SimxonCARE - a rational approach to maintenance optimization

If maintenance costs affect your bottom line, SimxonCARE gives it a direct hit

Contents:

After the publication of this web page, I have written two additional articles with background material.

A small FAQ to begin with

  1. What is SimxonCARE about? Prediction of component failure based on data and general understanding acquired through a structured dialog with the customer. The toolbox applied for that purpose spans the range from machine learning to rigorous mechanical simulations of all kinds.
  2. What is in it for me? Substantial savings in maintenance costs due to
  3. Where can I find material on your approach and its theoretical basis? Here, here and here.
  4. What is your delivery? A dynamic model.
  5. Do the pictures to the right illustrate customer stories of yours? No.

Shit happens. The only question is: when?

"Maintenance" is not the same thing as "repair". You repair something which has broken, that is, the thing suddenly went from a good running order condition to something less satisfactory. When that happens, the failure will have consequences of some kind.

Maintenance is what you do in order to avoid such consequences by taking action in due time. Only in fiction one can expect shit to happen at a time known in advance. In the real world, statistical modeling is the way to improve any situation at hand (or out of it).

SimxonCARE's delivery: a dynamic model

(See the web page "Some of the math" for a description of what SimxonCARE does up to the time of delivery)

We estimate that not every CEO will be able to appreciate the value of "a dynamic model". According to the experience of Gmech Computing, that problem is easily solved, however: protected by a Non-disclosure Agreement, we shall simply demonstrate some user scenarios with the models that we already have in our portfolio.

To the more technically minded: SimxonCARE's key delivery will be a "function with some static data in it". Select the programming language, pay us and you can have the model in any format you want. Our main IP lies in the processing that we do in order to obtain the model. When you have paid for the model according to contract, it is yours to keep (but not yours to disclose).

An abundance of possible spin-offs

When we have delivered your dynamic model to you, you can start to optimize.

If you own one hundred wind turbines, you can implement an algorithm running in the control room which will maximize the probability that all units reach their end-of-life simultaneously.

If you own one hundred train units, you can arrange things to maximize the probability that the number of functional units available during rush hours is the number that you need. If you are subject to restrictions concerning costs, substitute in the previous sentence the word "maximize" with "control" or "optimize".

If you own N instances of any other kind of device, you'll have to figure out yourself how to leverage a SimxonCARE dynamic model to suit your needs. Or to pay us for consultancy services...

Don't take our word for it: Before predicting your future, allow us to predict your past

If you own N instances of any kind of device, you and SimxonCARE can join forces to assess the applicability of our offering to your situation. The test may be as simple as this:

  • Randomly select N/2 of your N instances.
  • Negotiate with SimxonCARE the format of history data needed to establish the "network of causality" = decide upon a set of data which is assumed to be complete both when it comes to causes and when it comes to effects.
  • Gather history data for all N instances of the device.
  • Let SimxonCARE generate a model from the randomly selected subset only.
  • Test the model on the other N/2 instances.

The above validation procedure is just one out of many methods how to test a SimxonCARE model against its own data.

SimxonCARE is not an off-the-shelf product and never will be. Gmech Computing and Simxon will apply the SimxonCARE toolbox differently in each case. For those reasons, a test like the above will require a down payment. On the pro side, you will then have a proof-of-concept which may be unbeatable in your organizational context, too.

We have done stuff like this before with good results. With a suitable NDA, we will be able to tell you about it. We can also hand you a white paper.

We face competition with confidence

A companion web page and two articles document that "system identification" is an advanced discipline its own right. Neither Gmech Computing nor Simxon master all available tools, but we honestly believe that nobody else does. Big corporations may state that they do but - like us - they will have to prove that they master the tools which the customer needs. Ask any competitor if they have a consistent and transparent validation offer like we do (see the paragraph above).

We also believe that introducing the "network of causality" is a substantial innovation of ours.

On request, Gmech Computing and Simxon shall face competition with confidence. We are not the biggest player in the field but

  • we are agile,
  • we are competent, and
  • we are ready to collaborate with any person or organization designated by the customer.

Creative Commons License
Website for the company Simxon by Kim Ravn-Jensen is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Permissions beyond the scope of this license may be available at info@simxon.com.