Three separate case histories have been related to IST
recently in which torsion springs have given disappointingly
low fatigue lives. In each instance, a fault with the raw
material was suspected as the primary cause of the shorterthan-
expected fatigue life. One of the springs was made from
music wire, another from oil-tempered silicon chrome and the
fi nal one was made from 302 stainless steel, but metallographic
examination showed that all three wires were of good commercial
However, IST observed that there were a number of
common factors with these failures. All were definitely
fatigue failures initiated at the outside surface of the spring.
The fractures were at 180° from the point of load application.
None of the springs had an effective mandrel to support the
coils in use. CAD programs predicted that the springs should
not have been at risk of fatigue failure at the given operating
defl ections. interaction between the spring and its mandrel that leads to uncertainty about the spring fatigue life of torsion springs, and the need to test to accurately evaluate fatigue life.
Computer-aided design programs for torsion springs
assume that torque is proportional to stress, which is correct
when a mandrel is in place. Without a mandrel, the lever length
is signifi cantly longer for a given torque, and so the stress
without a mandrel can be as much as twice as high, as shown in
Figures 1 and 2, right. The lever length is the distance between
the applied force and the position of maximum stress.
The top figure represents a torsion spring with external
radial legs, which is not supported on a mandrel. The stress is
related to the induced moment within the body of the spring.
The moment is equal to the product of the applied force
multiplied by the distance from the point of application of
the force to the position indicated as “Position of maximum stress.” This distance is equal to the radial leg length added
to the outside diameter of the spring. Note that double torsion
springs that operate without a mandrel should be treated as
single torsion springs. Furthermore, torsion springs that are
supported externally, often in a round housing, should also
have stress calculated as if there is no mandrel.
The bottom figure represents a torsion spring with external
radial legs, which is fully supported by a mandrel. The stress
is related to the moment within the body of the spring. The
moment is equal to the product of the applied force multiplied
by the distance from the point of application to the force to
the position indicated as “Position of maximum stress.” In this
case, this distance is equal to the radial leg length added to the
radius of the spring.
For this case, the fully supported spring will have approximately
half the stress of the spring without a mandrel. The
explanation for the shorter-than-expected fatigue lives, the
moral of this Cautionary Tale, is quite clear.
Mark Hayes is the Senior Metallurgist
at the Institute of Spring Technology
(IST) in Sheffi eld, England. Hayes manages
IST’s European Research Projects, the spring fatigue failure analysis service, and all metallurgical aspects of advice and training courses given by the Institute.
Readers are encouraged to contact him
with comments about this column, and
with subjects that they would like to
be addressed in future installments, by phone at (011) 44
114 252 7984 (direct dial), fax at (011) 44 114 2527997 or
e-mail at firstname.lastname@example.org.