Hi everyone. In this video, we're going to be discussing the robust design process. Now you'll remember from your previous videos that the robust design process is where we take a look at the different noise factors and variations that can happen in the manufacturing or delivery of a service or product that can impact the potential performance. But how did we plan it to perform under non-ideal conditions? With robust design through design of experiments, we ID their control factors. What can you control? What can't you control? What are the performance metrics associated? An example would be, let's just say, for example, we want to make a plant-based carbon similar to carbon fiber type armor. Because it's plant basis, we could use a sustainable crop such as hemp or bamboo. When we're building this, we can say, "okay, we know that these materials are going to perform under these certain conditions, but what are some variations that we can account for?" It might be a really good year for a certain crop, really bad year for a different one. Then we take a look at, so how does trade impact us as well in the process that might suddenly be a huge demand from foreign countries or bios, which could jack up the price because supply and demand needs to hid in equilibrium. Then you start looking at, well, this is going to really cut into our profit margins and we're not going to be able to respond to investors. When you look at robust design and non-ideal conditions, it's not just in terms of the production itself and the specifications, but you need to apply non-ideal conditions across the board with your corporate processes, your sales team, legislation that can show up. There's so much noise when starting a company and so many different variables that you need to account for that even if you're like, a non-ideal condition for use could be like a car, one of those simple colors non-ideal condition is that you can't go and get it from the trunk, so it has to be accessible by hand. It can't block your field of vision or get in the way of your steering or changing gears. It needs to be small and is to be able to cut the receipt belt, how they can be durable as a seat-belt. Using the cutter and then what do you do then you need to get out the window, so crack the glass and needs to be able to puncture it. Those are the used examples. But then when you look at the variations and non-ideal conditions for the manufacturing and production and corporate processes. That's a whole different barrel of monkeys. Develop an experimental plan. When you start to develop your plan, this is where data comes back in again, year spreadsheets in the data you've collected is going to be critical. Control and noise factors, control factors of the design variables to be varied in a controlled manner during the experiment. This might have to do with how a certain material will compare to use of a different one. You know, how a certain alloy with a certain composition will behave versus one with a different one. How will copper react? What is the melting point of one material versus another? Same thing with service-based delivery as well. When you're looking at which software platforms and applications you use. What are these factors that you can control for? The noise factors are those that cannot be explicitly controlled. A lot of these have to do with human error. Human error comes into everything. Because once you've got certain specification set and you know how different materials and inanimate objects are going to behave. The human influence is one that is very noisy when you're trying to create a company. This is why again, we come back to metrics for accountability, for performance, and being able to make sure that you can get the job done. Takes the emotion out of it, less heartbreak. The performance metrics are those specifications of interest in the experiment. Let's just say that you do create a plant-based composite armor that you want to be able to use in mission environments, and you want it to be able to withstand a certain amount of impact and abrasion. If somebody is wearing it while riding in a Humvee, then you can test, okay, how will it work if the person is throwing 10 feet from the vehicle, versus 20 feet? How will it work if the vehicle is going 60 miles an hour, versus 80 miles an hour? Then you can examine, okay, how does that hold up? Where are the fracture points? Where can we strengthen it in certain areas and where may we have to use a separate material to ensure that it will perform as predicted or as you desire under non-ideal conditions? What are the ranges? In the case of AMA, you could say, okay, well, we know generally that people aren't going to be driving over 100 miles an hour, so you don't have to plan for it to work under conditions of 200 miles an hour. Same thing is, if somebody's throwing like 50 feet or 100 feet from a vehicle, they've got much bigger concerns than abrasion. You know that the mortality of that's probably not going to be their biggest concern as a few scuff mugs. How do you set the different metrics? What's the probability of a certain outcome? Then what are the noise factors associated? Remember when you sell planning for your company, you have to factor these things into your pricing and how you are going to cover yourself if you need to get course-correct and get back on track. Maximizing, that's used for performance dimensions where larger values are better, such as maximum deceleration before a belt slippage. These are the formulas that you'll use. Depending on where you want to go with your product, you may find larger is better. Eta equals mu squared. Smaller-the-better, these aren't going to be as relevant for what you're doing with your concepts. One of the reasons is that unless you're developing a highly technical physical product, you may not need these. I own a software company and I have not used these formulas in that capacity. However, I did have to use it in grad school, so use cases. Develop the experimental plan. Experimental plans and designs include: Full factorial, fractional factorial, orthogonal array and one factor at a time. Again, if you're developing a highly complex physical item that requires really precise specifications, then this is where this is going to be a lot more relevant. For a service providing, not so much unless your service is a highly technical one of course. Planning for your experiments. Remember, you start your datasets early and you bring up all your data together and you can sort, rank, pull everything into one place and then see where these different variables like. Because if you leave your design of experiments and data collection until this part of the process, it's going to be a nightmare, and it's not going to be accurate. Start your Spreadsheets now while you're starting to examine the performance metrics of certain software applications, talent, people's skill sets, labor costs. All of these data, just start from the very beginning. Because I can tell you from experience, when you wait till this point to start your data collection, it's not going to be accurate. You can forecast, you can make some predictions, but it may be as accurate as storing a dot at the wall sometimes and seeing what the outcome might be. But this is where I can stress this enough; start your data collection from the very beginning. It's annoying, but it'll save you a whole lot a heartache. In summary, ID control factors and noise factors. This can be a really fun process. I personally like it because then you can look at all ridiculous situations that may come up, and then what parameters you set as points where you will and will more be able to perform in those ranges. Formulate an objective function, develop the experimental plan and keep track of all the changes as well and the data. Create a good folder that your team can start collaborating and putting data in and be able to sort the data. It may be useful depending on how technical your concept is, to use something like Python for being able to analyze the data and perform a good technical analysis. Then conducting the analysis as well. If data science and analysis isn't your jam, then see who in your team may be able to do that. It can be a little pricey to find a good data scientists, but it can be really useful if you need to hit certain specifications. Like if you need to meet military specifications and you want to be able to sell to the military, then you know you've got to work within these frameworks. Thanks.
