Simcenter Testing Solutions Modal Impact Postprocessing: Getting the Best FRF

Figure 1: The Impact Post Processing workbook of Simcenter Testlab automatically sorts and identifies impacts based on desired force levels, presence of double hits, and overloads. In this case, two impact measurements (red areas in list and in time history) were rejected due to the presence of double hits while nine impacts (green highlights) were used in the final calculation of the FRF. 
 

In this article, current FRF measurement practices are discussed, and then the alternative FRF collection strategy is presented.

Article contents:
1. Background and Motivation
2. Standard Impact Testing Method
   2.1 Impact Hammers
   2.2 Workflow
   2.3 Manual Selection and Data Verification
3. Alternative Impact Testing Method
   3.1 Workflow
   3.2 Data Collection: Time Recording
   3.3 Automated Selection and Data Verification
4. Example Case
5. Conclusions

1. Background and Motivation

Often, the overall goal of a modal test is to correlate the modes of a Finite Element Analysis (FEA) model. If there are differences, changes are made to the model based on the test data. Therefore, it is important to have high quality FRF measurements as the foundation for the changes.  

There are fundamentally two excitation methods for modal testing, either with shaker excitation or with an impact hammer. Some of the differences between the two methods are summarized in Table 1 below: 
 

Table 1: Comparison of FRF measurements using shaker based excitation and impact hammer excitation.
 

The modal impact method is popular due to its simplicity and ease-of-use. Using the roving hammer method allows for a smaller amount of instrumentation and data channels from an acquisition system, hence saving costs and time.  However, the force levels are difficult to control, and the resulting measurements must be carefully screened for double hits, force dropouts, variations between hits, etc.

Shaker testing, due to its precise and repeatable control of the force input, often results in higher quality FRF measurements.  The setup effort for a shaker test is higher than impact testing.  Shakers must be carefully positioned at the appropriate locations on the structure, while an impact hammer can be used immediately at any location.

If the ease of impact hammer setup could be combined with comparable precision and control found in shaker testing, the result would be the highest quality FRFs with lower instrumentation effort.  The alternative FRF measurement method presented in the rest of this article explains how this can be done.

2. Standard Impact Testing Method

First, some more details about the current practices in modal impact testing.

2.1 Impact Hammers

Modal impact hammers come in a variety of sizes and weights, based on the force requirements and structure size, as shown in Figure 2. 
 

Figure 2: Variety of impact hammers for testing different size structures.
 

Some structures are easier than others, but in all cases, careful attention must be made in striking the structure precisely where the node of the geometry model is located. Also, it is important to strike in the correct direction. Sometimes this is difficult in tight spaces, as shown in Figure 3. 
 

Figure 3: Consistent impacts at the same point and direction can be difficult in tight or awkward spaces, especially if the operator must in parallel view a computer to accept and reject impacts.
 

It can get a bit clumsy to adhere to these requirements while attempting to strike at a consistent force level. Not following these guidelines leads to degraded quality in the results.

Automated hammers can eliminate some of the difficulties of manual hammer excitation. However, they add cost to a system, and sometimes are not so convenient in real test scenarios due to their size.  This takes away the simple and ease-of-use benefits. 

2.2 Workflow

During the standard impact measurement, a single operator must continually switch focus between two activities. The schematic in Figure 4 illustrates the flow: 
 

Figure 4: Schematic (left) of a common procedure for modal impact testing.  Focus is continually switched between two areas (right) – performing the actual impact hit and verifying the measurement. 
 

The operator first focuses their attention on the physical impact hit, then switches focus to the verification of the data collected from the hit. Once the data is verified and accepted, then the focus is switched back again to another impact, and so on until all impacts are made. 

Of course, solving the switching focus with dual operators requires more resources.

2.3 Manual Selection and Data Verification

After the hammer impact, at the “Data Verification” stage in the workflow, the focus of attention is turned to the computer, where the data will show characteristics which can indicate conditions for data acceptance or rejection. 

The data characteristics for impact that would be rejected are shown in Figure 5. 
 

Figure 5: The current impact average (red curves) has waviness due to a double hit that are not seen in the average (black curves) because this impact was rejected by the operator.
 

This impact would not be accepted by the operator, again forcing the computer to be close by for a single operator.
 

An example of the displays for an acceptable hit are shown in Figure 6. 
 

Figure 6: The current impact average (red curves) match the average (black curves).  The coherence is close to a value of one (lower right), the input force spectral response (upper left) is flat over the frequency range, and the FRF (lower left) is not noisy.
 

This impact would be accepted by the operator, again forcing the computer to be close by for a single operator.
Some additional information on these checks can be found in these knowledge articles:

Once the visual inspection of the data from the operator is done and the data is either accepted or rejected with a click of the mouse in the computer interface, the attention is turned back to another impact. 

The validation is done for each average (i.e., impact hit), and for one measurement location at a time.  This can be time consuming.  An alternative method with automated selection and validation is presented in the next section.

3. Alternative Impact Testing Method

What is the alternative to the current impact test procedures? This section explains a method, which utilizes the Impact Postprocessing workbook available with Simcenter Testlab 2021.1 and higher. 

3.1 Workflow

The schematic of the alternative impact test measurement workflow is given in Figure 7. 
 

Figure 7: Schematic of a procedure for a more efficient FRF impact acquisition.
 

During the measurement, the operator focuses solely on impacting the structure in a consistent manner. The operator still needs to do some cursory data validation checks like making sure that there are no overloads and that the accelerometers have not fallen down.

After the measurement is finished, an automated selection and verification is performed as is described in an upcoming section.

3.2 Data Collection: Time Recording

The first step in the alternative procedure is data collection. In this step, a time recording is initiated, and the operator will impact repeatedly at a single point for a certain number of detected impacts. 

Time recording can be turned on in Simcenter Testlab Impact Testing by selecting Tools -> Add-ins -> Time Recording. As shown in Figure 8, click on the “All Settings” in the upper right of the Measure worksheet, then turn on “Enable Throughput”. 
 

Figure 8: In Simcenter Testlab Impact Testing, click on “All Settings” in the upper right of the Measure worksheet, then turn on “Enable Throughput”.
 

Note: It is not always necessary to acquire the impact data with Simcenter Testlab. Time data files from another time recording system can be used as well.  The time data just needs to be imported into Simcenter Testlab.

Once the number of impacts is accounted for at a measurement location, then the operator moves to the next point and takes another set of impact measurements.  The placement of transducers is no different than any normal roving hammer or roving accelerometer impact test. This procedure continues until all desired impact points on the structure are measured. 
 

A graphical display of the resulting time data is shown for a simple impact test in Figure 9.
 

Figure 9: Time data for the impact hammer and accelerometer measurements for two runs of an impact test.
 

There are certain advantages that come with recording the raw time data. If any problems were determined like incorrect sensitivities, then this can easily be corrected with the recorded time data. There is also the possibility for conditioning the data with filtering or resampling.  It is also possible to change the applied window or change the frequency resolution (frame size).

Another advantage is to use a separate channel and a headset with a microphone that an operator can use to annotate the time data as it is being acquired.

3.3 Automated Selection and Data Verification

Once the time data is recorded, then the next step is impact post processing.  The dedicated workbook “Impact Post Processing” can be turned on under “Tools -> Add-ins” for 25 tokens as shown in Figure 10. 
 

Figure 10: The “Impact Post Processing” workbook can be enabled with the appropriate add-in in Simcenter Testlab.
 

Impact post processing enables the selection of impacts which meet certain criteria. This is an automated procedure which alleviates the operator from this manual burden. The results of this automated data selection are summarized in tabulated form, with an example shown in Figure 11. 
 

Figure 11: Tabulated results from an automated selection of impacts, showing ones that are accepted and highlighting the ones that are rejected. In this case, only four out of eleven impacts are being used to calculate the average FRF.
 

In this case, two impacts were rejected because of double hits, and seven impacts were not within the acceptable input force range between 10 N and 30 N..

Even though the selections are automated, the user can overrule the selections with the checkboxes on the left side of the screen.

The table in Table 2 provides a list of criteria used in automatically selecting and rejecting individual impact hits from the recorded time history in the Impact Post Processing module. 
 

Table 2: Criteria used to automatically accept and reject impact hits from time recording.
 

An example of the force range check is given in Figure 12. This shows a graphic representation of the required force level and the range that is specified as part of the selection criteria.
 

Figure 12: Top graph – Input force time history.  Bottom graph – Accelerometer response time history.  The shaded blue area shows a user defined acceptable range of force levels for inclusion into the averaged FRF.
 

Once the automated algorithm has done its job of sifting through the time data to select the impacts and tabulates the information, the next step is to present the results. Further human interaction can be used to validate the findings of the automation, and to override or otherwise modify the results. 

An example of the data displays for interaction is shown in Figure 13.

In the strip chart display, the individual impacts are shown by highlighting with color code all the time blocks. Double cursors are placed around each detected impact:

The size of the time block is determined with the spectral parameters specified by the operator in the Post Processing setup. If the option to show detected impacts and status is selected, the impacts can be visualized based on the following: color coding:

Once the data is selected and verified, the dataset is created which can then be used for modal analysis. All the data that was available before with conventional test methods using Simcenter Testlab Impact Testing module are also still available using the “Impact Post Processing” module. The "Impact Post Processing" module stores additional information including the time blocks for each impact. They are stored in separate folders within the project based each impact and response point.

4. Example Case 

Sometimes even for people with significant amount of experience in modal impact testing, it can be a challenge to make an acceptable impact. For example, if the structure is on a soft suspension, it can bounce back onto the hammer before the hammer is pulled away, causing a double impact. Also, it takes a lot of experience to avoid a double impact, and to strike the structure with a consistent force. 

These are things that take practice but are sometimes not easy for everyone. It takes focus to get good at it, so when turning the focus away after each impact to verify the data, it makes things even more difficult and inefficient. With the new alternative impact method, based on time recording and then post processing, the focus can stay fixed on each of these operations, making things not only more efficient, but help improve data quality also.

The following is an example of such a situation. The modal impact hammer and test object are shown in Figure 14.
 

Figure 14: Modal Impact Test setup: impact hammer and test object.
 

The test setup parameters and impact selection results are summarized below. A single measurement run was post processed in two different ways:

The result is two FRF sets referenced to two different force levels.  This can help understand the test setup and structure better.  For example, if the FRFs changed drastically from the 20 Newton force level to the 30 Newton force level, it would be an indication that the structure is not being tested in its linear range.

The graphical results of the Impact Post Processing for 20 Newton force level are shown in Figure 15.
 

Figure 15: Results in Simcenter Testlab Impact Post Processing shows the difference in the coherence that is possible by including or excluding a specific hit. In this case, highlighting the fifth impact (left side in the table) shows the coherence would degrade (red curve on lower right) versus the baseline (black curve on lower right)
 

In the lower right, a delta coherence function is shown.  It indicates that if the fifth hit (highlighted in blue on the left side table) was to be included in the average, the coherence would degrade (red curve) versus the impact hits that were automatically selected (black curve) for inclusion in the average.

While this post processing method shifts the focus of data validation during measurement to post-processing, it should be remembered that good quality time data needs to be recorded for it to be useful. For example, during acquisition it needs to be made sure that there are no loose accelerometers, no overloads, etc. To be on the safe side, record a higher number of impacts than needed.  The excess impacts can be rejected during post-processing.

5.  Conclusions

The alternative post processing method offers two areas of improvement: