Development of a Hopkinson bar for the evaluation of novel metallic materials

Abstract

A Hopkinson bar apparatus has been developed for the evaluation of hazardous materials which incorporates novel design features. Special considerations were made for the non-ambient testing requirement, with investigations into novel means of reducing the specimen cooling rate showing some promise but ultimately being rejected in favour of design alterations that enabled the rapid automation of the testing procedure. An ex-situ approach was taken, whereby the specimen was heated externally from the bars and then brought into contact with them a fraction of a second prior to test initiation. Specimen temperature homogeneity is investigated both numerically and experimentally. It is demonstrated that for small, conductive specimens, cooling can be effectively accounted for using a novel procedure. A validation study was completed in which identical copper specimens were evaluated on this new system and, using a more conventional in-situ heated Hopkinson bar apparatus at another laboratory, where the specimen and bars are heated together. Small differences in the measured flow stress at each laboratory were predominantly attributed to differing frictional conditions, loading pulse shapes and to some unusual variation in the strain rate throughout the test. This comparison demonstrates that the newly developed Hopkinson bar with unique capabilities for evaluating hazardous materials is effective and would be made more effective with minor design alterations which are being incorporated into a later prototype. After being fully validated, the newly developed procedures were employed to evaluate novel stainless steel 316L alloys across a range of strain rates and temperatures. Experimental data was used to calibrate some common constitutive models. Model calibration was achieved by determining the model coefficients individually using isolated data sets, and then using the determined values as an initial guess for optimising the models against the entire data set using a global optimisation algorithm. The accuracy of each model was quantified.