Thursday@KEMA

We prepared our presentation this Thursday.

We set up a small network with a loop and a radial structure as a numerical example to demonstrate our linear programming approach for DPG optimization problem. The algorithm works surprisingly fast in Matlab which gives us a lot of confidence in solving a more complex network problem.

Around 6, the nice smell of Pizza in the hall attracts our attention for a while. After a quick "avondeten", we begin to prepare our presentation. By 8, we've already got 16 slides. We decide to finalize our slides tomorrow morning.

Tomorrow is the show time!

Thursday@Marin: day of preparations

Arriving bright and early we went to work on our To Do list.


Synchronizing the optimization code with the visualisation code. Deciding on what to do when the seas are too rough, so that the capacity of the thrusters is exceeded. And then also that most pressing of studygroup questions: who is willing to give the presentation on Friday? And how to keep the attention of the audience past the first 5 minutes? Well, we have got a great movie to show you and we will reveal how much money can be saved by replacing linear theory by nonlinear theory.

Wednesday@KEMA

We try to solve the problem stochasticly with stationary distributed patterns this morning. We take the expectation of the energy production as our objective function, and reformulate our constraints. Rather than imposing upper bounds for the absolute value of the currents, we require the probability that the absolute value of the currents exceed their upper bounds to be below a certain level. However, we realize that constructing the probability distributions for the currents from the power production and consumption patterns will be difficult, partly because of the correlation of power production at different locations. Another difficulty is that it is questionable that the production and consumption patterns are close to some known distributions (such as Gaussian). Monte-Carlo simulation may provide some useful information, but it could be quite time-consuming.

In the afternoon, we meet Mr. Bloemhof and discuss with him about some of our results and our doubts. We realize that power production can not go beyond the capacity because of the safety issues, so the distributions will not have long tails.

After one day inspirational work, we go to the "Polder" and have some drinks together. The dinner later on are quite good.

Wednesday@Marin: day of visualisation

This morning half of our group had mysteriously disappeared. Or were they hiding in the library?

On the other hand, Ed from Marin unexpectedly showed up in the morning and we had a lot to discuss. As it turned out, yesterday's code encountered some problems (under more severe weather conditions), so we had hit on a nice new problem: what strategic decision to take when the required constraints cannot possibly be met? Another major point was visualisation. Ed was very specific about what he wanted to see, so one of us spent the entire afternoon on implementing Ed's vision. Everybody will be able to see the results on Friday. In the mean time, Lagrange's multiplier method was revived as an alternative to the black-box and houtje-touwtje approaches. Still lots of work to be done tomorrow!

Phytocare on Wednesday

Not much to write... Each group pursued their paths: more literature search, more modelling, some numerics and some analysis. We had some intermediate exchange about the state, news and how to continue tomorrow.

More SWI2010 Photos on a Picasa Web Album

Dear Participants and other Blog-Visitors,

Please note that more photos of SWI2010 can be found at (and will be put on) Picasa:
http://picasaweb.google.com/SWI2010CWI

The SWI-2010 organizing committee
Arnoud den Boer, Joost Bosman, Niek Bouman, Jason Frank, Rob van der Mei, Chrétien Verhoef

A note to the participants: You can upload your own photos to the SWI2010 Picasa Web Album, by logging in to Picasa with the same credentials as for the weblog.

Dynamic behavior of meat mass flow (27 Jan 2010)

We argued about the previous model obtained by group. The model is now better. Furthermore, we have obtained a dynamical model of meat mass flow for the unstable regime. As next step, we are trying to calculate an approximate solution of the obtained system of two coupled time-dependent nonlinear PDEs.

Tuesday@KEMA

Today we formulate the problem with deterministic but time dependent patterns as a linear programming problem for a selected time scale. For simplifications, we assume: the size of DPG is continuous; there is no energy losses; issues of reliability, dynamic stability and power quality is left out.

We consider the relationship between net generation of DPG at each node and currents between nodes a linear mapping, which can be found by exiting loadflow analysis techniques. Since the net generation of DPG is a function of sizes (which can be 0 if no DPG) for different type DPGs and consumption, we can express both the objective function and the constraints in terms of sizes, which are linear. Therefore, we can solve this optimization problem as a LP.

We decide to solve the problem using the stochastic approach tomorrow.

Phytocare on Tuesday

First thing in the morning was a briefing where each of the four groups presented their results. Then group work continued until we talked with Dick, the company representative to whom the groups reported and asked questions. The groups are roughly working on:

1) Ad hoc ODE model with some numbers from reality,

2) Diffusion-advection PDE model partly ad hoc, partly kinetics from literature

3) Literature search for data giving form of reaction rates in terms of relevant quantities.

4) Combining ODE models from literature and steady state analysis.

In the discussion with Dick it turns out that he wants also to have recepies how to fit the parameters/rates to data once they start doing the experiments. Let's see where it all takes us!

Tuesday@Marin: day of coding

Several more people joined our group overnight. Fresh blood brought novel ideas. We had a long initial discussion about what we could do. At long last we decided to write down the equations.


Then some of us almost started coding, but a faulty overhead projector turned attention away. Someone else then started writing a matlab code just before lunch. After lunch Piet showed up and explained a very nice idea how to reduce the number of constraints. He also proposed to start with a simpler problem and subsequently correct for the difference between the hard and the easy problem. By this time the matlab code was producing output. Great joy overwhelmed us and some people were temporarily deflated because it all seemed "too easy". Then a couple of us decided to go sightseeing in Amsterdam. The other ones finished the day by writing down a list of things we should do tomorrow. Among other things, we need to deal with the nasty issue of not choosing all physical quantities to be unity.

Dynamic behavior of meat mass flow (26 Jan 2010)

We have found a velocity profile for steady pressure driven chicken flow in straight cylindrical pipes. Next, we will consider the dynamics of compressible chicken flow in straight cylindrical pipe. The focus will be placed on effects caused by compressibility and viscoelasticity.

ESA@day1

Develop a Smart Positioning Algorithms for Indoor Satellite Navigation





ESA

The European Space Agency (ESA) is responsible for the construction of the Galileo system, an augmentation-system for GPS. Satellite Navigation. Satellite navigation has many applications, including route guidance for cars, the navigation of boats and airplanes, and the synchronization of telecom networks. It is has one serious weakness though, it cannot be used inside buildings because interiorly the signal is just too weak. The goal of the workshop project is to extend navigation to the interiors of buildings by using the knowledge of the position of users with respect to each other.

Workshop

Today all problems of the workshop were presented in the plenary session. The whole group also had the chance to get to know each other over lunch. Before lunch, Jaron Samson presented the problem of ESA: navigation inside buildings and how GPS and the European Galileo satellite system can still help with that. He explained his approach of letting users determine their relative distances to a few neighbors and fixed wireless routers. Given the absolute position of a few users outside the building this then determines the absolute position of each user in the network.

After Lunch we had second round of presentations, and people selected their favourite problem. The afternoon ended with some further explanation by Jaron. He once again explained the model. Abstractly the model is about

1. GPS devices which can

a. determine their absolute position with high precision in the reference frame of the earth, corresponding to users on parking lots and near windows.

b. Exchange information

2. Devices which can

a. determine the relative distances with each other and GPS devices, with typically rather lower precission

b. exchange some information.

3. Devices of type 2 (base stations) which are fixed with respect to the building.




This quickly led to some discussion on the technical feasibility of the abstracted version of the in building navigation problem, because in reality mobile devices, other than gps, are not actually very well suited to determine distances. We agreed however, (after some discussions which let Jaron to get some suspiciously foamy and yellowish glass of water to sooth his throat J) that this is a matter of technical progress. Indeed pseudolights with groundbased GPS transmitters are used in air and normal ports. It also involves breaking the chicken and egg problem, that there are no such devices and therefore there is no supporting infrastructure like transmitters with a known position and standard methods for getting a map of the building. In any case these problems are outside the skope of the workshop, and as an abstract model this works fine.

The algortithmic problem is now
Given an existing a seed network of size N (which could be just one device, GPS or otherwise)

• Find a probability distribution on (R^2)^N which suitably represents the probability distribution of the position of all the users. If we do not have sufficient information to determine absolute distributions this might well be invariant under a symmetry group of Euclidean motions and therefore NOT localized around a single point (Having one GPS device in the network breaks the symmetry to Orthogonal group O(2), tor the Z/2Z reflections in the line through the devices, no GPS devices mean that the symmetry group is non compact which is not nice, so it is probably best to assume at least one GPS node).

• Given probability distributions of the distances of a new node to M neighbors
  
• Broadcast information of this new node distribution to M’ neighbors.
• Let those M’ neighbors update their probability distributions
• Broadcast again…
• repeat
• Determine convergence, and if possible prove optimality of convergence rates

Some photos from Tuesday

On our way through the snow to Cafe-Restaurant Polder

Lunch in Polder

Coffee break in CWI's new wing...

Dynamic behaviour of meat mass flow (25 Jan 2010)

We discussed today the relevant processes that happen during the heat flow of chicken meat inside the portioning machine.
At the moment, we have focused on the description of simple geometry case without any restriction in model. Next to it, we are going to take flow of the meat in the model.

Monday@KEMA

We are a group of five, and we find the problem provided by KEMA very interesting and challenging. We need to maximize the total energy generated by Decentralized Power Generation (DPG) such that the maximum loads on the lines and components will not be exceeded, and find out which type, what size of DPG at which location should be inserted into the existing network. The Specialist in DPG from KEMA, Mr. Bloemhof, gave us a thorough introduction of the complex electric power system today, and provided some existing computational tools for the problem. We think that we can either solve this problem assuming deterministic but time dependent consumption patterns and production patterns, or solve it stochasticly with stationary distributed patterns. We are very excited to try out different approaches, and hope our results will provide some insights in the development of DPG.

Monday@Marin: day of optimism

We are a nice group of seven from four different countries. The participants form a very international bunch this year! As is usual on the first day, we are very optimistic that we can solve Marin's problem in the next couple of days.


Most probably, only tomorrow we will understand why it is difficult. Anyway, we quickly got the nonlinear optimisation problem with constraints (e.g. the ship should stay in position) on the white board. The white board is tiny by the way: good for keeping the discussion focused. It is down now to obtaining meaningful estimates for the parameters and then we should be off and running. Running some code that is; we seem to have numerical gurus as well as visualisation experts in the group.

Phytocare

We are quite a large group, something like 15 people. After an initial query session we split up into groups working independently, some reading literature that was provided and some searching for the important quantities and their interdependencies. One of the main points is to understand how algea growth is influenced by the various parameters such as CO2, oxygen, stirring, and nutrient influx. An important issue seems to be how these influence the photosynthesis, which then effects growth... Let's see tomorrow what these groups did!

SWI2010 has Taken Off - Impressions from Monday

Great to have you all here!

Here you can see some impressions of today:

The very first post...

Welcome!

I am delighted to present to you the official weblog of the Study Group Mathematics with Industry.

As a participant of SWI2010, you are encouraged to post on this blog a day-report, just some remark or even photos. Also, CWI's communication department will regularly post photos of the Study Group event.

If you are interested in posting on the blog, just ask us for the login credentials.

See you all on monday,

Niek

[on behalf of the SWI2010 organizing committee]