In this work we exploit the underlying dynamics of a turning process captured in force measurements for online flank wear estimation. We transform the sensor signals into feature vectors using recurrence quantification analysis and then estimate flank wear using a gradient boosted regression model. The data is collected by conducting two sets of turning experiments. The first set of data, which has 168 records, is used for training the machine learning model. The second set of data, which has 95 records, is used for testing the performance of the flank wear estimation method. The results indicate that the proposed method gives accurate flank wear estimates. The root mean square error of the flank wear estimation for the test data is 0.0497 mm.