Print this page
Thursday, March 02, 2017

Semantic Web Technology Could Help Feed Your Pet on Time and Much More

Written by 

Hungry Dog“No Treat for You: Pets Miss Meals after Auto-Feeding App PetNet Glitches” is the title of an article about an IoT connected pet-feeder, SmartFeeder by PetNet (www.petnet.io), that left some pets hungry due to a 10-hour server outage.[1] Another article[2] pronounced the PetNet SmartFeeder a “poor failsafe design that didn't take into account the most obvious failure mode, i.e.: no connection” and further speculated that the “reliability and design was tossed to the wind to focus on how wonderful a cloud connected technology solution to the mundane task of feeding your pet could be.” It appears, the third-party server employed by PetNet did not feature redundancy backups, rendering the SmartFeeder useless over the 10-hour period of the server outage.

The level of engineering complexity in the SmartFeeder is nowhere near what can be expected in systems-of-systems such as connected intelligent farms, mines, factories, health care systems, or transportation systems. In connected intelligent systems-of-systems, it would be much easier to miss failure modes due to low engineering intuition in the face of growing complexity of functional architectures, functional loops, multiple disciplines, and high software content. As a result, the complex systems-of-systems of tomorrow are much more likely to trigger product recalls and launch delays than the products of today. A robust and systematic way of mitigating the risks posed by low engineering intuition must be developed to make the connected intelligent products dependable.

In the CIMdata webinar entitled “Why Connected Intelligent Products Need Semantic Web Technology”[3] given on February 9, 2017, the potential for Semantic Web Technology to help avoid product failures, especially repeat failures, was discussed as a way of alleviating the problem of low engineering intuition caused by growing product complexity. Semantic Web Technology is seen as a possible way of capturing product failure knowledge by converting the implicit subject matter expert understanding into machine-readable knowledge, which can be reused through query and inference.

On February 9, 2017, some of the questions asked by the CIMdata webinar attendees could not be answered due to time constraints. Those questions are addressed in below. Hopefully, the questions and the responses will generate further discussion to enrich the understating of the topic.

For those unable to attend the webinar a replay is available here.

Let me know your thoughts.

Venki

Now to the Q&A

Question. 1 - How to integrate ontologies with engineering tools like CAD, Simulation, FMEA etc.?

In the cases mentioned in the question, the purpose of ontologies could be to capture and reuse the knowledge related to producing CAD files, input files for running simulations, or prepopulated FMEA tables. They could also be for post-processing of simulation results. Depending upon the existing expertize in a company, domain specific ontologies should be developed and instantiated using the existing engineering artifacts like CAD files, simulation input files, FMEA tables, etc. When a new engineering artifact needs to be produced, a variation of an existing instantiation of an ontology can be used to deliver the new artifact without missing existing understanding.

This process of using an ontology to capture and reuse existing knowledge may need two APIs, one for instantiation of the ontology and the other for generating the new engineering artifact. For example, in case of capturing and using CAD knowledge, one may want to use ontologies developed using OWL and instantiate them using CAD information from IGES, STEP, DXF, or BREP files. The new CAD file to be generated for a new product variant could then be produced from the ontology as one of the above file types, depending upon which standard the ontology was based on.

Question. 2 - Ontologies will change. How can these changes be managed during on-going product development?

Ontologies may change as the knowledge or capability in a company in any domain, e.g., product development, changes. Ontologies should be thought of as the IP of businesses and consequently they must be properly governed, updated, and secured. On the other hand, what is relevant knowledge and what is not directly relevant must also be decided carefully based on the company’s strategy and competitive positioning.

The main focus of CIMdata in ontology is to deal with the issue of changing knowledge during product development. This is of direct relevance to the development of complex connected intelligent products, whose failure modes are not known a priori and must be learned during the verification and validation cycles or in agile iterations. The subject matter experts (and perhaps ontology experts supporting them) in-charge of maintaining and updating the ontologies should constantly be in the design iteration loops, so that any performance issue that surfaces during simulation, laboratory testing, manufacturing, or field trials can be communicated for updating the ontologies. By employing higher level ontologies, this process may be automated so long as the change in the domain understanding is not drastic.

Question. 3 - Would an ontology based FMEA system require "governance?" If so, how is this governance done?

Yes, all ontologies need governance because they are meant to capture knowledge in any given domain and consequently represent the IP of the business. FMEA for example would need product or process experts in the business to agree on what the company’s standards should be and how they should be changed based on advances in technology and needs of competition. Consequently, maintaining and updating FMEA ontologies needs to have oversight by subject matter experts. Also, layers of ontologies would be needed going hierarchically down from the high-level product lifecycle management to specific engineering areas such as FMEAs. The higher-level ontologies would possibly constrain the lower level ontologies such as for FMEAs in terms of their ability to be changed and updated.

In summary, one can say that FMEA ontologies do need governance and the flexibility or rigidity of the governance depends upon where in the hierarchy of ontologies the ones for FMEAs are located.

Question. 4 - How does AI (capabilities like Watson) play a role in this domain?

Watson is an open domain question answering (QA) system that applies advanced natural language processing (NLP), information retrieval, knowledge representation, automated reasoning, and machine learning technologies.[4]

Open-domain QA deals with questions about almost anything, and mostly relies on general ontologies. On the other hand, closed-domain QA deals with specific domains e.g., automotive, aerospace, medicine, etc., and is easier to achieve because the NLP can operate on domain-specific knowledge where a limited type of questions is accepted.



[1] https://www.theguardian.com/technology/2016/jul/27/petnet-auto-feeder-glitch-google.
[2] http://www.freaklabs.org/index.php/home/reliability-in-the-iot-2.html.
[3] http://www.cimdata.com/en/education/educational-webinars/webinar-why-connected-intelligent-products-need-semantic-web-technology.
[4] https://en.wikipedia.org/wiki/Watson_(computer).
Venki Agaram

Email This email address is being protected from spambots. You need JavaScript enabled to view it.

Latest from Venki Agaram