The main operational focus in Reboot has been in a creation of a Phase I PoC Portfolio, resulting in 20 entries during the first eighteen months. All of them originated from company needs and utilised by our factory members Nokia, ABB, GE Healthcare, KONGSBERG and Ponsse. In addition, there have been several innovation spill-overs where factories learned and implemented good practices from each other or from research.

• Prospect to project enables better sales prediction to improve production planning, procurement and budgeting. It utilizes several kinds of business data such as CRM reports, sales history and market predictions.
• Supply chain transparency improves response times of supply chain in terms of production planning and material flow. Shorter overall production time gives benefit in the market.
• Extranet development PoC & B2B extranet data transfer PoC reduce manual non-value-adding work and increase information transparency in procurement processes at both factory and its supply chain.
• Digital twin technical development replaces physical sensors with virtual ones in end products, cutting the sensor maintenance costs in very demanding environments such as marine thruster systems.
• Tester predictive maintenance lowers the cost of electronics tester maintenance, improves production efficiency and lowers waste as testing is more reliable and anomalies can be spotted via data analytics. SCALE-UP
• Robotic process automation saves human daily working time by utilizing software robots in low value routine work related to IT-systems such as ERP and excel sheets.
• Value of service enabled by digital solution improves means to communicate the customer value of product digital twin to the potential customer. This gives manufacturer confidence when investing to digital service development.
• Factory acceptance test automates the data collection and reduces manual work in product final testing. It also creates a possibility for external classification agencies to give approvals remotely.
• Mobile robots in material handling reduce manual non-value-adding work and increase productivity in intralogistics. Components are transferred automatically/by-order from supermarket to assembly line with mobile robots.
• Automatic component quality control with machine vision increases capacity of assembly cells as inspection is quicker. It is also more reliable and less demanding in terms of human work.
• Standard robot interface increases the utilization rate of cobots, which are usually statically installed to a single workstation. New technology allows cobots to attach to and detach from individual workstations and to utilize task-specific grippers, significantly shortening the ROI related to cobot investments. SCALE-UP
• Automatic error handling reduces the downtime of robots that have entered in an error state through self-diagnostics based on machine learning from video data. Automatic error recognition also facilitates the development of self-recovery functionality.
• AI foreman PoCs produce automated production personnel allocation proposals based on multiple data sources to support faster day-to-day decision-making regarding production planning and resourcing. SCALE-UP
• VR to achieve more understandable instructions facilitates training of new employees and/or new products on manual assembly work and can also serve as competence development / maintenance tool for senior workers.
• Wearables in industry such as smart watches present new ways to inform employees about forthcoming tasks or alerts in production. Notifications can be tailored to each stakeholder group.
• Well-being at work increases understanding on what affects well-being of employees and how it could be improved. Studies show that productivity goes hand in hand with employee wellness.
• Gamification at factory floor improves work motivation and employee satisfaction. Work instructions or routine reporting can have game-like features such as rewarding trophies or visuals with 3D game engines.
• Legislation brings understanding on what conditions data privacy legislation (such as the GDPR & the Act on the Protection of Privacy in Working Life) sets for personal data gathering and processing within factory employees, as well as for automatic decision making through artificial intelligence.

Learning-based Machine Vision Tackled the Challenge

The challenge of ABB consisted of inspection of a one specific product on the automatic assembly line. They needed a system to make sure the product is assembled correctly. To make it easier, the factory already had a camera on the assembly line’s final inspection station. That camera collected the testing material needed to pilot the deep learning algorithm. Algorithms based on neuro-networks are one of the most recent trends on the field, but already familiar to Mika Saarinen:

”The development process was amazingly easy. This is a technical challenge we have been trying to solve with traditional rule-based algorithms for years, but it is not going to happen that way. The novelty of algorithms based on deep neuro-networks is that we can now solve problems that were previously very difficult.”

The solution’s Proof of Concept (PoC) is ready, waiting only for the paperwork to be completed until it can be utilised in production.

The project was called Cognex Vidi Deep Learning Machine Vision PoC. We tested learning machine vision instead of traditional machine vision in one automation cell. The co-operation with AP Vision went well and the results are promising. The learning machine vision was able to perform the inspection in this cell more reliably than previously used rule-based application. Programming the system was faster and easier than it was with the old method. We are considering expanding the utilisation of the new technology into new implementations after this project is completed. -Mikko Jukkanen, ABB

Seamless Co-operation

Reboot IoT Factory has been establishing ecosystem that creates easy co-operation among researchers, SMEs and factories. The companies have fresh networks, and chances to introduce their products and technology to the new audiences.

The project work has fitted seamlessly to AP Vision’s everyday business and there have been fruitful consequences. The company got another case, a chance to work with GE Healthcare Finland in a project that aimed to implement Cognex’s neuro-network AI camera into inspecting the coupling cables’ solders of an oxygen measurement sensor. This solution is ready and currently being used in production.

How the new technologies can be used to improve well-being and productivity of employees? Milla Immonen from VTT, Mikko Jukkanen from ABB and Anna Rohunen from the University of Oulu have been working on the Reboot IoT Factory Grand Challenge 3: Labor at Digital Work Environment. We had the chance to discuss with them how to develop a factory towards a digital working environment.

The big picture of developing the digital working environment consists of commitment of executives, rapid testing of new technologies and involving the workers on the factory floor. Gathering data about human resources requires empathy and legal control as well as appropriate tools.

How to Measure Stress?

The starting point of the work was tracking the well-being and performance of the employees at a factory – gathering data of the stress of factory floor work and limitations and preferences of the workers. The topic is sensitive, and it is important to explore thoroughly how to handle it with workers and their representatives.

“We have discussed the implementations of data collecting. Not big volume, just basics like how current algorithms can help measuring preferences. This begins with research”, explains Mikko Jukkanen from ABB factory.

“In the beginning we thought a lot of how the load of different work tasks could be measured. There are different kinds of workload: e.g. physical and cognitive. Together with the factory we have been building a concept of the worker information that could be used in work allocation to enhance the wellbeing and satisfaction of the workers”, Milla Immonen from VTT explains.

Mikko adds that other factors affecting the work have been mapped as well: the data of employee well-being, personal strengths and preferences completes the picture, and makes it possible to create a digital twin of an employee. The concept of employee digital twin is something VTT has created and it is in testing phase.

As Mikko says, “These measurements help taking care of well-being and tailoring the work for each worker. Creating a virtual pool of workers’ skills and combining the information to the order backlog helps making decisions and enhancing the production process.”

Listening the Factory Floor

The topic faces challenges in getting the factory workers receptive of the idea. The purpose is not to watch behind their backs but to support their performance and wellbeing. This requires a lot of background research especially in questions of privacy and confidentiality.

Anna Rohunen is a data expert who advices researchers in legal questions concentrating on privacy protection: “Collecting and using personal information is often considered sensitive in the context of employment. Privacy legislation regulates the data usage to guarantee the employees’ rights. We need to carefully make sure that the system is compliant with the legislative requirements”, she highlights.

It is very important to listen the workers and they have done exactly that by interviewing them. The pilot services have involved volunteers to test them and the workers got a chance to keep track of their own data and shape their own job.

”We have piloted gamification in production. The service tracks the production and records skill data to a matrix that collects the expertise of each worker. Each employee can then access and visualize their own personal data through GUI of the MES system. There have been promising results: we got good feedback from the volunteers and some of them continued to use the service,” Mikko talks about ABB’s pilot.

Another challenge is the attitude towards data collection. “When presenting the new solutions to the employees the focus needs to be on the positive effects and telling openly what data is collected. The solution is supposed to enhance well-being, safety, and satisfaction at work. The common fear is being constantly monitored and watched”, adds Milla.

Future vision: AI Foreman

Using AI in the production will increase in the future. AI Foreman is a solution developed by VTT that carries out work management with AI. Humanising AI is a future vision that these projects are aiming at. It is important to make digital environment and AI humane and user centred.

“The legal obligations are challenging and it is not yet completely clear how to interpret them in our context. References are still scarce in this field”, reminds Anna when speaking of AI-based data processing at work environment.

Mikko depicts a future vision: “AI foreman distributes tasks based on a worker’s digital twin. When a worker checks in the AI foreman greets and guides them. The worker’s preferences and skills have been mapped and the information is combined to the orders. The system monitors the day’s progress and how the goals are met. The AI basically takes care of resourcing and that everything goes well.”

Read more about Grand Challenge 3: Labor at digital work environment

Collecting data forms the groundwork for developed IoT solutions. We had the chance to discuss about data collection pilot with Jouni Vuorensivu and Janne Ouni (in the picture above) from Remion Oy. They have been working with ABB Porvoo Wiring Accessories on Remion created system that enhances the productivity of the factory with data utilisation.

Remion Oy is Tampere-based company specialised in industrial internet services, such as data collection, analytics and visualization using state-of-the-art technologies. The experienced company employs 23 people.

Piloting Big Data Collection

ABB Porvoo was looking for data collecting solutions and asked Remion to join the project. The solution was to collect, analyse and transfer production cell operation data into the cloud for further utilisation.

“Our PoC was a pilot concept, which aimed at building up experience of data collection. The main questions were what kind of information can be collected from the production cells and what can be developed of that data,” Janne explains.

Remion’s project with ABB Porvoo focused on big data collection. It was carried out with the service called Regatta®. It is Remion’s industrial IoT solution – a modular system offering different components based on the client’s needs. The ABB Porvoo project was to collect data for quality, predictive maintenance and production metrics and it was targeted to measure Cpk (quality) and OEE (production).

The service was tailored to meet the factory’s needs: the first module collected data, another processed it and the third module visualised the collected information.

Various Utilisations – from Maintenance to Visualising Trends

The pilot was built around injection molding machines that produce plastic electrification products for the building industry.

”We wanted to follow up the efficiency of the whole production cell – the OEE statistic. The traditional challenge with quality control is that the errors have been sorted manually, it is laborious and naturally less precise,” Jouni describes the project.

With the IoT solution, ABB Porvoo aims at higher OEE statistic and more precise follow up and control.

The data could be utilised in so many ways and different situations: in maintenance, quality control, catching errors and homogenising the products. Data helps minimising stops and interruptions; it helps monitoring the production. The users can react to alarms right away and fix the problems. The managers can follow the revenue and general efficiency of the plant or a machine on time. The data can also be integrated to other systems.

Putting the data on visual form makes it easy to analyse, browse values, and follow trends.

The Work Continues

The co-operation went smoothly according to Jouni: ”Working with ABB Porvoo was easy, everybody there were enthusiastic about the project and its goals. The challenges we faced were mostly concerning schedules. Although we had to solve the information security and privacy issues: we decided to keep the data collecting and transferring networks separate with manual transition for now.”

Remion has been discussing about expanding the system inside ABB Porvoo and will continue the work there. In the future they can utilise the factory’s own platforms to which the solutions can be integrated.

“We are in the early stages of updating and adding to the digitalization of our production equipment at the Porvoo factory, and the pilot project with Remion was a big leap in the right direction for us. The work that was done has been a great learning experience in how data from industrial equipment can be collected and processed, and this experience will aid us in configuration and scale-up of more comprehensive production data systems in the future.

Working with Remion was pleasant right from the start of the project. They are clearly experts within the industrial data area, but also very competent in project management and working flexibly within the constraints set by the client.” – Jesse Wickström, Production Development Specialist, ABB Oy WA

This project was part of the Grand Challenge 4. Read more: GC 4: Digital Production

Find out more about Remion Oy

One central goal of the Reboot IoT Factory project’s second phase is to transform routine factory tasks from people to robots. There is a noble societal objective underlying this: to keep factories and industry in Finland. Finland is a country of high labor costs. Automating tasks suitable for robots makes it possible to redirect humans’ activities into cognitively higher-level tasks. This, in turn, helps keeping the Finnish industry competitive. 

Robotics fusion combines several activities

The first phase of Reboot IoT succeeded in automating many activities. Now, during the second phase, the focus is on combining them. The used term is robotics fusion. Robotics fusion means not only connecting the activities of robots, but also advancing the co-operation between robots and people. Robotics fusion presupposes developing both software and hardware. The results are very concrete and demonstrable in real settings. 

Central areas for this work are for example machine vision, mobile robotics, and actions performed by a robot arm. Machine vision can be applied for example in monitoring how cogwheels wear out. This information can act as input for a digital twin. The twin, in turn, can perform estimations on future repair needs. Mobile robot Reetu transports components to appropriate locations in a factory. A robot arm can select proper parts and deliver them forward, in the right position.

People play central role in robotics fusion

Robotics fusion combines above-mentioned areas: a robotics arm can be attached to a mobile robot and machine vision can be applied in selecting the right parts. These days Reetu can even operate an elevator and dock itself with other carts. Furthermore, the fusion is also focused on the collaboration between people and robots. Traditionally, mainly due to security reasons, robots have been operating in separated fenced areas. Now Reboot IoT has started the first pilots where people and robots work together in shared spaces. 

Error anticipation and recovery are specific themes in robotics fusion. If a robot makes a lot of mistakes and always needs help from humans to recover, the whole point of automation goes down the drain. That is why it makes sense to invest in the robot’s fault tolerance and self-recovery capabilities. This is of particular importance in keeping the shared workspaces safe. 

Expert: Markku Hentula, VTT  

Text: Santtu Toivonen

Increasing the level of automation is a major trend in shipping, with advanced navigation systems and collision avoidance algorithms currently getting most of the headlines. To achieve business benefits and lifetime cost savings, however, it is equally important to ensure reliable operation of ship machinery and propulsion systems. Traditionally, maintenance of these ship systems has been very reliant on manning, reactive maintenance practices and scheduled overhauls. This approach has proved to be sufficient, as the crew has been constantly available to handle repairs and other maintenance tasks on-board. In the future, this may not be the case anymore, as the shift towards remote operated and autonomous vessels proceeds. This will potentially reduce manning and alter the crew tasks – whether they are located on-board the ship or in a remote control centre. 

Predictive maintenance is considered in the Reboot IoT Factory project as a key approach to mitigate the risk of failures as well as to improve the efficiency of industrial operations and the utilization of assets. Until recently, however, the research in the maritime sector on the implementation and use of predictive maintenance technologies, instead of continuous manning, has been quite rare. Thus, a pre-study on how to better understand the safety, legal, liability and insurance requirements, as well as the initial business considerations of using predictive maintenance in shipping was conducted in a collaboration between VTT, Åbo Akademi University and Kongsberg Maritime.

New predictive maintenance technologies need to be applied safely. This will require changes in the safety regulations, as well as in development and testing practices.

The maritime safety regulations are set by various actors, such as national flag state authorities, classification societies and the International Maritime Organization (IMO). The international and national regulations mostly do not set exact limits for the number of crew members on board, but some changes in the legal framework might be needed, as the idea of autonomous operation is not considered at all in the current regulations. The needed changes in international conventions are currently being studied, but the changes are expected to be very slow. Classification societies have taken an active role and have published some high-level guidelines to support the development of autonomous and remote operated ships. These guidelines are still quite high-level and do not cover many of the technical systems in detail. In practice, the first applications will likely be operated within the national waters of a single country. In such cases, the national authority plays a major role in approving the new technology and in issuing permits for testing activities. A step-wise approach, starting with small vessels operating in a small sea area, can be adopted to mitigate risks in new technology deployment.

The liability and insurance aspects are also just starting to develop, and they are internationally somewhat less harmonized than the safety regulations. The main questions revolve around the changing role of human and understanding who should be liable when an autonomous system makes mistakes. Currently, the liable party is usually the ship–owner, but in the future, this may not be quite as clear. Insuring the new technologies would need to start from consideration whether the predictive maintenance is predictable enough to be insured – in the first cases, this might be a challenge as there is no data available from previous installations.

Predictive maintenance is emerging in shipping, with the potential to improve efficiency, generate significant cost savings and create entirely novel business opportunities.

The potential business benefits of predictive maintenance in shipping can be significant. In addition to the reduction of the lifetime operating cost and extending the life of equipment and assets, predictive maintenance solutions can create totally novel opportunities for business activities. Moreover, better predictability of operations will enable more accurate decision-making. In the long term, the development may also lead to changes in the roles of different stakeholders in the maritime business ecosystem and expansion of the industry boundaries by enabling new, agile service providers to enter the market. As next steps in the Reboot IoT Factory project, the potential business models for deploying predictive maintenance solutions will be studied in more detail.

Experts involved in the study:

• VTT: Eetu Heikkilä, Minna Räikkönen, Pasi Valkokari 
• Åbo Akademi University: Henry Schwartz
• Kongsberg Maritime: Joni Keski-Rahkonen, Ari Vehanen, Mikko Mattila

Festum Software is a Finnish company that creates Robotic Process Automation solutions for companies. That means automatization of repetitive and monotonous tasks carried out by human workers. Software robots are not replacing people but letting them be the experts and concentrate on the work that requires thinking and intelligent decision making. Jorma Messo, Business Manager of Festum, tells us how to utilise RPA solutions in different processes with manufacturers.

Festum started to develop robots together with Scanfil, a global contract manufacturer and systems supplier. During the first development process they found out that Scanfil’s real need was automatizing purchasing. Purchasing needed to be faster and easier, and it contains tasks that could be taught to a software robot. The process included finding repetitive tasks, teaching the robot how to use the systems and perform the tasks. The selected Kofax RPA technology was based on Low Code technology and enabled fast implementation of the robot without traditional coding.

“Before Reboot IoT Factory, we had only little experiences in RPA mainly from the small-scale research we had done. Festum offered useful models how to build an environment in which RPA could be managed as part of the processes. They mapped and evaluated the potential targets for automation, formed and maintained the specifications and tailored the software”, says Sami Tervo from Scanfil.

“The project was ideal”, Jorma explains. “The work was open-minded and goal-oriented, and the technology used challenges interestingly the traditional solutions. Building trust was easy and working remotely went smoothly. Scanfil was very engaged in the project. Our company also got attention from new audiences.”

“We have only had good experiences: the standard office work is partly delegated to a robot and people have time to concentrate on more demanding tasks. The co-operation with Festum has worked well. Their experts have found solutions even to some more unusual cases”, Sami adds.

A Software Robot Goes Through Data

Festum was part of another development project as well: they worked with Nokia Oulu Factory to pilot RPA solution to material simulation. Manually performed simulation is a slow and demanding process. The amount of data is massive and there are several systems to go through. This task is strenuous to a human worker. The software robot was implemented in gathering the information and human expert comes along when it is time to analyse the data.

“Festum built the software based on existing model – how a human performs the simulation”, Jorma clarifies.

Daily material simulation creates visibility to material availability to production. This simulation is a manual process and requires high manual effort to complete. Together with Festum we have been piloting how Robotic Process Automation could be used to automate this process.

Collaboration with Festum has been working very well despite the covid situation and thus doing work mainly in remote mode. – Mika Kaivola, Nokia

Robotics from the Human Standpoint

When developing new solutions that involve robotics it is important to properly introduce the solutions to the people working in the factory. The robot is not going to replace humans. It helps to diminish the rush and unnecessary tasks. The benefits are clear: RPA helps to predict things; it enhances quality and helps controlling the processes. Humans have time and better resources to complete the analysis.

Anywhere there are repetitive tasks there are implementations for robotics. As Jorma says: “The technology is not the challenge anymore – at the start, a change in mindset to upgrade outdated work processes might be. There are right tools for every environment. RPA solutions are perfectly possible for the smaller companies as well as for the big manufacturers. The only thing that matters is data volume and the nature of tasks.

The next step of the developing industry is machine learning. Robots are connecting different technologies, industries, processes, and people. There is an exhilarating atmosphere in the field. These projects have taught us a lot of manufacturing. This is a very diverse and interesting field.”

Read more about Festum Software.

One of the most important tasks for a foreman is to ensure that the team has as little obstacles as possible disturbing their work. This applies also to “AI foreman”, a concept developed in the Reboot IoT project. AI foreman is supposed to assist the human foreman. It complements human skills with its own strengths, resulting in improved overall performance.

People working in factories these days spend way too much time in managing and scheduling orders. Furthermore, capabilities to react in sudden changes are not on a good enough level. Inventory management might fail if the storage runs out of a critical component, a worker with the needed skill profile might call in sick, a parallel project might reserve the needed factory equipment for a longer time than originally anticipated, and so on. Typically resourcing and scheduling is performed once a week or once a day, but because of the aforementioned incidents it is rarely realized in production as originally planned.

Together with VTT, University of Oulu, Gambit Labs, Parta Games, and FixUI, ABB is developing a system to assist order management and scheduling. This system, also called the AI foreman, has real-time knowledge of the factory’s state. It can also perform forecasts and notify of potential future incidents. It helps to speed up order management and improves the factory’s performance.

AI foreman can take into account the employees’ skill domains and levels. A future version will also start to gather and utilize employees’ own preferences in its reasoning. Light surveys can be conducted to collect feedback of various work phases. This can be used to plan an employee’s day so that it consists of pleasant and sensible tasks.

The same collected information can also be used for adopting new skills and career development. Diverse tasks often improve wellbeing and enjoying the work. In addition to the employee, also the factory as a whole benefits from multidisciplinary know-how because it helps resourcing and dynamic reaction to changes.

Gamification is in a central role in the project. When an employee completes a task enough times within the target schedule, she levels up and the AI foreman can use this information in its future planning. ABB’s factory has in the spring of 2019 had more orders than ever before, which has caused some challenges to make the deadlines. If the AI foreman had already been deployed, scheduling might have been easier to accomplish.

Standard Robot Interface is a device, which makes it possible to use one robot with multiple production cells. Robots, especially cobots are expensive, and utilization rate of those should be high. Standard Robot Interface connects the hardware interfaces (mechanical, electric and pneumatic connectors), as well as software interfaces so that it is easy to move a robot from one task to another thus improving its utilization rate.

Standard Robot Interface has first been tested at GE Healthcare factory in Helsinki. Based on the first tests in real factory environment, new features such as automatic gripper finger change system has been developed. Roboco (https://roboco.fi/) has made an implementation of Standard Robot Interface, specification and usage tests have done at GE Healthcare (http://www.gehealthcare.fi/), and research work has mainly been done at VTT (https://www.vttresearch.com/en). Standard Robot Interface has gained a lot of interest in Reboot IoT Factory project, and and there is plans to try it at ABB Vaasa factory.

Scale-up Proof-of-concepts are the finest the Reboot has to offer. They are the proof-of-concepts that are so successful that they are taken to another factory already during the project. 

Tester Predictive Maintenance proof-of-concept arose from the need of detecting when a tester is breaking down and not the products it is testing. Research was done towards how the testing data from the tester could be analysed in a way that it would indicate the wear and tear of the testing equipment. The first implementation was done at Nokia factory for electronics needs. The proof-of concept were a success and the algorithms could be applied to find the testing equipment failure. As data is gathered more the more efficient the analysis become.

Often the simplest solution to a problem is also the most effective. In order to discover such solution one has to take a step back and look for new points of view. To think outside of the box, as the saying goes. This is what happened at Nokia in Oulu, in Reboot IoT Factory’s TesterWatch project. As the outcome a novel way to test the functioning of base station products was invented.

Before shipping to customers, the proper functioning of base station products is ensured with test devices. Parallel identical test devices are used for scaling up the operations. If one tester gives systematically different readings than its peers, the problem is likely in that particular tester, not in the product which is being tested. For example the connectors used to attach test devices with products start to wear down over time. This causes contact failures and the test results start to fluctuate. Anomalies result in sending a real time alarm to the maintenance team. This way quality engineers can immediately engage in repair activities. Before TesterWatch they run reports in mornings and based on those performed the needed actions.

The algorithm itself and the solution based on it are straightforward. They were born in a brainstorming session among the project team followed by statistical analysis. The analysis and number crunching was completely agnostic to base station technology. Exactly this is the out of the box approach. The solution most likely would not have been invented if the project team had consisted only of the base station factory’s own workforce.

The approach can be extended to other industry domains. Reboot IoT consortium has introduced it in GE’s factory in Vallila, Helsinki where predictivity ability has been researched and implemented. Naturally also Nokia is going to implement it in its other factories. In Oulu it is already used in production 24/7.

TesterWatch is a prime example of digitalization project which delivers new process innovations rather than tangible gadgets. Testing base station products is now more reliable than it was before. Time and money get saved.

Even though the current version of TesterWatch is quite simple, it can be developed towards machine learning and AI. The aim is to enable predictive maintenance. Today’s maintenance team is called when something has already happened. In the future they can be paged before things go wrong. The system can for example keep log of test devices’ functioning and propose changing some parts to a particular device once it reaches certain age.

TesterWatch can also be attached with a broader digital system of a factory.

Digitalization has enabled consumer product supply chains take significant leaps during the past two decades. Information flows from one player to the next and the customer can observe the parcel’s journey almost in real time, using a computer or even a smartphone. Companies present in the ecosystem can estimate the supply, demand, and delivery times of products with high accuracy. 

In a business to business (B2B) environment the situation is different. Consumer products are typically standardized and volumes high. In contrast, in B2B supply chains the products are tailored and volumes low. Products are often unique and targeted to a specific need. Also the purchasing process differs: you can buy a consumer product by clicking a couple of buttons on a webshop. Purchasing between companies takes longer and is more complex, often including for example tenders. 

Balancing supply and demand 

One central objective of the Reboot IoT Factory project’s second phase is to estimate demand in a B2B supply network. As in the consumer market, also in B2B supply and demand need to be balanced. In a business like this, where the product varies a lot, estimations take for example the order backlog, ongoing sales processes, and tenders as input. 

Predicting demand is challenging, but it has to be done in order to evaluate future product configurations. Product configurations, in turn, have impact on the kinds of components and materials that need to be ordered by the manufacturer. And all this needs to get done in time so that the ordered products will be shipped according to the agreed schedule. In theory this problem would go away if everything imaginable was stored in the warehouse. In practice, however, this is not a realistic option. 

Inventory management is a central process to be optimized in many if not all business domains. In B2B it is even more relevant than in consumer market. This is because the products vary so much, as mentioned above, and this would require maintaining even bigger an inventory. Naturally, B2B is not homogeneous and even within the Reboot IoT project there are different companies in this respect. For example Ponsse forestry machines are somewhat standard compared for example with KONGSBERG propulsion systems, where almost each product is unique. This is why it is enough for Ponsse to update its forecasts relatively rarely. The electronics contract manufacturer Scanfil, in contrast, needs new updates on a weekly basis. 

Joint processes and information management in the core 

There is still a long way to go until B2B supply networks are fully automated. Various players in the network typically have their own IT systems and interfaces between those are only partly implemented. That is why it is central to concentrate not only on technical integrations, but also on defining joint operating models and processes. The ecosystem needs to function on common rules and share enough information to guarantee that the products get shipped in time and as ordered. 

There are roughly three main streams in a supply network: information flow, material flow, and money flow. The information flow is extremely important and has impact in the other two. Detailed and up-to-date information is crucial for the products to be manufactured and shipped as agreed. This is a basis for the money to flow from one company to another and profitable business to be made. 

Reboot IoT Factory concentrates mainly on automation, simulations, robotics, and other highly technical means for streamlining industry. While doing this, however, it does not forget the human worker: one essential objective is to apply various methods for helping people to cope at their work. 

Factory workers’ average age is increasing in many industry domains. This poses all kinds of challenges for carrying out the tasks at work. It is important to try and make the work easier for the older employees. At the same time job profiles are made more attractive also for younger generations. 

Taking care of job wellbeing means long-term optimization 

Often the productivity of a factory is measured in a short-term fashion, attempting to maximize for example the amount of products manufactured during one work shift. Such approach can easily overlook for example ergonomics. This can later on manifest as dissatisfaction at work and sick leaves. As a long-term rule of thumb one can state: the better the work wellbeing, the less sick leaves and the more production. 

Technology has an important role in advancing work wellbeing. Automating routine work helps to allocate time in tasks that are cognitively and socially more meaningful. As an example: when the team leader spends less time with filling excel sheets and reading pdf reports, she can direct his attention to organizing work and interacting with the team members. Reboot IoT’s AI Foreman concept aims at exactly this. 

On the factory floor the worker benefits from technology for example via automated work rotation so that the same muscle groups are not burdened for too long at a time. Worker’s “digital twin” tells what tasks he has performed earlier, what is his skill profile, and what are the limitations. The system can also recommend micro breaks and stretching the muscles. Collecting, utilizing, and storing data like this has to be done so that the personal privacy of the worker is carefully respected. 

Acknowledging personal preferences in addition to the skill profile 

One interesting approach in improving work wellbeing has to do with making use of the worker’s preferences in organizing tasks and work rotation. This will be piloted during the second phase of the Reboot project. Versatile and diverse tasks make the work more meaningful, in particular if the worker can influence on how his own workday is organized. 

Reboot IoT Factory works to re-imagine the automation levels of existing processes in the factories. Developing and using AI, intelligent algorithms, software robotics and robotics process automation the project helps the factories develop new intelligent ways for production. Automatic quality control is a field with promising applications of automatized visual inspection.

We talked with Ilpo Tanskanen and Markus Broman from ABB Porvoo and Vili Kellokumpu from VTT about the theme automatic quality control and more precisely, about visual inspection. The field is moving towards automatization with machine vision, learning systems and neural networks. Visual inspection is also a handy way to acquire measured information, data, and quality control.

“Opportunities created by developing digitalisation help the factories take robotics and automation to the next level. Reboot IoT Factory is a sure way to achieve this goal”, say Markus and Ilpo. “The project offers different approach to innovations: the participant SMEs create solutions for the factories and productize them. The researchers bring in the scientific approach. The way of working is different than we have used to in corporate life. In business it is usual to run rapid tests and set mind towards future needs. Now we have a large knowledge and research base and testing laboratories that we can use. We get help to our projects and achieve better understanding along with the better results.”

Factory as an Innovation Platform

Factories’ roles as innovation platforms make it possible to successfully merge different approaches to develop new solutions. The co-operation was fruitful. The first Proof of Concept developed with ABB Porvoo was about visual inspection and was completed in co-operation with VTT. It dealt with quality control of a surface of the product and finding aberrations. The purpose of this PoC was to find the right equipment for this task.

Before the Reboot IoT Factory project, 5 pairs of human eyes carried out the quality control procedure and the factory was eager to find another, automatized solution. The focus was on researching the methods, not building any device. The Reboot IoT Factory made it possible to work with a world class expert.

The expert, Vili Kellokumpu, worked with the technical solutions of PoCs together with the companies. His area of expertise are optical measurements, machine vision and imaging.

Vision: A Self-Correcting System in Quality Control

In the future, a self-correcting system lets the computer know how to develop production. The first step, finding the errors is already working. The technology for correcting the aberrations automatically is existing, but it has not been possible yet to utilise it in producing plastic objects.

Markus and Ilpo are enthusiastic about the project:

“Strategic PoCs can offer these future solutions. We will continue automatization and digitalisation in ABB Porvoo and invest in skills. Reboot IoT Factory work has been a great window to the future. We feel that sharing knowledge serves strategic thinking and the technological knowledge increases in this kind of co-working.”

Increasing automatization and digitalisation is an indisputable trend of industry. Vili clarifies that automatised visual inspection can use wavelengths invisible to human eye making it more precise.

“Neural-networks have made a breakthrough in image analysis and when suitable imaging techniques are applied the quality of the image is so good that you can spot the problems in quality of the product inspected.”

Collecting production data and utilising it is the future. Machine learning and AI complete the vision.

Responding the Factory’s Needs Ideal for All

The theme, automatic quality control and the PoCs related to it came from the factory. It is reasonable to start with them as the other operators are often more agile and the factory moves slowly.

The beginning of the PoC work started with samples. A prototype was created after that in the lab. The PoC was first demonstrated in a small scale. Rapid testing and productizing finalised the work. The results were good.

“Results on a larger scale include delivering important knowledge based on scientific research, but also actual products. Technical development of machine vision made this possible”, Vili explains.

Markus and Ilpo unwrap the current challenges: “Objects with shiny surfaces and 3D objects. The solution for them must be tailored as there are no such equipment in the market that would be ready and suitable for our needs of quality control of surfaces. Now we know which technology can be applied and which not. The process was smooth: rapid testing, minimum investment and scientific validation were the cornerstones.”

The work continues with the second PoC. The financial investment is on totally different level now when it is clear what methods to use. The situation is ideal for the factory and the company developing the solution. The product development moves ahead based on the needs of the factory. That is exceptional in corporate life.

This project was part of Reboot IoT Factory’s Grand Challege 2: Robotics Fusion

Text: Veera Kiurujoki, Design Inspis Oy

Picture: Design Inspis Oy

Voice assistants, such as Amazon Alexa and Google Assistant, are being used daily by hundreds of millions of people at home, in the car, or on the go. Even though speech is natural and effortless way of interacting with devices and services, voice control is not yet widely used in factory environment. As part of the Reboot IoT Factory challenge, Nokia and Creoir wanted to bring voice control into Nokia Conscious Factory to validate the voice user interface in a factory environment.

Creoir Oy is a technology and product development company specialized in speech technologies and voice-enabled products and services. Creoir Offline Voice Solution SDK (Software Development Kit) enables a flexible and effortless way of bringing Automatic Speech Recognition (ASR) and Natural Language Understanding (NLU) to devices that need to work in challenging noise conditions.

Hands-free at the Factory

The operators and workers at the factory have typically many concurrent tasks at hand. The challenge is that operators need to interrupt the current tasks and free their hands to access instructions and manufacturing data, or to control computers and machines.

Security and reliability are extremely important in any factory. Therefore, it is not possible to use major cloud-based voice platform providers like Amazon or Google. Also, to reach robust speech recognition performance, attention needs to be paid to noise reduction and voice processing.

Smart Factory Voice Assistant

The solution in the PoC project was to create a Smart Factory Voice Assistant, which enables a hands-free voice user interface to access Nokia factory data. Creoir Offline Voice Solution SDK was used to create custom wake-word and needed voice commands to access the Nokia factory system. Special voice processing and noise reduction algorithms were deployed to guarantee far-field voice control even from a distance of five (5) meters.

The Smart Factory Assistant is a stand-alone product with an integrated speaker and two microphones. All speech processing is done on-device, and MQTT (Message Queuing Telemetry Transport) interface is used to access the Nokia factory back-end.

The deployment in the project was straightforward; at the beginning of the project MQTT topics based on intents were agreed upon between Creoir and Nokia. Creoir was responsible for speech to meaning conversion and the Nokia team implemented the needed action code at the Nokia back-end.

The noise conditions were recorded on-site, and these recordings were used in the configuration and parametrization of the noise processing algorithms.

Benefits and Next Steps

Smart Factory Voice Assistant is now in use at the Nokia Conscious Factory control corner to provide an easy way of getting quality data of the ongoing production. The plan is to extend use cases to support operators during the assembly process on the factory floor. Also, voice control for mobile robots is being investigated.

There are also many other opportunities for voice control in a factory environment. Feel free to contact Creoir if you would like to learn more about adding voice to your factory!

“Speech is the most natural way of interacting between people. At Nokia, we wanted to find out how we could use voice control for machines in a factory environment. Working with Creoir was a good experience for us – their expertise in speech technologies and embedded products was essential to achieve the targets of the PoC project. We are looking forward to continue co-operation in the future” – Jouni Herronen, Technology Program Manager, Nokia.

This project was part of the Grand Challenge 4. Read more: GC 4: Digital Production

Find out more about Creoir Oy here.