I'd like to discuss the question of the Role of Digital Shadows and Digital Triggers in the Digital Twin Concept. I think both terms play an important rule but they are not exactly defined til now.
You can see my point of view about in an attached poster for an order-oriented Digital Control Twin.
The Digital Shadow (DS) is only a Snapshot of a real object at a certain time, a certain location and for a certain 'cutout' that represents the actual state but not the entire object itself. To interpretate the DS it is very important therefor to define the exact time, location and cutout of the DS. and to store this data.
The Digital Trigger (DT) is only an impuls of the virtual twin at a certain time, location etc., to 'influence' the real twin but it is not the entire virtual twin itself.
The virtual and real twin are connected by a certain kind of regulator to detect and evaluate deviations of DS and DT and to harmonize them. The kind of regulation can be very different and depends on the object itself, the environment and the purpose.
These digital shadow and digital trigger concepts seem equivalent to the data acquisition and control aspects of SCADA systems.
Re: interpretation of the digital snapshot:
Most SCADA systems have mimic diagram displays = a simple sort of digital twin. More complex control centers such as those for power systems have more complex twins. State estimation applications are used to combine measurement data with the digital twin (=model) of the power system in order to provide a better estimate of the overall state including values that are not measured.
This can then be presented to human operators for "situational awareness", or as the input to applications that try to regulate or optimize the real system.
Thanx for answering, is the conclusion, that DS and DT should work (only/mainly/best/ ...) with SCADA-Systems resp. on SCADA-Level ?
Hello Wilmjakob,
I personally think the definition of a DS that you give here is valid, however I believe we can go further.
A recent paper "Digital Twin: Enabling Technologies, Challenges and Open Research" has used the following definition "A digital shadow is a digital representation of an object that has a one-way flow between the physical and digital object. A change in the state of the physical object leads to a change in the digital object and not vice versus."
With that said, I would think that a comprehensive collection of Digital Shadows could enable the creation of a black/grey box model or a probabilistic representation that could be used to validate a Digital Twin.
Meaning that once enough data is collected of how the physical model affects the digital model, the reverse operation can be done and in this sense a digital twin can be established.
From my (debatable) point of view, the difficulty is that the term DT originally had two meanings. In a narrow sense, this is a special type of model that is always in accordance with its original (the synchronization speed depends on the problem being solved and can vary in a wide range). In a broad sense, this is a system that includes the original, the digital model itself, the connections between them, and, perhaps, also data and services. Wilmjakob Herlyn uses the term "DT Concept" for this system, although it is not very clear how a Concept can consist of physical and information objects. I would prefer to call it a cyber-physical system with a model in the control loop. This is a special kind of CPS: its model is always in a "hot " mode relative to the object, but the requirements of a rigid (almost automatic) connection of the physical and digital parts can be weakened in comparison with the initial understanding of CPS.
As for the scheme presented in the question: it is unclear what the term "digital shadow" refers to. There have been attempts to say this about DT in a broad sense, but the shadow metaphor itself is not very successful, in my opinion. The shadow always simplifies the object, it is passive (how to imagine, for example, a predictive shadow?), the connotation of a shadow is rather negative. Communication from the original to the DT (as in any cybernetic scheme) is the transfer of data about the current state of the controlled object.
The feedback called a "digital trigger" in the scheme seems to be some simplification. Feedback doesn't always work as a trigger. In the recognition of failures (diagnostic DTs), this is correct: a critical mismatch of the original and the DT generates a signal about a potential failure. But in operational/ optimization/predictive tasks, DT is used to evaluate control variants and choose the best one based on the task - where is the trigger here?
Hy Victor, thanks for your detailed argument, I think the trigger is the result of one or more simulation runs, e.g. in the automotive industry there are a lot of simulations for the stiffness of the chassis and strength of individual parts or for the absorption forces of bumpers etc.. before real crashes are performed. In this case, the results of simulation can also be named as trigger because it is used to change the design of the chassis or a certain part etc. Can you agree with this?
Hi Wilmjakob, I think your interpretation has the right to life – in the end, any digital control can be represented as a trigger. To be precise, I have not met the term “digital trigger” in this context before – it seems this is your term.
If I understand correctly, your goal is to give a general cybernetic scheme of DT in a broad sense (Concept of DT). I agree that the regulator (as in any cybernetic scheme) is determined by the features of the control task. But in the classical control theory, a model (in our case, DT in a narrow sense) is a part of the controller (“a model in the control loop”). The model (namely its structure, detail, accuracy) is also determined by the characteristics of the control task (that is, it is not a comprehensive model of the "whole" original, which is both difficult and unnecessary to build). At the same time in your scheme, DT exists independently, and a regulator is located between it and the original.
I understand and welcome the desire to separate DT. Psychologically, it seems that DT is more than just a model: it is a model that is permanently synchronized with the original, often (although not necessarily) in the form of a machine learning model (which was not previously in classical control schemes); DT may differ from the original (it is also an extension of the classical understanding).
In a word, your attempt (and the whole discussion) it seems to me very timely and productive, but it requires further research.