types of people: people with domain knowledge such as process engineers who understand what happens in the platform; automation engineers who know how to automate those processes; and an AI expert who can operate the data.

Better Efficiency as the Core of Future Manufacturing

The panel discussion was joined by Shu Huang, Division Director of Industrial Technology Research Institute; Johnny Chen, Chairman of Solomon Technology Corporation; David Huang, Manager of Techman Robot Inc.; James Chen, Regional Sales Manager of Landing AI ; and Jack Wang, TCE Application Development Manager of Rockwell Automation. The panel was moderated by Connyn Chang, Director of Taiwan Artificial Intelligence Association and discussed the roles of AI in smart manufacturing.

Compared to traditional manufacturing, smart manufacturing is capable of producing multiple product types and offers more flexibility as robots will be able to perform different functions with simply changing its modules and AI models. As manufacturing today moves from mass manufacturing to small batches, flexibility becomes a benefit and is perhaps one of the main draws to smart manufacturing and why AI is being sought after as a professional/ industrial solution to existing problems.

However, while AI can solve some issues, it cannot solve all existing problems and other solutions might be needed instead. In essence, users must find the right scope for the right solution to solve the right problems. Starting a small scale AI project will be the most effective way to find the right scope and solutions.

One of the foundations in smart manufacturing is data standardization which will be needed to make sure data from the various machines, sensors, IoT devices, etc. can be utilized for the AI project. As data preparation often takes a great majority of the AI project, providing the right data will be able to help the organization reduce the cost of ownership and make smart manufacturing more workable for the organization. In this regard, data documentation is vital to the project and must be conducted at a consistent standard.

Currently as industries start to execute their Industry 4.0 programs, there are now ideas on what the next evolution - Industry 5.0 will entail and where AI will be the core of it. Regardless of the significance or the role that AI will take, efficiency and quality need to be considered as it is one of the most important aspects of industry. Efficiency can be achieved through AI and better tuning the parameters and configurations in the manufacturing lines and better quality will also help manufacturers reduce material waste, operating time, and energy usage.

Higher efficiency will also help factories achieve their sustainability goals as better fine tuned systems will be able to reduce or even eliminate waste. To retain customers and build trust, companies need to manage their escape rate and overkill rate in the production lines to make sure customers receive good products without unnecessarily scrapping good products which can be achieved via AI visual inspection systems. AI implementation will be able to expand the human possibility as plenty of tools are now available to be used in place of humans, especially for 4D tasks. With proper systematization and standardization, it is possible to provide people with better and more fulfilling work than the ones involving 4D tasks.

Efficiency in the process can be achieved by minimizing machine downtime such as through predictive maintenance and fine tuning parameters that will rely on data and image vision. With optimized processes and machinery, production will not need as much resource input as waste can be minimized. As one of the world’s manufacturing powerhouse, Asian companies often focus on cost reduction instead of improving efficiency which might offer better results for the companies; such as lowering lead time, system integration, lowering cycle time, and employing automation integration.

While manufacturing used to focus on hardware and machinery, software’s importance in manufacturing and robotics is growing. Every industry no matter the level of tech involved, size, or scale will have their own domain know-how that can be leveraged to improve the industry as a whole or even exported to other locations either as reference or starting point. In terms of smart manufacturing, companies now have the option of hiring or contracting many different companies or consultants; each with their own specialization that has become a way for them to differentiate and excel. However, expansion will need proper marketing efforts as trust, history, and credibility are vital for a successful expansion which makes partnerships important, especially with other companies such as System Integrators (SIs).

Watch the full forum video here.

For more InnoVEX forums, please click here.