The Research And Realization Of Blade Inspection System Based On Denoising Using Concave Hull Algorithm

Turbine is an equipment which changes the power of kinetic energy of vapor to electricity and its blade has significant influence on the efficiency of the transformation. How to realize a quick detection of turbine blade has become a hotspot of research. In this article,data process problems are discussed which may occur during the blade detection period begins with blank blade and ends with finished,including denoising of data,matching of points,adjustment of mould and extract of front&rear edge of the blade.This article includes all of the following aspects:Blank blade data process: In this part the problems including denosing and smoothing of data are discussed. The article proposes a new method based on concave hull to sort the point of a blade section and then, on the basis of sort, we realize the denosing of the blade section data. When it comes to the smoothing of data, the article suggest to take feature points into consideration in order to maintain the integrity of blade features.Matching of data and adjustment of mould: The article researches the method to match the points and adjustment of the mould. In this article, we use PCA(Principal Component Analyze) method to get the initial position of the data, then ICP(Iteration of Closest Points)method is processed to get the final match. After that, the matching of data is analyzed to get the adjustment of the mould.Finished blade data process: The article researches the fit of front&rear edge and dispersing of curves. The key point of finished blade detection is to get the parameter of a particular blade section such as max thick,middle arc etc,and the problem is how to determine the area of front&rear edge of a blade. In the article, a new method based on the error control theory is proposed to find the area. As the point of blade is so sparse, dispersing method of the curves is processed by using parameter spline curve.The development of blade detection software: the article introduces the principles of software development and module designment. The modules should contain data process,matching of points, analysis of curve's error, adjustment of mould and extract of feature parameters. At last, the article shows the procedure of blade detection by using a sample blade.