抄録
In product development process, we use experiments, 1D calculations, and 3D calculations (3D Computational Fluid Dynamics etc.) to find the design parameters solution that satisfy the performance goal. Product development is a task of transforming experimental and calculation result data into design information that can determine specification. In mechanical engineering, some data have interactions between design parameters and the performance and show strong nonlinearity. In order to show the effectives of machine learning method to analyses mechanical engineering data, we apply 3D calculations which simulate the room air conditioner’s air flow control in the room for following two data analytics. (1) The airflow distance is controlled by discharge airflow temperature, angle and volume rate. These 3 parameters have interactions in each other. We show decision tree method can classify airflow distance and 3 parameters data. (2) The temperature increase is affected by discharge air flow angle, but it show nonlinearity. Gaussian Process Method is visualizing the nonlinearity of data. We propose that the machine learning method is 4-th method of product development tool in efficient product design practice.
