1. MBSE. Vision and fields of use.

The first area we think of for digitization is industry, and manufacturing in particular. This is a very good starting point, but we must open our minds to all possibilities.

Think, for example, of a smart city. Industry 4.0 uses automation, digitalization and big data to increase productivity, BUT Industry 5.0 focuses on humans. People are the main beneficiaries of digital twins and artificial intelligence, as they help us make the best decisions and create products tailored precisely to our needs. Industry 5.0 wastes no material and reuses our limited resources, while producing fully customized, people-centric products.

Model-Based Systems Engineering (MBSE) introduces virtual models into the entire product or service lifecycle; from the first line on the drawing board, to the end of its useful life and its final recyclability. These models do not simply replicate 3D designs, but also incorporate the entire functioning and behavior of complete systems and, more importantly, of interconnected systems of systems.

MBSE makes extensive use of the virtual world to enhance all aspects of product engineering.

• Conceptual design and specifications: We can generate our first ideas on a computer, and from the outset predict the performance of the item, component or assembly we are designing. We can even know if we meet specifications according to customer requirements and legislation. It’s all available from your PC or laptop.
• System engineering and structural design: It is possible to complement our design with all the accessories attached to the system. We can simulate the coordination between all hardware and software components and subsystems. We can even develop complete simulations, called digital twins, for virtual testing. This allows us to cost-effectively analyse how our systems will respond in any scenario in a risk-free virtual environment.

 

Our large-scale virtual replica will receive and provide feedback to all elements of the system through IoT arrays. 5G communication is essential to ensure full connectivity and, according to the most respected sources, 6G is just around the corner.

• Mechanics and electrics for smart cities, vehicles and factories: With MBSE we can virtually simulate a complete city, a complete manufacturing plant or a complete vehicle with all participating elements. The means we can design complete vehicles, data transmission and the general working conditions of any machine and process we can imagine.
• Design verification, validation and final deployment: In a virtual environment we can test everything to its limits and discover the maximum capabilities and performance of systems, vehicles, factories and even entire smart cities.

With manufacturing engineering in mind, the virtual MBSE approach can be applied to:

• Process engineering, assembly line and tooling simulation: all kinds of processes, lines, tools and workstations can be simulated down to the smallest detail. It is very important to note that simulators and digital twins not only obtain information from the real world, but are also able to generate their own virtual data, replicating reality.

This is essential for training the artificial intelligence at the core of a digital twin. Algorithms are specially designed to learn by themselves and can adapt and change their logic according to the information they receive.

• Supply chain management, quality and inspections are included in the Industry 5.0 concept. Think of computer vision and its ability to identify objects and especially faults and defects of any kind. Of utmost importance in key sectors such as aviation and electronics.

A digital twin is able to foresee unexpected circumstances and risks long before they happen.

It is important to note how digital twins greatly improve predictive maintenance of vehicles and assets. It is already extremely useful to be aware of the operational condition of assets online and in situ at any given time, but it is critical to know in advance how the asset will evolve over time and WHEN a potential failure may occur. Measures can be taken to prevent these failures at an early stage, saving a lot of time and money and improving user experience and customer satisfaction.

In terms of systems engineering, digital twins and simulation can be useful for all kinds of industrial processes and machinery.

They can be used to create training devices, develop system models or user applications with powerful graphics engines (such as Unity or Unreal). It is possible to integrate any solution that the customer is already using while combining real and simulated systems in a ‘hardware in the Loop’ approach. Test benches can also be included in this virtuous circle.

Continuing with the technology deployed, we highlight the ability to generate synthetic data from simulations to train artificial intelligence. Cooperative simulations are possible by integrating simulators from different platforms, gathering input from different types of simulators. This is especially important for both civil and military applications.

It is also possible to create twins using a digital Live, Virtual, Constructive (LVC) approach. LVC links real (live) entities with simulated and virtual (virtual) environments and then introduces new virtual actors to interact with them (constructive). The last step is the validation work with the client.

Figure 1: Schematics of a simulation incorporating virtual and real scenarios and variables.

2. MBSE core competences.

As a core competence, we must be able to develop any kind of software, including applications and user interfaces, databases and communication between systems.

To model any system we apply deep and machine learning, virtual models and real-time digital twins, acquiring telemetry data from machines and vehicles via IoT. This is where the ability to predict exactly what will happen in the future comes into play, establishing behavioural trends and seeding the field of predictive maintenance. We can combine 3D tools from other companies. It is possible to create accurate copies of test fields to replicate test conditions and the actual operating life of assets and machines.

It is a good idea to enhance information visualisation with augmented reality, overlaying virtual information on the real world and linking virtual reality applications to enhance training, creating an immersive experience amidst a reliable virtual reproduction of the real world.

3. MBSE Design Competencies.

At the design level, we can develop cockpits and ground control stations (GCS) with human-machine interfaces (HMI) for industry in general and vehicles in particular. This can be linked to the remote operation and automatic driving of all types of machinery, under human supervision: for example, forklift trucks in warehouses or taxi fleets in urban areas.

This contributes to the final stage of validation and verification (V&V) in unison with the end user.

4. MBSE transversal competencies.

Among the transversal skills that are useful to develop are integration skills: the ability to connect different systems from different suppliers and different platforms. This would allow us to certify any system.

If we want to achieve a complete virtual ecosystem in our industry, we must move from conceptual design to cyber security, through simulation, digital twins, IoT, predictive maintenance and final validation. If we are able to embrace all of this, either on our own or with the help of partners, we will be able to take full advantage of virtualisation.

5. Integrating virtual twins in Industry 5.0
Virtual integration requires a visual engine that can generate 3D environments and also provide the three most important capabilities for a virtual twin simulation.

• Hardware in the Loop, introducing real assets into a virtual world.
• Software in the loop, introducing any solution into the main virtual twin.
• Human in the loop, introducing humans into the virtual world.

Note that the two most powerful aspects of artificial intelligence are the ability to capture and interpret images, what we call computer vision, and the ability to help us make the best decisions. Artificial intelligence will not take our place, but it will help us in our daily work.

Figure 2: CAD design and MBSE incorporating operational behaviour are key to predictive maintenance.

 

6. The fundamentals of predictive maintenance.

Preventive maintenance includes the actions necessary to keep machines running, minimising downtime due to breakdowns, machinery breakdowns or unplanned stoppages. The main objective is to anticipate unexpected downtime and minimise downtime.

Predictive maintenance goes further. By monitoring the production line we can monitor machine data such as temperature, humidity, pressure or uptime to assess and detect problems before they occur.

The use of artificial intelligence or deep learning allows for more effective and faster predictive maintenance management.

All the data captured in real time can be used to improve predictive maintenance, allowing machines to learn and detect problems autonomously, displaying alarms, indicating when a part needs to be replaced or, in extreme cases, stopping production to extract this knowledge. and generate predictive models that detect when a defect will occur in production.

Implementing predictive maintenance requires algorithms that are continuously learning and improving the models, and this is where artificial intelligence and deep learning come into play.

7. How to implement predictive maintenance.

First, all relevant data is imported, stored and transformed into a usable format.

The second step is to explore this historical dataset, before developing the models in which the variables used to identify patterns are defined.

Once we have the model we move on to the training phase and, when it is sufficiently tested, it is deployed in the real environment. Thus, it works with new data and can perform predictive maintenance, generating the necessary early warnings of failure or decrease in the quality of the parts produced in the event that the predicted metrics far exceed the defined thresholds.

Different combinations of recurrent and convolutional neural networks can be combined to generate models that can process real-time data and provide failure predictions with high levels of accuracy.

The TC approach consists of using neural networks and artificial intelligence algorithms to establish a predictive model based on all the operational data of the asset under examination, as well as that produced by other connected elements.

 

Three types of maintenance are currently used:

• Run to Failure (RtF): maintenance is carried out after the failure occurs. It is used when equipment failure does not significantly affect operations or productivity.
• Planned preventive maintenance (PvM; Preventice Maintenance): Reviews are carried out to prevent equipment from failing unexpectedly.
• Predictive Maintenance (PdM): The performance of equipment is monitored over time through IoT systems, this data is collected over time, allowing us to determine how much life a vehicle, machine or part has left; by anticipating when the machine will fail, we avoid unnecessary maintenance or unexpected failures.

Predictive maintenance consists of four phases:

• System data collection: There are two main methods for collecting data, on-board and off-board. The on-board method relies on sensors to collect data directly from the equipment, while the off-board methods rely on large existing databases.
• Data processing: Because large volumes of data are constantly being produced, we must weed out irrelevant information and extract what is most important for our needs. This is especially relevant when working with artificial intelligence, as the data we use to train the model strongly influences its performance.
• Fault diagnosis: RUL (remaining useful life) is normally used. Once we get the outputs of our artificial intelligence, we can run them through a health indicator, which will allow us to know the state of our machine and predict its useful life.
• Maintenance strategy decision: Once we know the RUL of our machine, we can make the corresponding maintenance decisions.

 

Figure 3: Predictive maintenance can be complemented with applications using augmented reality and virtual reality.

8. Predictive maintenance applied to machines and vehicle fleets.

There are four models for implementing predictive maintenance:

• Physical model approach: a physical or mathematical model of the system can be used to assess the degradation of components. This type of model is difficult to implement in complex environments.
• Knowledge-based approach: This approach uses expert systems or fuzzy logic. Expert knowledge is required to build the system.
• Data-driven approach: This approach uses computing power and a large amount of data. There are three models: statistical, stochastic or machine learning.
• Digital twin approach: This approach relies on data and models, creating a link between the digital and physical world.

The data-driven approach is the most widespread and uses a wide variety of algorithms and applications based on deep and machine learning. Deep learning methods guarantee good performance in fault prediction, but require a large amount of data compared to traditional machine learning techniques.

The digital twin is a more modern technique, with a strong ability to identify abnormal conditions and predict the lifetime of a product, as it combines physical models (to create the digital twin) with a data-driven approach.

Our predictive maintenance approach allows us to anticipate maintenance operations and failures before they occur. In addition, we can design operations to improve any system. In these cases, CT Engineering Group will analyse all available information from any source, perform root cause analysis, conduct functional investigations, develop model hypotheses, identify relevant variants and propose validation methods. In short, we coordinate the development of an entire system. ECOMOVIL ‘23 is a good example of this, applied to an electric city bus.

We can imagine the complex connections with telemetry, transferring data to a database, calculating predictive maintenance and creating digital twins. The digital twins then interact with each other.

In this case we have chosen what we believe to be the key elements of an electric vehicle: the electric motor, the battery, the ADAS system for emergency braking, the air conditioning unit and finally the brake and tyre wear.

We first created a model of each subsystem and then linked each individual model within a single platform and finally represented the interactions between the subsystems. We all agreed that road conditions or driving style can greatly affect battery or tyre life.

This knowledge has implications for the operating costs and efficiency of a vehicle, and with the help of our ability to predict what will happen to the vehicle in the near future, we can reschedule maintenance operations, reduce downtime, optimise the availability of spare parts in a just-in-time model, thus avoiding the accumulation of parts in thousands of cubic and square metres, saving a lot of money, minimising breakdowns, as well as extending the actual life of the vehicle fleet.

Predictive maintenance lets you know in advance, based on date or mileage, when components such as the starter battery or brake pads need to be replaced, receiving a report whenever there are signs of crucial component malfunctions, with artificial intelligence models for failure detection.

Figure 4: Key elements of an electric bus to which predictive maintenance can be applied.

Remote diagnostics is like having the most advanced diagnostic equipment installed on each of your vehicles running at all times. You can view active DTCs and customised alerts on your vehicles based on their severity, with visibility into key maintenance metrics by comparing history and checking sensor behaviour in real time with messages, emails or notifications.

With automated maintenance plans, mileage and run times are automatically recorded, so you can receive alerts when vehicle repairs are needed. In addition, plans can also be automatically adjusted based on faults detected through remote diagnostics and predictive maintenance.

It is also possible to monitor and control vehicles, all in one place, in real time on the same map. Consumption and other metrics and alerts can be anticipated based on fleet configurations.

Technologies are available that can establish a scoring system to ensure that drivers make fleets safer and more efficient (possible through a driver identification plug-in). Complete visibility of fleet performance can be obtained for any type of operation.

Even specific training needs can be identified by monitoring driver performance, to help minimise wear and tear on vehicle components.

9. MBSE for Smart Cities.
Moving on, I would like to mention how important digitalisation will be for a smart city. Here are some examples of key projects CT is developing, using digital twins and appropriate simulations.

• Road inspection and management. We can inspect roads by taking advantage of all the sensors integrated in vehicles (radars, lidar, cameras…). They can be used to see the condition of signs, horizontal markings and even the quality and degradation of the pavement. In this context, a digital twin can detect and prevent the evolution of cracks, potholes and ditches in the asphalt to prevent accidents, while providing a complete and updated model of our environment in real time. In other words, we can have incredibly accurate models of our streets and their actual state.

All the captured information can be analysed with big data and artificial intelligence techniques and then displayed to the end user on portable devices, such as PCs, tablets or mobile phones.

We will be able to see on our screen what the next street will be using the sensors installed in our vehicles. Imagine the possibilities that arise from thousands of cars providing pristine information about what is going on around them. These anonymous vehicles could provide an enormous amount of information that could be analysed to produce a gigantic dataset to train predictive models that would improve, in the short, medium and long term, the safety and efficiency of our cities.

Figure 5: Road detection is well suited to using vehicles to capture information.

• Self-calibration of ADAS sensors thanks to IoT. It is also feasible to acquire information from other vehicles, which would make it unnecessary to send vehicles to a workshop to recalibrate the sensors with specific tools.

With this technology, vehicles can communicate with other vehicles and the surrounding infrastructure to receive information and compare it with the information they themselves are collecting. The data is acquired from the telemetry of the vehicle and connected via V2X and 5G protocols with other vehicles and infrastructure elements (such as traffic lights, cameras on gantries, etc…).

If the data captured by our vehicle matches the information transmitted by other vehicles and infrastructure, the probability that our sensors are well calibrated is very high. If there is a match, our ADAS calibration is fine. Otherwise, the information provided by surrounding assets can be used as a standard to recalibrate our sensors online in a matter of minutes. Online and on-site.

Sensor accuracy is key. Static calibration can cause up to 2 hours of downtime. With our technology, sensor calibration takes a maximum of 5 minutes. In addition, we perform a continuous dynamic calibration by comparison with other road users. In this particular case, we focus our attention on the ADAS system that assists in emergency braking. We can see the relationship and connection between information from other vehicles and infrastructure and our own vehicle safety suite. A clear example of collaboration between intelligent systems talking to each other.

Figure 6: Self-calibration of sensors based on data received from infrastructure and other users.

• It is now possible to drive vehicles remotely in urban areas and open roads using IoT connections and the power of simulators and digital twins to replicate real-time environments with extremely low latency. How much will businesses grow if they can safely manage a fleet of taxis, vans or small fleets remotely? Imagine driving a real vehicle in the middle of a real city or highway, through real traffic consisting of other cars, cyclists, skateboarders, pedestrians, etc…, all from the comfort of your home or office.

As a disruptive technology, we are developing communications between sensorised vehicles (ADAS suit and control actuators), remote driving stations and control centres using MQTT and UDP protocols. We are also developing simulators for real vehicles.

It is important to highlight the fact that we need to develop a robust system, resilient to general communication network failures and able to operate smoothly even in poor communication conditions with ultra-low latency in information transmission, especially for video signals and commands to control the vehicle.

Note that a latency in the order of 50 milliseconds at a speed of 120 kilometres per hour would mean that the vehicle runs uncontrolled for four to five seconds.

The main advantage of this technology would be seen in fleets of taxis, vans and small trucks. Obviously, a single person can NOT manage more than one vehicle at the same time, but one person can seamlessly switch from one vehicle to another in a matter of seconds. This will increase the operational efficiency of the vehicles and reduce, especially in freight transport, the number of journeys where the truck is completely empty. In short, remotely controlled vehicles bring us to the threshold of fully autonomous driving and are an excellent approach to managing traffic in a smart city.

We will develop two types of simulators: a basic one that provides an initial integration of the different systems involved in vehicle operation and a comprehensive one that provides a fully interconnected and highly realistic simulation.

• Smart city traffic manager. Vehicle-to-vehicle (V2V) and vehicle-to-everything else (V2X) IoT communication will be essential for sharing billions of bytes of information in the cloud so that we can receive information about what is happening in front of us and around the corner. Vehicles can transmit and receive data that will warn of any dangers or help calculate the best routes from point A to point B.

Real-time digital maps built from data provided by vehicles and UAVs can help improve the planning of the smart cities and factories of the future. It bears repeating how revolutionary it will be to combine real-world information with virtual data generated by our digital twins. Ultimately, they will help to better train our virtual models.

We also participate in the European ECOMOBILITY KDT JU project in which we are developing a traffic manager for a smart city. Three-dimensional maps of the environment are generated from data captured by vehicles through their ADAS systems and infrastructure data, through V2V and V2X communications adapted to 5G protocols.

We are developing predictive models to streamline traffic in a large city and avoid incidents, traffic jams and accidents, making journeys faster and more agile, significantly improving the user experience and contributing to greater eco-sustainability.

We take advantage of vehicle sensors to capture data and images of our own environment to create an exact virtual replica. The next step is to generate a mapped virtual environment using graphics engines such as Unreal or Unity to simulate our smart city.

And then we complete the simulation with a fleet of virtual vehicles. Our goal is to give the vehicles the ability to autonomously decide how best to get from origin to destination based on incidents and traffic. The roads are not predetermined. It is about deploying intelligent models capable of deciding in situ and online the best routes, the best average speeds to avoid red lights, the best parking spaces, etc… All this is going to be achieved with extensive use of communications between ground vehicles and aerial platforms, what we call UAVs or drones. In this way, smart cars could rule the smart city.

Since a key aspect of this project is the communication between assets and vehicles, we also need to secure these communications. Messages sent and received by travellers are not the same as communications between military or police vehicles. We all agree that strategic vehicles, such as police or military, need much stronger cyber security and encryption than those used for civilian applications. We are also implementing this value-added layer of protection, backed by blockchain for seamless traceability of communications and links.

So the perfect combination for a smart city is:

• 5G communications
• V2V and V2X connectivity
• Artificial intelligence
• Predictive models based on machine learning and deep learning
• Cybersecurity
• Blockchain

Figure 7: V2V and V2X communications to manage smart city traffic.

10. Benefits of Simulators, Virtual Twins and MBSE.

To conclude, the main benefits of virtual twins and MBSE are quality, security, speed and reliability.

We can improve the quality of product and process design, as it is possible to immediately correct initial errors. This is really important as we can bring a product to market or start up a machine much faster.

• Humans and machines can be introduced into the loop to evaluate all scenarios. With this approach we give the simulators all the necessary functions to work in unison, rather than separately, which would be less useful.
• All simulated assets are safer. We can detect early on any design and day-to-day failures.

Thousands of simulations can be run without the need for large investments. So the advantage of using simulators over real tests, based on errors and redesigns, is more than obvious.

This is the approach we have adopted at CT Engineering Group, making extensive use of MBSE strategies to deliver our projects faster and more reliably.

Figure 8: MBSE virtualisation can be applied to any Industry 5.0 and smart city project.