To achieve a method for flame exposure testing of high-pressure cylinders in automobiles that allows fair evaluations to be made at each testing institute and also provides high testing accuracy, we investigated the effects of the flame scale of the fire source, the fuel type, the shape of the pressure relief device shield, and the ambient temperature through experiments and numerical simulation. We found that, while all of these are factors that influence evaluation results, the effects of some factors can be reduced by increasing the flame size. Therefore, a measurement technique to quantitatively determine the flame size during the test is required. Measuring temperatures at the top of each cylinder is a candidate technique. Furthermore, flame exposure tests to be conducted on cylinders as single units must ensure safety during a vehicle fire. To demonstrate this, we conducted vehicle fire tests on vehicles equipped with cylinders and compared the results with the flame exposure test. As a result, we found that the flame exposure test differed from the vehicle fire under all test conditions, so evaluation of safety through a flame exposure test on the actual vehicle level is recommended to improve reliability.