I'm trying to develop a machine learning model to solve this problem, and am unsure of where to start.
We begin with some user-defined settings. The settings are used by a machine to create a product. The product will have imperfections (due to the user's choice of settings). Ultimately the goal of the algorithm is to recommend changes to the settings to reduce imperfections in the product. Due to the nature, environment, and noisy measurements of the machines involved, there are no "universal best settings" that can be applied, so changes to the baseline must be recommended (like temperature += 10 degrees instead of temperature = 30 degrees). Two machines in different locations can produce different results with the same settings. Many of the settings are somewhat dependent on each other - such as the machine can run at higher heat if there is also higher cooling, but an imbalance of heat and cooling will introduce imperfections.
Is there a name for this specific type of problem? Are there any similarities to other problems here? I'm not really sure what to even google to learn more about this.
Available data could be pictures of the product from a standardized angle, the settings used to create the product, and a calculated similarity score between the created product and ideal product.