Thoughts on a possible "Quality Control" mod.

[mrudat]

Here's some notes on a mod I'm contemplating making, just in case someone else thinks it's an awesome idea and implements it before I get a round tuit.


Create a new "Quality Inspection" furnace-style entity (automatic recipe selection).


Create a new "Rework Report" used to perform research to unlock better levels of quality inspection.


Take all existing items and:

• Create an untested-<item> that has not been through quality control
• Create a failed-<item> that has failed quality control.
• Create a series of quality-inspection-<level>-<item> recipes that have a chance to return either an <item> or a failed-<item>, increasing level of the recipe has an increased chance of success.

Take all existing recipes and create a new rework-<recipe> version that takes a failed-<item> in addition to the original ingredients, and produces an untested-<ingredient> for each original ingredient, to represent disassembling the item and replacing individual components until it works, where the individual components replaced might be faulty.

For some recipes, this might not work. For <ore> to <plate>, for example, it would probably work to have <ore> + failed-<plate> => 2 untested-<plate>.

For all rework-<recipe>s, add as a result a new "Rework Report" that can be fed to a laboratory to research higher levels of quality-inspection, to improve the effective yield of a crafting recipe.


Remove all existing technology, unlock all recipes from the start, and generate a replacement technology tree based on the original tree that uses "Rework Report"s to unlock higher levels of quality-inspection.

[Impatient]

I try to think how this would affect gameplay and base layout.

1. After each production line there would be the QI facility (because, there the most items of this type have to be tested)
2. After the QIF there would be the rework-facility for that item type.
3. The only reason to rework faulty items and not to blow them up in a chest, as the rework costs more than making new ones, is to get the rework reports to improve quality. Right?
4. The rework reports have to be transported to the QA lab for quality research.
5. The other outputs of the rework facility have to be routed back to their respective QI facility. Most likely on rainbow belt(s) running opposite the bus direction.

That is how I imagine the basic incentives and factory components a mod like this will add to the game.



What I like with this idea is, that there are residual products in a production result, which have to be taken care of. This adds an additional layer of complexity. There are other ideas and I think mods which use that kind of game mechanic. I guess the specifics make the difference between players adopting a new playstyle or ignoring the mechanic or being unnerved by it.

[Impatient]

Three questions come to my mind.

1. What initial percentage of faulty items do you have in mind?

If the percentage is low (let's say 10%), there might be a small incentive to rework faulty items. Although over several production steps the return on final products diminishes multiplicatively ( 0.9 x 0.9 x 0.9 ... ), players might try to ignore the quality problem until they get the logistic network, where thy can ship everything around with bots.

If the percentage is high (let's say 50%) this might be a heavy burden on expansion capabilities. The player would need to implement rework-, routback- and research infrastructure right away. That infrastructure would be heavy. As there would be a lot of reports, the burden could be lifted to an acceptable level quickly-ish. In fact the percentage of faulty items is self-balancing. If there are a lot, a lot of research can be done, which lowers the amount of faulty items and less research can be done. ... Makes me wonder if there would be a point, where having QA related infrastructure would not pay off anymore.

2. Are you planing of having one "rework report" item tpye as a result of the reworks of all faulty item types?

That would imply, that I can use reports received from reworking electronic circuits, to improve the quality of eg rocket control units. I did not do the math on this, but this might turn out to be misusable.

3.

...
Remove all existing technology, unlock all recipes from the start, and generate a replacement technology tree based on the original tree that uses "Rework Report"s to unlock higher levels of quality-inspection.

Can you elaborate on this? Why does the original tech tree have to be removed? Or is this a design decision? When I read the specifications for the mod up until that point, I pictured the normal tech tree and quality related branches of tech for each recipe forking from the tech that unlocks these recipes. The quality related techs in a branch are prequesites to only the next tier of quality tech in that branch.

That way, I as a player have to do the normal research, but I also have the choice of doing research on quality or not.

[Impatient]

And here is an opnion of mine:

The QI "furnace" and the QI recipes together should be fast enough to do the QI at the throughput of one belt. That way the player can use loaders to funnel the items through a QI unit. At least that is what I would feel comfortable at the moment I write this. Imagining that I need an extensive complex of slow QI units after each production line takes some fun away. Maybe several tiers of QI units or a basic QI unit that is speed modable to adapt it to higher belt speeds would do the trick.

[Ranakastrasz]

https://mods.factorio.com/mod/ProductionScrap2
Not the same thing, but quite similar. instead of needing to test for flaws, the produced items are either flawed or not, and require recycling and reprocessing.

[mrudat]

The thought I had was that it would replace the vanilla tech tree; the higher up the tech tree the higher the starting chance of failure, and the more expensive the research, ideally balanced to be around the same amount of effort to complete as the original tech tree. Each technology would asymptotically approach no failures.

As an example, iron-gear-wheel, which you get from the start, there would perhaps be a 5% failure/rework rate, which after research you could get down to under 1%, still producing a slow but steady stream of reports.

On the other end of the tech-tree, with blue circuits, initial failure rate may be up to 98% (because logically, you haven't yet solved the engineering challenges with setting up a circuit manufacturing plant), and you'd work your way through 50% to 25% ... until you get all of the bugs sorted out and you have under 1% failure rate for blue circuits as well.


Oh. I might not have been clear, the rework-<item> recipe would take failed-<item> + ingredients and yields <item> + untested-<ingredients> + report. There's no net loss, other than in needing to run the (possibly faulty) ingredients through QC again.

Perhaps there should be somewhere, though, in order to act as a resource-sink, as opposed to just a time+effort sink, though perhaps not quite as severe as the destructive investigation of a failed item + the ingredients to craft a new one?


It might work well/better as an addition to the tech tree? I hadn't thought much past the mechanism to generate science-packs for arbitrary recipes.


I figure that there could be an upgradable set of inspection facilities; not sure what sort of tiering would be best? Probably have a tiered set that can be run at belt-speed, similar to the stacking beltboxes mod? You'd probably need a 2x2 entity, as you want the room to comfortably connect 3 loaders to it.


I figure that, unless you actually reduce the failure rate to zero (and thus require reworking the now redundant parts of your factory), you will continue to generate a small stream of failure reports, and they need to be gotten rid of somehow. I don't think that's any different to the materials-science of improving copper purity and manufacturing-quality of iron-gear-wheels being used to help research, say, batteries.


The main thought is that you could place down a blueprint of a starter base, build some of the initial infrastructure by hand, including at least one roboport and construction robot, set it to automatic research, and come back to have the base filled in over time as the various manufacturing steps become more reliable over time.

[ZombieMooose]

What would be the penalties for skipping the QC phase? Would it increase the chance of other items if say you have a faulty circuit go through to produce something else?

How will this impact entities? Would it stop them from working, or glitch out occasionally?

Would this factor into a maintenance system where faulty items would break down quicker needing to repair an entity sooner?

[Impatient]

...
Take all existing items and:

• Create an untested-<item> that has not been through quality control
• Create a failed-<item> that has failed quality control.

...

mrudat's specification is, that for each <item> (eg car, icon: car) an untested-<item> (eg untested-car, icon: car?) and a failed-<item> (eg failed-car, icon: car!) is implemnted. So I imagine, that after production you get car?, which you can not do anything with, except run it through QC. Only after QC you either get car or car! . And only car can be used to place a car in the game world. So in my understanding QC can not be ommited.

I think the whole specs are built around that mechanic. Thus no expensive scripts have to be run in any event handler. Nothing expensive at all. It seems everything can be done in data. Pretty well thought out imo, if my understanding is correct.

[mrudat]

...
Take all existing items and:

• Create an untested-<item> that has not been through quality control
• Create a failed-<item> that has failed quality control.

...

mrudat's specification is, that for each <item> (eg car, icon: car) an untested-<item> (eg untested-car, icon: car?) and a failed-<item> (eg failed-car, icon: car!) is implemnted. So I imagine, that after production you get car?, which you can not do anything with, except run it through QC. Only after QC you either get car or car! . And only car can be used to place a car in the game world. So in my understanding QC can not be ommited.

I think the whole specs are built around that mechanic. Thus no expensive scripts have to be run in any event handler. Nothing expensive at all. It seems everything can be done in data. Pretty well thought out imo, if my understanding is correct.

That is exactly the case. Do you want to drive a car if you don't know if it will just work, or immediately explode when you get in? =)

[darkfrei]

The main problem is that the game has bad support for probabilities and amounts.

The recipe has ingredient "not-tested-gear", and you can get results: 90% "good-gear", 10% "bad-gear", but you can get both of them simultaneously! Or get nothing.

[Impatient]

...Or get nothing.

?????????????????????????????????????????????????????????

[darkfrei]

...Or get nothing.

?????????????????????????????????????????????????????????

Exactly! The 10% and 90% are independent and you can produce no item at all.

[Impatient]

Exactly! The 10% and 90% are independent and you can produce no item at all.

https://lua-api.factorio.com/latest/Con ... ml#Product

Very interesting! I thought the probabilities always have to add up to 1 and one of the products is going to be it. But reading the doc, I realise, the api allows for pretty crazy recipes.

Imagine making a power armor mk2 or a nuclear reactor and getting nothing after QC.

Luckily products with probabilities can be ensured by simulating the probabilities with stack recipes. Eg for two results with probabilities of 85% and 15% the recipe has to require 20 inputs and has to make 17 of product A and 3 of product B.

Edit: I just realise,that in case of very expensive single items, like the mentioned mk2, simulation of probabilities by using stack recipies is only a tad better than using the probability settings themselfs and risking getting nothing at all. Because the recipe would require 20 mk2s to make 17 goode ones and 3 failed ones. I imagine, selling that to the players, will pretty tough.

[Ranakastrasz]

I seem to recall a mod that scaled up recipes, which worked great, except for nuclear ore refining, since it just multiplied things. Ratio of like 50, so you had to run it 50 more times to have the same chance of getting ANY refined uranium.

My opinion is that the data api should have the ability to entangle multiple results, such that they use the same random number, and treated like a grab bag (with replacement)

[Hiladdar]

The game already has something like that, programmed into it, with uranium ore to uranium 235/238 processing.

Beyond that, what would make sense is to allow the player to determine the failure rate, something between 0.01 and 10%. It would make sense that the more complicated the assembly is and the longer it takes, the closer to the player set max failure the combine would be. One other variable would be which tier of assembly machines is used to make the combine. Lower tiers should fail more, then upper tiers.

One way to address this issue, is if the assembly machine puts a broken item on the belt, then the it is up to the player to figure out how to remove that off the belt or logistical system, and what to do with it. The other question is what to do with the broken item.

The other question, is what happens when the player is dealing with fluids. Do you want to touch that, and have this process also extended to chemical plants plants and oil refineries. What about centrifuges for non uranium ore conversion? What about smelting, do you want to have defect in the smelting process? Finally will the miners also have QA issues?

I really like what you are doing, which is developing a parameters as to what the mod will do and will not do.

Hiladdar

[darkfrei]

What about smelting, do you want to have defect in the smelting process? Finally will the miners also have QA issues?

Poor, normal and reach iron ore with probabilities 50%, 40% and 10%, that can be smelted into plates with probabilities: 1*50%, (1*100%+1*50%) and (5*100%+5*50%), maybe with small chance for sulfur, stone or iron ore.