| 摘要 |
A computer executing a software algorithm may be used to detect a depression in a temperature profile. The temperature profile may be smoothed to eliminate noise. Next, the temperature profile's center may be extracted. A polynomial may be fitted to extracted data. An algorithm used to fit the polynomial may guarantee that the fitted curve's peak may be below the actual temperature data's peak. Next, residuals may be calculated by subtracting the fitted curve from the actual data. If there is a dip at the center, then the residuals in the center may be less than zero. The software algorithm executing on the computer may then make a decision based on a sign of the residuals. For example, residuals less than zero may indicate bar porosity. Residuals above zero may indicate no porosity. The magnitude of the residuals may then be used to classify a size of a detected defect. |
