Multi-stage semiconductor process virtual metering method based on Gaussian process and convolutional neural network

Abstract

The invention discloses a multi-stage semiconductor process virtual metering method based on a Gaussian process and a convolutional neural network, and the method comprises the following steps: (1), collecting an output signal of a process variable sensor related to a to-be-detected variable in a to-be-detected control process; (2) preprocessing the acquired signal data of the process variable, and removing an abnormal value; (3) rearranging the preprocessed data and reserving stage information of the data; (4) performing feature extraction on all data and establishing a regression model; (5)storing the current parameter weight, calculating the final maximum posterior value, and if the final maximum posterior value does not meet the stop condition, updating the parameter and repeating thestep (4) until the stop condition is met; and (6) storing the parameter value of each layer, recalculating the new prediction point and obtaining the probability distribution of the geometric quality. By utilizing the method and the device, the virtual metering result with higher precision can be obtained, the uncertainty of the prediction result is calculated, and a numerical basis is provided for further improvement of the model.

Description

technical field [0001] The invention relates to the field of data mining in industrial systems, in particular to a multi-stage semiconductor process probabilistic virtual metering method based on Gaussian processes and convolutional neural networks. Background technique [0002] Semiconductor manufacturing involves many stages. For example, in the production of electronic chips, a wire saw first cuts a silicon ingot into segments, then goes through several flat stages, including cleaning, polishing and grinding, and then transfers the treated wafer to the front and back where the final chip is formed. end process. Due to the high-throughput nature of semiconductor manufacturing and the high cost of measuring wafers, it is not possible to measure all quality variables of produced wafers at every stage. Due to the limitations of physical measurements, wafer-to-wafer modeling is increasingly being used to predict the quality of the final product, allowing timely adjustment of...

AI Technical Summary

Problems solved by technology

[0004] On the other hand, past methods can only obtain scalar predictions of outcomes, which means that the model cannot reflect the confidence of the predicted values

In theory, the higher the uncertainty, the less reliable the predicted value at this time

Images