Industrial Internet: Towards Interoperability and Composability
Key to making this happen is to change from our top-down engineering approach to a bottom up, ad hoc construction of systems from parts, orchestrated to solve a particular problem at hand. This will require a description language for each software or hardware component that allows such composition. That language must include descriptions of how each part is expected to behave in various circumstances. Since we cannot know, a priori, how software will behave in all circumstances( except under very limited conditions) it is important that this task be augmented with machine learning and, leveraging the digital twin approach, can perform experiments in a simulated environment that can be validated in the real-world if the results appear promising. Not to understate the complexity of producing such a simulation, however, is it really any different than that required for an excellent digital twin? The complexity of a simulation depends on the questions we are trying to answer with it. If we build a simulation to understand how gears mesh, we may have left out the parts that would let us answer questions about gear wear. If we allow gear wear, we may still not understand how shavings from wear will interfere with the bearings, etc. As we approach perfection, we usually find it is simpler to just build the thingamabob and measure it! But nothing prevents us from doing exactly that – having physical experimental setups that allow such experimentation by the system prior to operational deployment.
Because a common operating picture 39 isn’ t needed – we don’ t require a single standard to which all software must be developed for interoperation because we can generate appropriate translation and adaptation software to insert between interoperable but not immediately composable pieces – we lessen the longer term disadvantages of being locked into a single model we have to select prior to broad application. The system should satisfy and adapt to circumstances similar to an insect colony or even a( non-congressional) committee.
6. CONCLUSION
In our‘ analytics in the cloud’ approach, we need a standard way to represent information about common processes. However, such approaches in the past( e. g., IEEE SUMO 40, Cyc 41, Ontolingua 42) have generally only succeeded at the syntactic level. We think there is a common way to reason about anything, but the reality is that this is a trick of our own brain – we do not really reason using common methods, but have an ad hoc collection of hacks that work well in
39 https:// en. wikipedia. org / wiki / Common _ operational _ picture
40 Pease, Adam; Niles, Ian; Li, John;“ The Suggested Upper Merged Ontology: A Large Ontology for the Semantic
Web and its Applications,” AAAI Technical report WS-02-11, 2002. https:// www. aaai. org / Papers / Workshops / 2002 / WS-02-11 / WS02-11-011. pdf
41 http:// opencyc. org
42 http:// www. csee. umbc. edu / csee / research / kse / ontology /
IIC Journal of Innovation- 73-