Part machining surface prediction method based on space-time full convolutional recurrent neural network
A cyclic neural network and parts processing technology, applied in the field of image processing, can solve problems such as inability to share information between layers, meet the requirements of production takt, improve prediction accuracy and training speed, overcome the insufficiency of fitting ability and feature extraction ability Effect
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Example Embodiment
[0051]Example 1
[0052]A surface prediction method based on a space-time full convolution cycle neural network, the method comprising the following steps
[0053](1) Generate a grayscale image: Measure the top surface of the engine cylinder by using three-dimensional high resolution surface morphology measurement technology, sequentially obtain three-dimensional high-density point cloud data, which is three-dimensional coordinates in x, y, z The format shows that the X, Y direction resolution is 150μm, the Z direction measurement accuracy is 1 μm, the point cloud data density can reach 40 measurement points per square millimeter, the total number of points can reach 1 million, using the 3D point cloud Data reflects the high value mapping of the surface topography of the part to the pixel gray value, generating a grayscale image of the top surface processing process of the cylinder;
[0054](2) Get a subset: The image coordinates (211, 511) are initial sampling positions, 50 × 50 pixels are ...
Example Embodiment
[0061]Example 2
[0062]A component processing surface prediction method based on a time-space full convolution cycle neural network, the method comprising the following step (1) to acquire grayscale image: This embodiment is a direct-obtained bearing side gray image by industrial HD camera shooting. Therefore, three-dimensional high-density point cloud data conversion is not required.
[0063](2) Get a subset: The image coordinates (668, 300) are initial sampling positions, 64 × 64 pixels are initial sampling size, 32 pixels are sampled interval isometric samples 10 times, resulting in 10 samples, sampling resultsFigure 7 Indicated.
[0064](3) Space autocorrelation analysis: Analyze the spatial autocorrelation of 10 sheets. After the 10 sub-maps are vacuied, the corresponding Molan index is calculated separately, and the maximum 256 × 256 pixels of Molan Index are the highest dimension as the spatial autocorrelation.
[0065](4) grayscale image slip window cut: Take 256 × 256 pixels as the sl...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap