Richard Gedney of Admet says that a good test engineer must have an excellent understanding of the sources of errors that may be introduced during a test
Mechanical testing is widely used in engineering design, development and research, as well as during manufacturing to ensure a material or product meets some predefined specification.
A universal testing machine (UTM) is used to generate a stress vs strain diagram from which the mechanical properties of materials in tension, compression, bending or torsion can be determined.
Common properties of interest in tension are offset yield strength, Young's Modulus, ultimate tensile strength and total elongation.
A true picture of the stress-strain diagram can only be obtained through accurate measurements.
This requires that the testing machine is configured properly and that the operator performs the test according to specification.
The accuracy and repeatability of the sensors are affected by the test frame, power transmission, grips and fixtures.
Sensors that are mounted in the wrong position, are heated up, or are deformed by mounting bolts all introduce measurement errors.
Before commencing testing, one should know how to measure the errors in order to keep them from creeping into the results.
Accuracy, repeatability and resolution.
There are three basic definitions to remember with respect to how well a testing machine can measure stress and strain.
They are accuracy, repeatability (precision) and resolution.
Crosshead position, which is sometimes used for strain measurement, can be used to explain the meanings of accuracy, repeatability and resolution:.
Accuracy is the ability to tell the true position of the crosshead on a testing machine.
Accuracy is the maximum error between any two crosshead positions.
Repeatability (precision) is the ability of the crosshead to return to the same position over and over again.
Repeatability is the error between a number of successive attempts to move the crosshead to the same position.
Resolution is the larger of the smallest programmable steps in crosshead position or the smallest mechanical step the crosshead can make.
Factors that affect accuracy, repeatability and resolution.
Hysteresis is the maximum difference in sensor output between measurements made from 0-100% full scale output (FSO) and from 100-0% FSO.
Although hysteresis is easily measured, its mechanism is not fully understood.
Linearity is the variation in the constant of proportionality between the sensor's output signal and the measured physical quantity.
It is often expressed in terms of a percentage of FSO.
No sensor is truly linear so a least squares fit is commonly applied to fit a straight line to the sensor's output graph.
The least squares line is the line drawn through the sensors response curve such that the sum of the squares of the deviations from the straight line is minimised.
Microprocessors can perform multi-point calibrations to effectively map out the non-linearities.
Noise is the magnitude of any part of the sensor's output that is not directly related to the physical quantity being measured.
Force and strain resolutions on most testing systems are user programmable.
The programmed resolution should always be greater than or equal to the noise.
Sensor location.
The most important consideration for a sensor is where to mount it in order to ensure that the desired quantity is accurately measured.
The sensor can be mounted on the input or output ends of a transmission.
If the sensor is mounted on the input end of a transmission along with a motor, the resolution of the system will be enhanced by a factor equal to the transmission ratio.
However, backlash and compliance in the transmission, belts, ballscrews, test frame, grips and fixtures will also affect the output of the sensor.
On the other hand, if the sensor is mounted on the output end of the transmission, it will more accurately measure the process but the resolution will be reduced.
Example - using a linear displacement transducer between the moving crosshead and machine base to measure flexural strain.
The mechanical properties of plastics, composites and concrete in flexure are frequently measured.
A specimen of rectangular cross section is placed on two rigid supports at spacing, L, apart.
Halfway between the supports a downward force is applied causing the specimen to flex.
A flexural stress vs flexural strain diagram is generated.
One approach to measuring flexural strain is to install a linear displacement transducer between the machine base and moving crosshead.
No structural member is truly rigid.
Therefore, analysis of the sensor location relative to the transmission of forces through the testing machine will show strain measurement errors are introduced by the following:.
- Compliance in the opposing tapered roller bearings.
- Stretch in the ballscrews between the machine base and moving crosshead.
- Backlash in the ballnuts mounted in the moving crosshead.
- Flex in the moving crosshead and machine base.
- Compliance in the load cell, loading nose and specimen supports.
- Hysterisis, noise and non-linearities in the displacement transducer.
Most of these errors will cause the measured flexural strain to be greater than the actual specimen strain.
Some of the errors are also non-linear making it even more difficult to characterise the behavior of the material.
Two alternative methods of measuring flexural strain that will yield less error are:.
Method 1: Halfway between the supports, measure the relative displacement between the machine base (point 5) and the underside on the specimen (just below point 3).
Method 2: Attach a rigid bar along the neutral axis of the specimen (red line between points 1 and 2).
The bar contacts the specimen at points 1 and 2 only.
Connect a displacement transducer to the bar at point 3 to measure the relative displacement between point 3 and point 4 on the loading nose.
Method 2 will yield more accurate strain measurements (Method 1 measures the deflection of both the specimen and supports.) but requires more complicated fixturing.
A careful analysis of each test setup and results is required to determine if the errors are small enough to be inconsequential.
Conclusion.
All experimental measurements include errors.
Before testing, a good test engineer will always ask the question: "Are my measurement errors small enough to not matter?" A thorough understanding of the sources and magnitudes of the errors is paramount to making accurate measurements.
