A high-density dense back-drilling PCB processing control method and system
By combining multidimensional signal standardization, adaptive Kalman filtering, and bidirectional long short-time memory network with local weighted regression, the problems of inaccurate interlayer interface identification, inaccurate layer prediction, and unsuitable parameter adjustment in the processing of high-density back-drilled PCBs were solved, achieving high-precision and high-yield back-drilling processing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- AN HUI SHENG PENG JIE ZHI NENG KE JI YOU XIAN GONG SI
- Filing Date
- 2026-04-23
- Publication Date
- 2026-07-21
AI Technical Summary
Existing high-density back-drilled PCB manufacturing technologies suffer from problems such as limited signal acquisition and processing methods, insufficient accuracy in interlayer interface identification, insufficient accuracy in layer prediction probability distribution and confidence level, and a lack of intelligent dynamic adaptation capabilities in adjusting processing parameters, resulting in low processing accuracy and product yield.
Multidimensional raw signals are collected and standardized. Interlayer transition events are detected by adaptive Kalman filtering and bidirectional long short-term memory network. The feed rate and spindle speed are dynamically adjusted by combining local weighted regression method to achieve accurate layer thickness analysis and machining parameter optimization.
It improves the accuracy of interlayer interface identification and the reliability of depth prediction, enhances the dynamic adaptive adjustment capability of processing parameters, realizes precise control of residual pile length, and significantly improves back drilling processing accuracy and product yield.
Smart Images

Figure CN122432982A_ABST