AI systems can be helpful assistants for factory workers. Plamen Kiradjiev and Jochen Benke from IBM explain in an interview how this can reduce machine downtime, why augmented reality still needs time and how to get people to accept technology.
IBM has developed an intelligent manufacturing assistant for the automotive industry. What is it all about?
Plamen Kiradjiev: At the beginning of IAMA (Intelligent Automotive Manufacturing Assistant), the idea was to develop a solution for the shop floor with a large automotive company, using technologies such as artificial intelligence, augmented reality and the Internet of Things. We provide a marketplace of various services in this context. One focus is to transmit relevant information to where it is needed.
What does this mean specifically?
Kiradjiev: What does a worker do when a machine stops? He goes back to the office and uses computer to browse for the relevant document within a certain structure to correct the error. But this structure is 10 to 20 levels deep. Whoever created the levels can handle the structure. But a normal service technician has big problems finding the right document quickly. With IAMA, we transmit the information directly to a machine or mobile device. And we make it easier to find with an intelligent search.
How does artificial intelligence come into play?
Kiradjiev: AI makes a semantic and contextual search possible. The user is also guided directly to the appropriate text spot and does not have to search through a document with 1000 pages to obtain his information. For this purpose, the solution is trained based on an ontological arrangement of terms in special filters that are important for the respective employee. As a result, a distinction is already made in advance in topics such as factory, work area, building, station, machine or component. Intelligent optical character recognition, i.e., OCR, is also performed in preparing the information. This is used to digitalize paper documents, such as signed security protocols.
What is the concrete benefit?
Kiradjiev: We were able to calculate that the solution can save up to 30 minutes in searching for relevant information. This dramatically reduces the downtime of a machine and saves millions in downtime costs. Companies need to rework less. For example, additional costs due to extra shifts are eliminated.
How do workers access such manufacturing assistants?
Kiradjiev: In principle, the employee can use the wizard on all devices: via a desktop computer, tablet, smartphone or on a screen at the respective workstation. We have also developed a service where we can get a remote expert to help, based on augmented reality. In the event of a problem with his mobile device, a service technician can make image sequences of his surroundings and send them to the expert. The latter can then mark certain areas in the images, for example, and consequently assist the worker in resolving the problem.
Data glasses are also usually used in augmented reality. What role do you think this technology will play in manufacturing in the future?
Jochen Benke: One project developed a vision system in the logistics field. The worker was shown the number of the shelf into which he was to sort a part as well as a directional arrow via augmented reality glasses. As a result, the error rate has virtually fallen to zero.
Kiradjiev: However, the glasses currently available are not yet suitable for use in work. We're not talking about a computer game here. Augmented reality glasses basically provide a communication channel between man and machine. But you have to use them in such a way that they fit ergonomically to people.
Is a worker's acceptance one of the biggest obstacles when introducing technologies to support him in his work?
Benke: In all the things you introduce, you have to make sure that they are not only accepted by IT or by the manufacturing engineers, but also by the workers who have to work with them. And you have to make sure that a solution is easy to use. The operations team should not have to deal with any additional complexity.
Kiradjiev: For example, an AI solution must be designed so simply that the user does not have to be an AI specialist.
Benke: We try to implement technical assistance systems in our projects in such a way that they manage themselves. You hardly have to intervene. Everything is rolled out and monitored automatically.
What else do you have to pay attention to in addition to usability?
Benke: Already in our first reference project with an agricultural machine manufacturer, the manager responsible handled this in a very clever manner, because the employees on the line were very skeptical at first. But when the management team from the USA came, he had the workers present the solution as their system. This increased acceptance.
Kiradjiev: Employees were involved from the start. And we also got feedback from the workers and talked about the desired adjustments. It's their system. Why should I tell employees what to use? The workers know what they need. People are the focus. Those are not must empty words.
Where do you currently see the greatest potential for AI in assisting workers?
Kiradjiev: I see many possibilities, especially in inspection activities and also in quality assurance. It can relieve the burden on humans there, for example, in controlling circuits in electronics. An example is a service with which quality assurance can be carried out spontaneously. A worker can take a picture with his smartphone – for example for checking rims – and intelligent algorithms then detect any errors. Once these have been found, the employee also receives the appropriate instructions via the service. AI basically has great potential to support employees in a factory. However, it must not be used purely as a goal in itself.