Dilir

https://www.chemistryworld.com/news/metallic-hydrogen-reveals-itself-under-mounting-pressure/4011102.article Nature paper from a french team that claims a significant advance in the search of metallic hydrogen

Just as a reminder : But yeah, the orbit is a bit off: http://spacenews.com/satellites-placed-into-incorrect-orbits-by-ariane-5-can-be-recovered-owners-say/ SES-14 stated that it will indeed take 4 more week to get to operating location, but it should not impact it's intended lifespan : https://www.ses.com/press-release/ses-14-good-health-and-track-despite-launch-anomaly As for Al Yah 3, no estimate was released yet.

Definitely a bad day at work tomorrow if it did fail.

Return with higher velocities / from higher altitudes than they usually do because of the fuel needed to land and the weight of the legs ?

Maybe i'm saying something stupid but what if they keep burning up after S2 sep with just the booster, to go really high and acquire data that would help them toward the goal of S2 reentry ? Edit : at least, it's coherent with the emblem.

Same here, themorris... :'(

They also want both the liftoff and splashdown to occur under daylight.

Wow 0.38m/s at the rebound. Not that far from escape velocity.

The global idea is good. However, keeping only the best set for offspring, while it allows faster convergence, often causes to fal into a local minimal. By keeping some "less good" individual, you ensure better genetics diversity at the price of computation time. I am not sure of the complexity of my algo. It's roughly number of individuals * number of generation to convergence, but i don't know how the latter relates to the former. I agree with you that for high complexity (3+ stages) rocket, if you limit the numer of generation, you might only get "goods rockets" and not the absolute minima. This is a tradoff, because for n stage, exhaustive search is, assuming you divide your rocket in 1000 dv units, (n-1) among 1000 which is roughly 1000^(n-1)/(n-1)! Anyway, genetic search was something fun i wanted to try out (because i never used it before) and seemed to be interesting for the problem i am trying to adress. I do not say it is the best solution, and I would be happy to discuss with you about other algos.

You got the tricky part right. I am really open to new ideas here, as it's my first go at theses algos. Mutating is for most of the case transfering some dv from one stage to the next or previous one. Sometimes, it's adding or removing an engine. Crossing is chosing a cutoff delta-v and take the upper part of rocket 1 and lower part of rocket 2. I got the idea of using these algo from 1 phd and 1 paper found here: https://etd.auburn.edu/bitstream/handle/10415/837/BAYLEY_DOUGLAS_5.pdf?sequence=1 and here http://strategic.mit.edu/spacelogistics/pdf/AIAA-2006-1720-231.pdf And now i'm really going to sleep!

Ok, here is what i do Function 1 (1 stage optimization) Imputs : required TWR, required size Harcoded : all engine specs (for now, i only use vacuum isp). Code Step 1: discretize dv and payload to n and m values Step 2: create two n*m matrix. First will contain best engine for each (dv,m) couple, second required fuel for that engine. Step 3: for each (dv,m), for each engine, compute the required mass to get dv. Store best (engine, mass) into dv,m point of storages matrix. Retrun the two n*m matrix. Obviously, this function is quite time consuming. But, as we hard store the result for a large number of dv and mass, it only need to be run once per required TWR in rocket design. Function 2: main algo Step 1: create N rockets (a rocket is a partition of dV). Step 2: evalutate mass of rockets using function 1. Step 3: select (with mass), cross and mutate the rockets. This is the genetic part. The idea is to use "good' parents to produce next generation. Good but not necessarely best allows better coverage of the optimization space and to avoid local minimas. Step 4: iterate until convergence. Step 5: keep best individual. Dv of each stage is given by definition of my rocket, corresponding engines are computed using function 1. I am going to bed now, but could discuss that furthermore tomorrow.

Well, we are talking optimization genetic algorythm, i don't know what computer scientist are calling genetic ^^. I can't take credit for the idea, it's well documented. I am not determining ideal child specs. I use my first function (only once per differents twr required, that's the trick) to generate two matrix. First give mass as a function of delta v and payload for single stage rocket, second give related optimal engine from those available. Then, a rocket is just defined by how you break your total DV. Each subset engine can be easily extracted from the matrix previously computed. This is still under testing and improvement and I will give more precise info when it will be a bit more polished.

Genetic, not generic :-). I am neither a programmer nor a specialist in optimization problem (my field is material science), so don't be too judging. The idea is to create a random dv repartition for a population of rockets. They would be the parents. Then with the previous function you compute what is the mass of each of them. You then use genetic algorithm techniques (selection, crossing, mutation) to generate a child population. After a given number of generation, you look for the best individual in your population. Oh, i also did an exhaustive search algorithm for 2 stages rocket, and the two give matching result. If you only have 2 or 3 stage, i would however suggest exhaustive over genetics, which is more suited to 3+ stages rockets.

I am working on something with matlab. Here is how i proceeded: - Build a 1 stage evaluation function. Imputs : minimum TWR (and later some engine size limitation). Output: optimization variable (IE mass or funds (currently mass), optimal engine as function of (dv, payload). The function evaluates the best couple of "engine + fuel " that match the required dv and twr. - Set global criteria for mission : total dv, payload mass, specific requirement (dead weight winglets for the first 2000m/s, twr >1.2 at launch, etc). - Create a random set of rocket (a rocket is a sucession of dvi, sum(dvi)=total dv). Compute with previous function total mass - Use genetic algorythm to find near optimal values. I am testing this in a new career. Works pretty well so far (i can save a few kg but that's due to algo rounding). I have done mostly 2-stages rockets for now, but the algorythm was designed for more. Right now, with a very coarse time consuming programation, i have rougthly 1min computation time. (2min if i add all the engines from KW, Novapunch, AIES) Concerning thrust plats and multiple engine, i guess that if you want to incorporate them, you should just define that as a new engine. Oh, you can also eliminate some dominated engine (for example, LV-T45 is dominated by LV-T30 if you don't use gimbaling).

French ESA tweet : further analysis show that harpoons were in fact not fired. Otherwise, Philae is in very good shape. ESA France @ESA_fr · 3 min il y a 3 minutes TT @esaoperations Une analyse + poussée de la télémétrie montre que les harpons n'ont finalement pas été tirés, @Philae2014 en très bon état :

Horrible french accent guy talking (I can bash, i am french )

Seems good!

Hello, Thanks for the good work! I am having questions about wing / winglets mass calculation. I know their is a mass/area involved, which i think is this one massPerWingAreaSupported = 0.065 My first question is : how can we see/compute a wing area? Second question : can we modify this ratio in game? (i understand that it will modifiy wing strenght).

I don't intend to hardcode it. The purpose of the tool is to do something like this - I want to go to X with payload Y, to do that, i need: - Burn 1 with Dv=... and minimum TWR =... - Burn 2 with Dv=....and minimum TWR=... .... With theses data, the optimal rocket design will be a N stage rocket with engine A=..., engine B=.... fuel A=...., fuel B=...., .... For now, it can compute the optimal single stage design (engine+fuel) for a tripplet (TWR, DV, payload), and choose between a 1 stage or 2 stage design to minimize total mass. Concerning the math, I use procedural tanks designs, so fuel tanks are not a problem. Engine have discrete performance, so their are obviously efficiency jumps at some point, but the code take that into account. The main benefits is for multi-stage designs. It is not trivial to divide the total DV in the right way. My code will tell you how to do that, and which engine to put.

Hello! I am designing a Matlab piece of code to compute optimal design rockets (mainly for fun, but i could be useful for a hardcore career). In the current state, with an imput of dv and payload mass, it will choose the most efficient design with either 1 or 2 stages (design=fuel + engine). The main problem is that to be able to do that, you have to imput a TWR. The point being, the lower TWR, the cheaper / lighter rocket you will ends up with. But TWR to low will make manoeuvres difficult. So my question is : what are the typical TWR you use (using Kerbin gravity), for theses manoeuvres: - Circularization burn - Kerbin orbit manoeuvre (or burn to get to mun / minmus SOI) - Munar landing - Minmus landing - Escape burn - Transfer to another planet

KSP ranks into my top 5 if I account for playing time. For records, i would put (with no order). Civilization Heroes of might and magic (3,4,5) Total war KSP Baldur's gate

When you reinstall all you mods, then manually erase the parts you don't need to squeeze in a few more mods.

I do like the Kerbal Experience very much, but when it comes to the effects, it is more delicate. What could an experience system bring to kerbals? My first idea is certification : kerbals needs to be trained on the ground, and it has a cost. Each command pod, descent lander, rover, piece of scientific equipement would require a certified user to actually work. You can add some situation as orbital EVA, surface EVA, docking. You can need certification to launch certain ships (nuclear ship, heavy ship, very high/low TWR ship). All of this could be obtained on the ground, with some money/time cost. They would be basic levels (limited EVA time, maximum mass for docking maneuvres, limited SOI). Second idea is mission gain experience : - Part of experience would be used to improve certifications - Second part would give bonuses. I am not against bonuses that breaks the physics games, as long as they are small. I would also want them to be ignorable without abandonning the whole experience system. It would also be great to have more realistics bonuses (limited automation seems a good idea, as well as RCS / SAS efficiency)

Hi and many thanks! This mod is one of the best RAMsaver / easy design mod. As you pointed out, cost are for the moment really unbalanced. I achieved some negative costs with emptied fuel tanks :-). Take your time to balance it, and if you need help for numbers/testing, i'll be glad to help.

Question. It was at some point mentionned that 0.25 would include a flight history log for kerbals. Is that done yet or will it be in another uptdate?