Abstract In general, flexing tests are subject to large errors, some of which have been studied using a test of the du Pont belt type. An improvement in the standardization of the test conditions can be effected by using a suitably designed belt balance arm and regularly adjusting the length of the belt. The rate of cracking and also the reproducibility of the results is affected by the design of the test-piece and also by the type of fastener used. Although the error between the results on replicate test-pieces is extremely large, more important are the large errors due to lack of reproducibility of replicate molded slabs, and this variability probably arises during vulcanization. Extremely large differences can also occur between nominally identical repeat experiments. This may be due to differences in the judgment of the stages of cracking in different experiments. Graded photographs were found not to be of any value in improving the reproducibility of results between two different operators. Attention is drawn to the differences in types of cracking obtained with the du Pont and De Mattia test-pieces. It is pointed out that with the du Pont test-piece the cracking tends to develop in depth more than is the case with the De Mattia, and therefore it is more difficult to construct or use graded photographs with the du Pont test. Accepting the fact that variations in the disposition of the stages of cracking do occur between two different experiments, a method of treating flex-cracking results graphically is suggested. By this method random variations in results are ignored and the average flex-cracking results obtained. It is suggested that flexing properties may be defined by two characteristics, viz., (1) Crack initiation time, and (2) Crack propagation time, and an example of the method of determining these characteristics is given.
Read full abstract