Digitalisation

From data science to digital twins

Digitalisation

From data science to digital twins

Digitalisation has long since become a key part of the industrial value chain – and it offers new opportunities for efficiency, optimisation, automated processes and smart systems. Bertrandt is supporting the digital transformation in production processes and beyond.

We are involved in a wide range of sectors, from the automotive industry and medical technology to mechanical and plant engineering. Whether it is a question of data science, digital twins, artificial intelligence or predictive maintenance, we offer our customers a complete digitalisation process from a single source, from consulting, development, and implementation, right through to support.

 

Data science opens up many possibilities to acquire new knowledge from data, to optimise existing systems and processes, and even to develop new business fields. We advise and support companies in using data science for their purposes and to take the next step in the digitalisation process – whether it is the next digitalisation step, the right machine learning approach, optimisation potentials in processes, profitable data evaluation or making data-driven decisions.

In addition to technical expertise, successful data science projects also require specialist knowledge of the industry itself and a deep understanding of development, design, production and support. Bertrandt’s 40 years of know-how in various industrial sectors is therefore a major advantage compared to pure big data service providers. Our data science solutions are applied in particular in the fields of manufacturing, autonomous driving, medical technology, Industry 4.0 and business analytics

Some of the services in our data science portfolio

  • Determination and exploitation of optimisation potentials
  • Efficient management and structuring of large quantities of data
  • Training of artificial intelligence for automation
  • Pattern recognition and prediction of deviations and systems
  • Clear representation of parameters (dashboards)
  • Efficient troubleshooting in the event of system failures

 

Examples of data science projects

Anomaly detection in the vehicle powertrain

Unpleasant jerking of unknown origin occurs during gear changes. The jerk is correlated with other system parameters. The strength of the correlation provides information about the cause of the jerking. A key additional benefit is that automated anomaly detection will result in higher cost efficiency in future vehicle series.

Validation of data quality

Corrupted camera images are recorded during test drives. (Runtime) optimized image analyses are used for error detection. The improvements thus achieved include the early detection of quality problems in test data, (time-) optimised testing procedures, the avoidance of unnecessary test drives, and cost efficiency due to the filtering out of bad data. 

Management scoring

The quality of a company’s management can only be evaluated subjectively. In order to provide a measurable evaluation of management quality, mathematical optimisation is applied to develop a reliable score. The consequences of introducing such a score are a data-driven derivation of necessary preventive measures and predictive calculations.

The application of virtual representations of objects and systems opens up new knowledge and opportunities for process design:

  • Improvement potentials: digital twins can quickly detect weak points in development and design, thus enabling improvement potentials to be utilised.
  • Implementation of new service processes: mechanical engineers who use a virtual twin can continuously monitor their machines at the customer, optimise their operation and also use intelligent algorithms to perform predictive maintenance. All this is done on the basis of data sent from the machine to its digital twin.
  • Use of simulation: virtual installation space tests show how new components can be integrated. Critical components can already be subjected to thermal loads at an early stage in the development process.

Bertrandt not only has the necessary know-how, but also has extensive experience gained from successful projects to enable it to produce representations of digital value chains. This includes, for example, machine data acquisition, data storage and the analysis of large quantities of data, as well as the development of virtual models and their visualisation.

Mixed reality, which consists of a combination of virtual and augmented reality applications, is no longer a vision of the future, but instead an effective and viable technology and can be found in all of Bertrandt’s customer sectors.
Under the slogan “Creating Next Reality”, the national and international company locations will work together to offer their customers needs-oriented solutions, even for more complex applications.

More information

Over time, machines become more susceptible to malfunction and require more frequent maintenance. Integrated, automated monitoring opens up new possibilities, such as predictive maintenance. Bertrandt equips machines and plants that have no interfaces with an IoT device for machine data acquisition:    

  • The IoT device includes sensors and the transmitter unit as well as software and computing capacity for pre-processing the data.
  • Parameters such as temperature, pressure, vibration, machine characteristic values and throughput are continuously recorded.
  • The machine data are stored locally in a data storage system or transferred to the cloud.

Bertrandt develops the software for all process steps. We have carried out numerous projects in particular for the processing of the data in the cloud.

Which data are required and how can data quality be measured? How can the interrelationships within the data be recognised and what added value can be achieved by analysing them? Bertrandt advises companies on how they can use sensor data from machines to draw conclusions about the causes of faults, power optimisation or design improvements.

With the Bertrandt Industry Cloud (BIC), we offer you a basic modular system for recording and analysing machine data which can be adapted to any customer requirements. Our industry know-how and our understanding of all questions of development, design, production and support enable us to distinguish between useful and non-relevant data more reliably than big data service providers.    

Our service portfolio includes

  • Production analyses and fault visualisation
  • Utilisation of machine learning and neural networks
  • Selection, configuration and data modelling
  • Microsoft Azure Cloud Services (IOT Suite), Microsoft Gold partnership for cloud services
  • SQL, noSQL, Hadoop, Datalake technologies
  • Kafka
  • MQTT data transfer

Progress in the development of autonomous driving is directly linked with the development of virtual testing methods. Bertrandt has applied its expertise in the validation of driving functions to design and build virtual testing environments. We use virtual testing methods to test individual components, integrated systems and the interaction between a large number of complex systems during automated driving.

Our virtual testing procedures include

  • Vehicle dynamics simulators
  • Scenario-based testing
  • Hardware-in-the-loop (HiL) testing, e.g. for virtual crash tests
  • Model-in-the-loop (MiL) testing, e.g. for driving functions
  • Software-in-the-loop (SiL) testing, e.g. for conventional functions


Examples of projects for data-based services

Automated labelling of image data
The aim of this project was to train a neural network for an autonomous driving project in order to improve camera-based environmental recognition. For this purpose, more than 50 million images of road scenes had to be automatically labelled with machine-readable descriptions. We developed powerful algorithms and a toolset for the automatic labelling of image data.

Targeted data analysis for fault detection
In one case, cracks were found to reproducibly occur without apparent cause during the operation of a vehicle component. In our analysis, we correlated the vibration data of the component with the lateral and longitudinal acceleration of the vehicle. This enabled us to determine that incorrect sizing of the component thickness were the cause of the vibrations that destroyed the material.

Industry know-how and specific algorithms to provide reliable knowledge
The aim of this project was to derive information about the road surface quality from the behaviour of the vehicle’s shock absorbers. In spite of the application of machine learning and neural networks, an external specialist for data analysis was unable to deliver plausible results – due to a lack of industry knowledge and understanding of vehicle dynamics. Our data analysts were able to draw meaningful conclusions about the quality of the road surface by using mathematical-statistical methods and application-specific algorithms.

Your Contact

Michael Schneider

Key Account Manager