The real hero of the Fourth Industrial Revolution may be generative Design

As early as 2015, when Industry 4.0 was in full swing, an Airbus partition design was unveiled at many exhibitions and also won numerous awards. This is the complex and aesthetically pleasing structure in the image below.

8440-2150 The bionic partition wall is optimized with generative design, which greatly reduces weight while maintaining the same stability. The following video will help you understand the generative design process.

In general, the designer first specifies the optimization scope and constraints, so that the computer on the outline line of this partition, according to a certain distance to set nodes, and then connect the nodes. If all nodes are wired, then the structure is very rigid, but there are too many support rods and too much weight; If you reduce the number of support rods, the economy will be better, but the structural rigidity will also be reduced. Thus, the trade-off between programming balance is the crux of the problem. In fact, generative design is to generate the appropriate parameter combination through genetic algorithm, and then select the excellent parameter combination by simulation calculation. Finally, the designer is given several groups of reasonable design schemes, and the designer makes the final judgment.

The program eventually generated more than 100 feasible solutions, which users can filter according to optimization goals. The most important criteria for Airbus to evaluate this project are rigidity and weight, so in the end, the best balance between function, cost and beauty is achieved. 8440-2150 This kind of design is impossible for humans to complete, and it also relies heavily on computing power, because every time a scheme is generated, its structure must be simulated, and the parameters of the simulation are different for different optimization goals. So the whole process is very time-consuming, but also consume resources, ordinary enterprises can not afford, but also rely on cloud computing subscription software services.

When we talk about “Generative Design”, we have to mention “Topology Design”.

8440-2150

Topology optimization design also relies on 8440-2150 finite element, so that the program constantly calculates the least stressed element, and removes it from the grid, and then computes, and constantly removes unimportant materials. This cycle of iteration leads to the most efficient structure. This feature is already widely used in European enterprises, the algorithm is constantly optimized, and the computing speed has been reduced from weeks 10 years ago to minutes now.

So I don’t subscribe to the idea that the Internet and AI in Europe are stagnating. European industrial software is very developed, and the process of large-scale and intelligent has not stopped for a moment. But the focus of manufacturing in Europe and the United States has been on research and development and design, because this link has been a huge growth potential.

The following picture is a simulation of the milling process done by a tool 8440-2150 manufacturer 15 years ago. By comparing with actual turning data, they continuously optimize the material and process parameters of the simulation model, and finally make the simulation effect close to the reality. In the video, the process of chip generation to curling is perfectly simulated, and the heat, stress, strain and other parameters generated by the whole process are accurately calculated.

In the process of optimizing the simulation model, the physical experiment and the simulation model form a pair of CPS (Cyberphysics system), which can be beautifully named Process 4.0. The difficulty of parameter adjustment of this simulation model is no less than that of the most popular training large language models at present, which requires powerful computing power, excellent algorithms and a large number of manual experiments. Once the simulation model is successfully optimized, the problem of part trial-production can be solved almost once and for all, and the size of the blade can be arbitrarily modified until it is satisfied.