On the Characterization and Optimization of Machining Chip Washing System Using Multi-Variable Response Surface and Gradient Descent Method

The final step in manufacturing high-precision parts for internal combustion engines, such as cylinder heads and blocks, is the removal of machining chips from the finished parts. This step is essential because the machining chips and cutting oil residue left on the surface after machining can cause quality issues in the downstream engine assembly process and impact the cooling system’s performance during engine operation. This chip removal process is especially critical for parts with internal cavities, such as the water jackets in cylinder heads, due to the difficulty of removing chips lodged in the narrow passages of these internal channels. To effectively remove chips from water jackets, high-velocity water jets are commonly used, creating flows with enough kinetic energy to dislodge and carry away the debris. For a machining chip washing machine equipped with dozens of water nozzles, optimizing washing efficiency is a significant challenge due to the large number of variables involved in the process. Additionally, the optimization objective and procedure could vary depending on the specific constraints of the process and the chosen performance criteria. The goal of this paper is to develop an engineering process with two main objectives. The first is to develop a systematic procedure for determining the characteristics of the machining chip washing system, based on different performance criteria. These criteria include both a local perspective, which focuses on individual nozzle performance, and a global perspective, which assesses overall machining chip washing efficiency. The second objective is to use an algorithm to optimize the system based on these characteristics. For local nozzle performance, the response surface method will be employed, while global washing efficiency will be optimized using the gradient descent method. By utilizing this approach, the complex high-dimensional optimization problem can be broken down into smaller, more manageable sub-problems, which can be solved using conventional optimization algorithms.