Chip performance degradation trend prediction method based on multi-step robust prediction learning machine

Abstract

The invention discloses a chip performance degradation trend prediction method based on a multi-step robust prediction learning machine, which combines an extreme learning machine and a recurrent neural network, has extremely high information fusion capability and rapid information processing capability, constructs similarity based on correlation entropy by establishing an error code book, establishes real-time prediction model updating according to chip degradation diversity and dynamics, and overcomes the influence of interference on a prediction result. Therefore, compared with the existing method, the method has higher online prediction precision, and the multi-step prediction result is more accurate compared with the existing method.

Description

technical field [0001] The invention belongs to the technical fields of electronic device health management and machine learning, and more specifically relates to a chip performance degradation trend prediction method based on a multi-step robust predictive learning machine. Background technique [0002] With the rapid development of chip technology and the diversity of chip working environments, existing methods for predicting chip performance degradation trends face challenges in terms of time complexity and technical flexibility. On the one hand, with the improvement of chip information interaction speed, the prediction algorithm needs to quickly give the evaluation of the health status; on the other hand, due to the complexity and diversity of the chip application environment, the prediction algorithm needs to flexibly evaluate the health status according to the actual operation of the chip. However, existing chip performance degradation trend prediction methods usually ...

AI Technical Summary

Problems solved by technology

However, the existing online degradation trend prediction methods still have the following problems in engineering practice: First, the existing online degradation trend prediction methods cannot guarantee the efficient transfer of effective information in the historical data during the multidimensional prediction process, resulting in some effective Information cannot play a role in actual prediction; second, although the online pre-degradation trend measurement method can make corresponding adjustments based on real-time data information, parameter adjustment requires iterative calculations or re-checking of some models, which cannot adapt to the operation of related components Speed ​​requirements; third, since most models assume that the degradation trend of components satisfies the Wiener process, the model is vulnerable to non-Gaussian noise and singular points in the data, which greatly affects the prediction results

Images