Soft Sensor Model of Phase Transformation During Flow Forming of Metastable Austenitic Steel AISI 304L

This paper deals with the modeling of a soft sensor for detecting α’-martensite evolution from the micromagnetic signals that are measured during the reverse flow forming of metastable AISI 304L austenitic steel. This model can be prospectively used inside a closed-loop property-controlled flow forming process. To achieve this, optimization by means of a non-linear regression of experimental data was carried out. To collect the experimental data, specimens were produced by flow forming seamless tubes at room temperature. Using a combination of production parameters (like the infeed depth and feed rate), specimens with different α’-martensite contents and wall-thickness reductions were produced. An equation to compute α’-martensite from both specific production-process parameters and micromagnetic Barkhausen noise (MBN) measurements was obtained using numerical methods. In this process, the behavior of the quantity of interest (namely, the α’-martensite content) was mathematically evaluated with respect to non-destructive MBN data and the feed rate that was used to produce the components. A combination of exponential and potential functions was defined as the ansatz functions of the model. The obtained model was validated online and offline during the real flow forming of workpieces, obtaining average deviations of up to 7% α’-martensite with respect to the model. The implementation of the soft sensor model for property-controlled production represents an important milestone for producing high-added-value components on the basis of a well-understood process-microstructure-property relationship.