In the industrial world, when it comes to logistics 4.0, innovation is often the result of contamination: there is no doubt that each project has unique starting and finishing points, but often it is only the intersection between sectors and disciplines that multiplies each result, applying to many different industries.
This is basically what Aramix faced with freight trains, developing a more efficient system that was able to weigh them in motion, collecting the physical data in the field and integrating them with the know-how of the experts.
The result of our Use Case is an innovative method that provides realistic cost ranges, based on the actual weight of the loaded train, but above all it is a useful approach for any company that needs to understand a phenomenon by relating labeled data and unlabeled data of a system, coming from different sources and in different “formats”.
For example, we have applied the same method to the Risk Prediction of failures and malfunctions of helicopter rotation systems, based on vibration, noise and the overheating data of a single component.
And again, this was the same approach when we developed a Decision Support System for the supply of machinery components in the heavy industry, based on wear data and the replacement of macro-elements of the systems, such as metal drills, augers, turbines etc.
Cost prediction based on mixed data: weighing a train is above all a question of method
If you are not an expert you could be wondering: who cares about weighing a freight train?
You would be surprised by how many stakeholders are actually involved.
- those who manage the infrastructure, i.e. the tracks, to know how much they will be worn out
- those who take care of the hardware, i.e. the trains themselves, to organize transport
- those who rents the convoys to transport their goods by rail, to predict the costs
- citizens, because overweight trains compromise the safety of the infrastructure itself.
The cost of transport is determined with a series of different variables.
The main variables that define the cost of rail transport are the time,
the traffic on the route, the duration of the journey and the occupation of the route (how busy it is at a given moment), total weight of the wagon plus the goods it carries.
Regarding this last point, why does the weight of the train affect it?
Because the greater the weight of the wagon, the greater the wear on the rails and the energy required to move it.
Beside all that, safety is also a main issue, because railway companies need accurate load estimates to control overloads, identify potentially unsafe carriages and optimize maintenance planning.
Until now the traditional method widely used was the static one; basically the train was positioned on a special “intelligent” track.
This track has optimal slope and curvature conditions, and is equipped with sensors capable of recording the weight of the convoy. On the other hand, this procedure turns out to be long, expensive and complex. It is performed only periodically and the recorded evidence is used to make estimates on subsequent loads.
Innovating the weighting method for a more efficient cost prediction
However, a rough estimate is not enough when you pursue efficiency in logistics 4.0, especially in terms of energy consumption.
The only alternative is prefigured as a DWIM system: Dynamic-Weighing-In-Motion.
Unlike static weighing systems, DWIM systems in fact clearly improve circulation times and optimization of the commercial activity.
One of the major players in the world of rail transport commissioned us a preliminary study on the development of simple and inexpensive DWIM systems, based on easy-to-install strain sensors, to be applied practically anywhere.
We have developed an experimental system on the railway infrastructure near Milan, consisting of sensors installed on a 10m stretch of straight track, with standard and constant gauge.
The sensors are placed where the installation and de-installation processes are quick and easy, facing each other, on the right and left rails of the track.
They are fiber optic and acquire data at an impressive speed. The acquisition system identifies the train passing on the track and cuts the corresponding signal with a time margin of 10s before and after the passage. Furthermore, the sensors are able to detect different types of “collateral” data, including the temperature and the deformation of the rails, which depend on various local factors ( the conditions of the ballast and the railway ties, etc…)
The output: a replicable and re-applicable system
The core of the methodology was combining labeled and unlabelled data.
Although axle weights were rarely available, because it is difficult to arrange the passage of a train with a known weight on the line, we were able to calculate the weight of hundreds of trains passing on the dynamic ladder each day.
We have certified this DWIM system both for axles and for bogies and trains: a service that can be easily replicated and resold even by infrastructure builders.
Actually, the greatest result is a promising approach that paves the way for future investigations, for a considerable saving of time and resources applied to new industries.