Naast de keynote sessies, zijn de live demonstraties van onze laatste innovaties 1 van de belangrijkste aspecten van het Flanders Make Symposium. Hieronder vind je een overzicht van de demo's die je op het event kan beleven. Je kan deze demonstraties vrijblijvend bezoeken gedurende de dag, net zoals op een gewone (vak)beurs, maar je kan je ook inschrijven voor een georganiseerde demo tour tijdens de lunch.  

Neem deel aan een demo tour

Tijdens de lunch bieden we drie georganiseerde tours (1 per competentie cluster) aan. In het overzicht hieronder kan je zien welke demo's onder iedere competentie cluster vallen, zo heb je een idee van de inhoud van de tour. 

Je kan je keuze voor de demo tour doorgeven tijdens het registratieproces. 
 

Demo's in de Production tour

2022-17: Adaptive ergonomics Work Instructions

Heavy and repetitive assembly tasks can lead to injuries for the operators. Using the right ergonomic way of working could prevent this. Therefore, we need operator support systems which provide operators with correct information and instructions to work in an ergonomic way. 

In this demo, you will find a newly developed Digital Work Instruction (DWI) which monitors the operator’s ergonomics in real-time and provides the operator with the required ergonomics information. In addition, the system is capable to adapt the content of the work instructions to the detected potential risks. Using this DWI, you will achieve an overall reduction of your average Rapid Entire Body Assessment (REBA) score for a manual assembly.  

2022-27: Robotic kitting operation

Programming a kitting robot is a complex and challenging task. In this demo, we illustrate our newly developed skill-based robot programming framework. This framework makes automated kitting operations with minimum programming effort feasible. It is supported by a world model generated from a CAD model and a planner to provide an optimal sequence of robotic movements.  

Using our framework, companies will save time as the robot programming could be done in mere seconds (assuming that the CAD-modelling of the world model and the hardware configurations are already available). In a traditional approach this could take hours to days depending on the complexity of the scenario.  

On our Symposium you will find an AGV with a robot arm which will kit workpiece variants and transport them from a warehouse to an assembly cell. To this end, the operator defines the goal of the kits which automatically triggers the generation of the optimal sequence of robot movements. 

2022-32: Active vision for defect detection

Due to the high variance and quantity of products, mass customisation makes automatic quality inspection challenging. Customers ask for an increasing quality, often meaning that human quality control is not sufficient anymore. To meet these demands, robot can be used. However, these robots need programming, resulting in a need for robot experts.  

Therefore, we have developed an automatic path planning based on the CAD-file of the produced part, and (pre-trained) defect parameters. An auto annotation tool lets you annotate on one image and automatically annotating the other images. Implementing our path planning results in:  

  1. 50% quicker decisions for inspecting complex parts and up to 100% correct detection for low-contrast defects. 
  2. More than 50% reduction of configuration time of automated inspection using viewpoint optimization.  

On the Symposium we will show you the entire workflow, from path planning to defect detection.  

2022-36: Assembly of parts using AI models trained on synthetic data

Digital work instructions are indispensable in today’s industry. In this demo, we offer operators stepwise instructions based on AI models for a complex parts assembly. This way, we avoid wrong assemblies and increase your productivity. The cost-effective AI models that we have developed are trained using synthetic data. Using these models, we can avoid wrong assemblies and assist operators with details regarding the orientation of subparts.  

On our Symposium, you can use an assembly station with digital work instructions augmented with a 2D camera and our AI model. Additionally, you can compare the performance of a traditional DWI system that does not consider part orientation.  

2022-41: AR (Augmented Reality) visualisation of IoT data of a Smart Factory

When performing maintenance or troubleshooting machines or a production line, it is important for operators or maintenance engineers to have access to the correct data. Ideally, they receive an in-context visualisation of this data, as this will fasten the maintenance process and improve the productivity rate of the machine or production line.  

That is why we have developed an Augmented Reality visualisation of IoT data. In a smart factory, the machine (or production line) data is uploaded to the edge or cloud. Using our tool, we can visualise this data without the operator having to switch context all the time (e.g., between machine and laptop). This makes for a faster access to data and thus a higher productivity.  

On our Symposium we demonstrate this technology using our SmartFactoryBE . The SmartFactory assembles and disassembles flashlights and publishes its data through an OPC-UA server. The actual main contribution of Flanders Make lies on transferring the OPC-UA data to a time series database on a modular IIoT platform that can be implemented both on the edge and on the cloud. The data is then transferred to an AR application that can run both on a tablet and on a smart glass (Hololens).  

2022-42: Converter for Digital Work Instructions

Creating digital work instructions (DWIs) is a time-consuming task which makes their adaption on a large scale within a company difficult. In addition, instructions are programmed in a platform-specific format and are therefore very platform-dependent. However, a lot of information from existing procedures, instructions or checklists is sufficiently structured to be automatically converted to Digital Work Instructions. To do so, we have developed a generic model for work instructions, which can convert the collected information immediately from e.g., a PPT, Word or Excel file to one or many commercial or in-house DWI platforms. 

By automatically parsing several (semi-)structured document types to generate DWIs in a uniform format (ISA95), we reduce the start-up efforts and allows easy interchanging of DWIs between platforms. This way we can create DWI’s up to 90% faster. 

On our Symposium you will find this demonstration in our Make Lab. Here, you can:  

  1. Select different input documents for the same assembly procedure: Excel, PowerPoint, Images, Videos, etc. 
  2. Select an output format to show the instructions: Proceedix (Smart glasses/Tablet), Arkite (Projection), Azumuta (PC/Tablet), etc.  

Then, we start the conversion (+- 1min) and highlight the way the information is collected from the different sources in the uniform ISA-95 standard. 

2022-43: Robot programming by XR demonstration

Programming industrial robots and cobots is a tedious and time-consuming task and requires a lot of technical expertise. This demo displays the possibility to capture operator skills, using XR technology, and have them being translated in robot trajectories. Supported by an intuitive and user-friendly XR interface, operators can create robot trajectories without the support of robot experts. This:  

  1. Decreases robot programming complexity such that it can be performed by a process expert without robot programming expertise. 
  2. Facilitates offline programming, thereby minimizing production downtime for the robot. 

Overall, our tool decreases the robot programming time, and therefore production down time, by 50% on average.  

2022-59: Infraflex – Flexibility by reconfigurability

Mass customisation is a growing trend. Adapting your production line to it can be quite challenging. Therefore, we have developed Infraflex. Infraflex is an open architecture of multiple reconfigurable workcells for flexible assembly. This allows for a flexible reconfiguration of the physical assembly infrastructure which facilitates the fast product change over. The goal is to achieve a one-piece flow (Cfr. moving products/parts efficiently by planning your workflow based on the product and its needs instead of the organization or available equipment.  

With Infraflex, we can achieve a 50% faster change over depending on the availability of the resources. Discover it live in action at our Symposium.  

2022-61: See where your money leaks out

Pressure leaks in a company's systems can be problematic and need to be addressed as soon as possible. To inspect pressurized containments like valves, tubing, or kegs, we have developed an inspection tool with minimum setup time, that uses Mixed Reality (MR) visualization of ultra-sound emissions combined with optical perception. 

Using our tool, we achieved a 70% cost and time savings in inspection and servicing pressurized containments.

Demo's in de end-to-end design operation tour

2022-10: Hybrid AI applied on a foil processing machine

In this demo we present how the use of hybrid AI combines the strengths of data and models to identify unknown parameters more quickly. Knowledge of such parameters allows to better operate and tune machines, increasing their performance. The physics-based backbone allows for faster training of the hybrid model. 

There is a significant need to identify unknown parameters in manufacturing systems using less data. So instead of doing dedicated measurements of material properties, we offer a solution to identify those parameters live during the operation of a machine. Moreover, our approach requires less data compared to traditional AI, leading to accurate results in less time. Possible parameters to quantify are ‘reduction in data/training time’ and ‘increase in performance’. Preliminary research shows that a hybrid model requires up to 6 times less data/time to achieve the same model accuracy compared to a non-informed neural network. 

At our symposium you can generate your own oscillating training timeseries around a working point, by controlling the speed and the minimum and maximum displacement of the web foil. The oscillating trajectory is generated and the hysteresis loop on the displacement-force plot is visualized on the screen. The elasticity and damping of the material are derived from the generated oscillating trajectory in real-time. 

2022-20: First time right profile design (for Production) by CAD feature extraction and evaluation

“First time right” product design can save a company a lot of costs. To assist CAD designers in their tasks, we have developed a feature extraction on CAD files. This makes it possible for designers to automatically evaluate production constraints during product design. Which leads to less design cycles, first time right designs and it makes expert production knowledge accessible to designers.  

Our software:  

  • Automatically detects the areas on which the check must be applied  
  • Checks the constraints 
  • Visualizes the outcome to the designer.  

This evaluation takes only seconds, is up to 95% correct and removes dependencies on expert human evaluators. At Reynaers this speeds up the design cycle (per profile) with about 2 weeks. 

At the Symposium, you will be able to use the software and first-hand experience its advantages.  

2022-21: Slash commissioning time by 30% more machines using AI based hot starting

We have developed an AI method to speed up controller commissioning. This method provides a faster way to commission controllers, for cases where a machine is added to a fleet of similar machines which are already commissioned and for which historical data is already available. Currently, operators typically choose the initial settings – based on their experience - to try on the new machine. But as this is quite difficult, they must spend more time afterwards to fine-tune the settings. For machines that perform multiple tasks, this is even more difficult as each task requires specific tuning during commissioning.  

Using our AI method, we save (on average for all task/system combinations) 35% of iterations and increase the performance during the first trials by 70%. 

For systems performing many similar tasks, the expected impact is to bring the commissioning time down from multiple days to a single day. Also expected is that since commissioning can be performed with improved initial performance, the tests can be run with less safe settings, further reducing the needed commissioning time. 

On our Symposium, you will be able to apply this AI method to a bar linkage. Combining a task to the bar linkage you will see how the method learns to complete each task more efficiently and faster. You can compare our method to benchmarks and get visual proof in the GUI.  

2022-22: Make new controllers quicker using a virtual 3D environment

In this demo we will demonstrate how a model- and operator in the loop approach can be used to develop and validate controllers offline, resulting in significant reduction in development time, and reduction in cost of field testing. Realistic yet complex models can be made nowadays to do this for complex applications. 

To develop new controllers that interact or support operators, working directly on the real machine is not preferred since it is impractical and very costly. By emulating the machine and the environment and putting an operator in the loop interacting with this emulation, it becomes possible to get operator feedback offline, and to fine-tune control software based on this feedback. 

An agricultural emulator was built including photorealistic visualization. This was used for model- and operator-in-the-loop testing, allowing to offline develop new controllers for an agricultural application that an operator can already experience. At our symposium, you can take place in the emulator and operate a forage harvester driving over a field. You will be able to enable or disable automation features, so either the user must do several tasks himself simultaneously, or so he only must drive forwards while the automation does the rest. 

2022-45: From data and tacit knowledge to actionable insights in the production process using knowledge graphs

Manufacturing companies capture enormous amounts of data but are often struggling to easily create added value out of these data. This is because: 

  1. The data is stored in different locations;  
  2. The data is stored in different ways that require technical understanding that exceeds the problem domain knowledge and;  
  3. The knowledge that is necessary to turn this enormous amount of data into actionable insights is spread over different roles in the organization (manufacturing engineers, data scientists, operators, ...). 

A knowledge-graph-based digital twin cannot only help data scientists in retrieving data stored in different ways and locations, but also by explicitly modelling production knowledge. As such, the knowledge graph supports fast and agile creation of AI-models for root cause analysis, prediction, and optimizing production parameters. 

Therefore, we developed data models and supporting methodologies that allow the creation of knowledge graphs that include intuitive (tacit) production knowledge on top of providing easy access to heterogeneous data. 

By using the production knowledge that is stored in the graph to explore the enormous data haystack, we can now find actionable insights that were previously undiscovered. Visit our Symposium for more details. 

2022-50: Topology optimisation for stiffening of thin-walled structures

How can I obtain the best structural design for a certain application with minimal engineering effort? If this is a question that your company is struggling with, this might be an interesting development for you.  

Structural topology optimization is not just a high-end approach for 3D printed parts. It can also enable thin-walled designs and allow for faster and better designs than what human designers can typically produce. To design stiffeners in thin-walled structures (as common in injection moulded components), we have developed a shell-based topology optimization approach which leads to: 

  • Shorter optimization times 
  • More manufacturable designs.  

This development results in a:  

  • 50% reduction in design time for a +/- 50% percent increase in performance (stiffness) compared to manual design; 
  • A 90% reduction in optimization time for similar performance, and much more manufacturable design compared to regular topology optimization. 

On our Symposium, you can compete with our topology to design a stiffening structure for a thin plate. You will have to select the location of the stiffeners, while accounting for a maximum weight. Both you and the topology optimization routine will start at the same time. When both are finished a numerical simulation of the deformation of both designs is shown to demonstrate who performed best.  

2022-51: Digital Twin based estimators for hybrid vehicles

In this demo we will show how virtual sensors allow to extract operational parameters from low-cost sensors during the operation of mechatronic systems like vehicles, leading to a 50% reduction in uncertainty on (safety critical) parameters like tire cornering stiffness. A digital twin-based approach moreover allows a significant reduction in the setup time of these virtual sensors. 

We will use an LMSD concept car, a hybrid development platform, to demonstrate the real-time implementation of a wheel cornering stiffness estimator under low excitation conditions. First, we will describe the concept of the stable estimator and how it has to potential to increase the safety of vehicles. We will highlight the differences with respect to previous estimators (mainly long-term robustness and performance under low excitation). 

We then perform a short electric ride to showcase the robust and online cornering stiffness estimation on the vehicle. If desired, we can modify the inflation pressure to show that the impact on the cornering stiffness can be perceived by the estimator. We can also add a vertical tire disturbance to show a second estimator running in parallel.  

2022- 52: How the formalization of manufacturing knowledge will accelerate the flawless CAD design

How can you extract the manufacturing knowledge from your senior experts? This is an important question regarding the continuity of your company. By making manufacturing knowledge formal, it becomes less dependent on senior experts who are not always available and whose knowledge might get lost e.g., when they go in retirement. In addition: 

  • It accelerates the learning curve for junior process experts;  
  • CAD Design engineers get quicker and more independently to a design free of manufacturing errors; 
  • The learned design lesson can 1) improve the manufacturing knowledge for the process expert, and 2) improve the formal knowledge. 
  • It streamlines the interaction between the CAD design experts and the process engineers, bringing benefits to both parties.  

2022-60: Maximum performance with minimal energy

This demo shows how you can unlock the optimisation potential in machine design to minimise energy usage in a convenient and accessible way. This while accessible tools already used by machine designers can be used to maximize performance and at the same time minimise total cost of ownership. 

At our symposium, a corona ventilator prototype will be in use. The prototype clearly visualizes the design parameters. Moreover, you will be able to tune the design parameters in a corresponding simulation to immediately visualize and quantify the energy usage of their design. In an overnight simulation we achieved an RMS torque reduction of 67%. 

Demo's in de Motion Products tour

2022-03: Help autonomous mobile robots find their way

In this demo we will go deeper into robust localisation without adding infrastructure in dynamic environments. At the Symposium, you will see an autonomous mobile robot – MIR 200 (AMR) that drives around in a warehouse-like environment facing challenges that state-of-the-art Simultaneous Localization and Mapping (SLAM) cannot handle. At the Symposium, we will demonstrate 2 scenarios: 

  • For state-of-the-art SLAM, a dynamic object (e.g., a person or other AMR) may impact the localisation of the AMR. Coupled to the position control of the AMR, this can make the AMR move unwanted. You might have experienced this yourself when travelling by train, as the train next to you starts moving, believing you to think that your train is moving. We tackle this by adding object recognition to SLAM (making the SLAM aware of what it is seeing), which adds robustness against this situation.  
  • To demonstrate the smart recovery strategy you, as a visitor, can press a button while the AGV is moving from A to B which will initiate a localization loss. To solve this problem, the AGV will track back 5 metres and will block the area where the localization loss happened. Subsequently, the AGV will plan a new path, avoiding this to happen again. 

In summary, our research results in a more robust SLAM in dynamic environments by using semantic information to better understand the context.

2022-05: Zero-effort indoor localization

In this demo we use Ultra-Wide Band localisation to automate localisation and tracking of assets and people in the industry. This easy-to-install localisation technology is ideal when you have a temporal localisation need, for instance when taking inventory, maintenance on machines, harvesting, etc. Not only does the system has an accuracy of +/- 10 cm, but it also has a very short installation time. In a matter of minutes, you can set up this self-calibrating system by placing 3 to 4 fixed reference points (anchors) in the area. All people or assets that you want to localise and track must wear a small, portable device. On a screen you can track the position of every device, allowing a quick overview of all assets/people. Furthermore, upon moving the anchors, the system automatically updates, allowing you to cover any additional area without any set-up time.  

In summary, our UWB localisation technology reduces set-up time with 90%-100%, is self-calibrating and updates automatically when the anchors are moved.  

2022-07: Multi-machine energy management over a DC grid

In this demo you will discover a cheaper, more robust and future-proof energy supply for machines. We have developed an industrial DC grid and implemented various forms of real-time energy-management. Additionally, you can discover a tool we developed to design a DC grid and quantify the impact compared to other state-of-the-art solutions. This research resulted in several advantages for the industry: 

  • Integrated uninterruptable power supply (UPS) 
  • 30% less peak power/connection capacity 
  • 10% lower losses 
  • 85% fewer harmonics 

Additionally, DC grids are better compatible with electric vehicle chargers & photovoltaic production.  

On our Symposium you will see a live DC grid on which an energy management algorithm is running. We will visualize the main KPI’s on a monitor, which show the results of several scenarios that will be demonstrated.  

2022-08: Smart machine monitoring based on event-based raw data sets

In this demo we will highlight a new hybrid approach for machine monitoring based on raw data sets. Typically, the monitoring of machines is done using feature data on a certain time base, e.g., an average temperature or calculated RMS value. However, with this method you are missing valuable information throughout the intervals.  

This why we developed a new approach that transmits raw, high frequency data, but limits ourselves to a complete yet compact data set. We created a framework that stores raw data sets in the cloud to provide more useful insights, while redundant data is discarded. On our Symposium you can experience this research first handed. 

The goal is to have compact but representative databases that are sufficiently detailed for transient faults. A quantitative number is highly depending on the application and the functioning of the machines itself. Minimal dataset will reduce the cloud and transmission cost which can become significantly when applying it on a fleet of machines. 

2022-09: Online reconfiguration in modular motor drives

In this demo we will visualise and demonstrate how the concept of a multi-agent control in modular motion drives realizes fault-tolerant operation. The requirement of fault tolerant operation will be relevant for critical applications, e.g., compressors, where a fault in a winding or a PE module is a dominant source of failure. 

At our symposium, the demo set-up contains a modular motor and a modular PE converter. The different modules are arranged in such a way that you can clearly distinguish the different agents. The DC bus voltages of each module and the torque and speed of the motor are visualized on a screen. You will be able to set the desired motor torque and observe that the setpoint is followed, speeding up the motor. When activating the fault trigger, the functional isolation of this module is initiated. When you again change the torque setpoint, you will be able to observe that the decentralized motor control can still track these setpoints. 

2022-14: Modular Motion System

In this demo we will show you how modular actuation can increase productivity up to 25% because it can provide solutions for high speed and highly dynamic motions to increase production rate, scalability, fault tolerance or to lower internal dynamic of these systems. 

The demo holds a comparison of the non-modular reference with a modular reference on reciprocating crank motion and a demonstration of the internal dynamics level in the continuous crank motion. At our symposium you will be able to select 3 motion profiles and control the speed, through a button and throttle (both will run at the same speed). While that happens, KPI's can be monitored on a screen.  

2022-15: Multiload real time virtual sensing in flexible mechatronic systems

In this demo you will see how a real-time joint input force estimation framework allows the accurate identification of multiple concurrent loads on a complex industrial structure. We will show you how Kalman filters for joint input can estimate multiple concurrent loads on a flexible structure where it is practically infeasible to place force transducers. 

The interactive setup at our symposium consists of the subframe structure of a Range Rover Evoque. To this subframe there is a shaker attached which causes an (unknown) baseload, and several positions are marked where a participant can excite the structure through hammer excitation. A display shows the (real-time) estimated inputs on the system. When you excite the structure you will see the forces from the shaker, as well the estimated hammer impact at the identified location.  

2022-19: Up to 3 times less maintenance through prognostics without the need for historic data

Predicting the remaining useful life of industrial machines is crucial for companies. However, due to a limited amount of available data predicting this can be very challenging. In many cases, existing state-of-the-art solutions are not sufficient. Therefore, we have developed algorithms for fault detection and prediction of remaining life span of bearings, based on data-efficient AI and digital twin technologies. Our algorithms perform significantly better than current state-of-the-art benchmark algorithms, even when there is only a limited amount of training data available.  

On our Symposium, you will find an accelerated bearing lifetime test. Here, we run two novel approaches based on vibration data for the estimation of the remaining useful life (RUL) of components & systems. These novel approaches exploit novel features, weighted feature merging and synthetic data, and provide accurate results. The RUL is computed in real-time from historical data pulled from the cloud (in real-time) and shows the results in an interactive dashboard.  

2022-47: Exponential decrease in development time of systems with variable topology

This demo demonstrates and validates tools to develop a controller that accommodates varying dynamics as well as handles switching/variable topologies on a pneumatic system. These tools can then be beneficial to the industry to handle control applications that require to both varying dynamics and varying topologies.  

At our symposium we will demonstrate this principle in real-life. The user will be able to interact with the system and change its topology in real-time, while seen the controller react to this change, as compared to a classic control tuning. The results can be viewed online and will indicate an increased efficiency in control by limited system experiments due to efficient system modelling and controller architecture. 

Bestel je tickets hier