AC 20-146A CHG 1 8 COMPUTER MODEL VALIDATION.
8.1 Overview.
8.1.1 The validation process determines the degree to which a model is an accurate representation of corresponding physical experiments from the perspective of the intended uses of the model (refer to ASME V&V 10-2019). The ability of the model to represent a physical phenomenon is evaluated by comparing the model predictions with physical test data. This process relies on high-quality test data and a quantitative comparison of test and simulation results. Validation is not an iterative process of refining the model to reduce discrepancies between the experimental systems and model outcomes. Refining the model is called calibration; it does not provide confidence that the model can accurately predict the system’s response to untested conditions.
8.1.2 This AC provides guidance on the numerous parameters that deserve consideration when comparing the results of numerical analysis to actual test data. However, in performing the validation of a numerical model, there is no substitute for good engineering judgment. An applicant should not use the guidance in this AC as a substitute for communicating and coordinating with the applicable certification branch when validating a numerical model. Sections 9 and 10 of this AC define the conditions where an applicant may use the results from computer analysis for certification purposes.
8.1.3 Any virtual anthropomorphic test device (v-ATD) used in the analysis should simulate the same type ATD that was used in the physical testing, which is defined in § 2X.562, TSO-C127x, or FAA policy. The applicant should demonstrate that the v-ATD is calibrated to behave similarly to the physical ATD or that it meets the guidance in SAE ARP 5765, revision A or later. Before any of the validation criteria are calculated, check the occupant simulation software for the reference location of all sensors of the v- ATD compared to the corresponding locations of the sensors on the ATD.
8.2 General Validation Acceptance Criteria.
8.2.1 The model should be validated against the original dynamic tests in accordance with § 2X.562(b) or TSO-C127x. The FAA will not accept a model validated against another validated model.
8.2.2 Model extrapolation should be limited to conditions that are similar to the model validation conditions. Occupant trajectory and head injury criteria detailed in appendix A of this AC. The current seat analysis should be similar to the test and analysis used to validate the numerical model, including loading conditions, seat type, and worst-case conditions. For example, the applicant should not use test results from a four-legged seat to validate a three-legged seat model. As another example, a model validated against longitudinal test data should not be used to evaluate vertical test conditions.
8.2.3 Conservative simulation results are encouraged, but not required. In the context of simulation, conservatism relates to the model over-predicting a particular response or 10
AC 20-146A CHG 1 under-predicting a particular strength. For example, if in the test, the calculated HIC was 700, but the model denoted 725, then the simulation is conservative.
8.2.4 The level of correlation required should not be more stringent than the level of accuracy of the test data, which is dependent on the test instrumentation.
8.2.5 The FAA considers the computer model validated if an acceptable agreement can be shown between the analysis and test data for those parameters critical to the application of the model. The calculation methods are detailed in appendix B of this AC. Test data used to validate the model should be included as an appendix in the validation and analysis report (VAR), refer to section 12 of this AC.
8.2.6 When precise occupant trajectory information is not required, a visual comparison of the event can be sufficient.
8.2.7 In addition to the general validation criteria above, the applicant might need to validate the model to some or all of the application specific criteria defined in paragraph 8.3 of this AC to ensure that the design requirements of the computer model are correct.
8.3 Application-Specific Validation Criteria.
The applicant should validate relevant parameters to the application of the model. The applicant and the certification branch should identify and agree on the validation criteria specific to the application, and the certification plan should list those criteria. The applicant and the certification branch should negotiate any additional validation criteria not listed in this AC. The following paragraphs provide guidance on the validation parameters to consider and suggested correlation levels. However, the final levels should be coordinated with the certification branch prior to the initiation of any program. If no acceptable rationale is available to make this determination, then the details listed in the following paragraphs may be followed:
8.3.1 Occupant Trajectory.
8.3.1.1 The general occupant trajectory should correlate to the test data. The trajectory should be specified by time history plots under SAE J211-2.
Occupant trajectory describes the overall translational and rotational motion of the occupant. SAE AS8049 defines use of the seat reference point (SRP) as the datum to determine occupant trajectory or position. The trajectory of the occupant can include head path, pelvic displacement, or torso displacement. When the pelvis is not visible, the knee can be used as a surrogate. If there is a concern regarding femur injury, occupant trajectory can include leg motion as well.
AC 20-146 CHG 1
8.3.1.2 Head path trajectory can be, in and of itself, a validation item. For example, if the applicant conducted a validation effort to support a claim that no head contact occurs, head path is a unique validation item. It can also be used to support another parameter, such as HIC.
8.3.1.3 The numerically derived occupant trajectory should be compared to high-speed video obtained from dynamic tests. The ability of the computer model to predict an occupant trajectory can be established by comparing time-history plots, for example, x-position versus time, and z-position versus time to calibrated photometric data obtained from the baseline dynamic test. Acceptable limits for model validation may be based on criteria established between the applicant and the FAA.
8.3.1.4 Quantitative evaluation of extremity flail is typically not required, specifically for forward- and aft-facing seats. However, arm and leg interaction with other seats or monuments can have a significant effect on the motion of the head and pelvis and, therefore, should not be completely ignored.
8.3.1.5 For side-facing seats, the applicant can use computer modeling to demonstrate that only incidental body-to-body contact will occur when the occupants are exposed to the accelerations and velocities pursuant to § 2X.562. This assumes that the seat structure and occupant motion have been validated to a baseline side-facing seat test.
8.3.1.6 Appendix A of this AC illustrates some of the items to consider when evaluating occupant trajectory and HIC. Although appendix A is an example of those items to consider in this type of validation, it is not a universal example. Other situations might require more or less stringent validation efforts.
8.3.2 Structural Response.
Quantifying the structural response of the computer model includes evaluating the internal loads and structural deformations of the seat. Validation of the computer model should include a comparison of the structural performance criteria.
8.3.2.1 Internal Loads.
8.3.2.1.1 The applicant and the certification branch should work together to establish what floor reaction loads, if any, are critical to the application of the analysis. Correlation of the floor reaction loads demonstrates a properly modeled load path from the occupant to the restraint system to the floor. The peak critical floor reaction loads between the analysis and test data should correlate to within 10 percent unless a different level of correlation was determined prior to the initiation of any program. In addition, the applicant should provide data showing that the critical floor 12
AC 20-146A CHG 1 load reactions correlate to the test results as described in appendix B of this AC.
8.3.2.1.2 There may be times when the applicant introduces a unique design for a primary load path member. The applicant may choose to install test instrumentation to monitor the internal loads or strains on this member.
This instrumentation is not required for certification but may be useful in validating the numerical model. In this instance, the applicant and the certification branch should work together to determine how or if these loads will be used to validate the model.
8.3.2.2 Structural Deformation.
8.3.2.2.1 A comparison of the planar space plots of structural deformation obtained from the analysis and photometric data obtained from dynamic test can help validate the model. Acceptable agreement should be obtained between the shape and magnitude, within 10 percent as described in appendix B of this AC unless a different level of correlation was accepted prior to the initiation of any program, for members that are critical to the overall performance or structural integrity of the seat or seating system.
Not all safety margins or all modes of failure need to be examined, only those the certification branch and applicant believe will be critical. The applicant can use visual comparisons between the test and analytical data for non-critical structural members.
8.3.2.2.2 If analysis of dynamic tests requires the determination of post-test permanent deformation such as that required for a structural test, then the analysis should differ by no more than 10 percent, unless a different level of correlation was determined prior to the initiation of any program, from the test and should not exceed any allowable deformations.
8.3.2.3 Failure Modes.
Failure modes of components may be validated through full-scale dynamic tests or component tests following a building block approach. See figure 2 of this AC. Modeling of tests that show structural failure can provide confidence in the model’s ability to capture relevant failure modes. If the model is only validated against tests with no failure or permanent deformation, there is little confidence that the model is capable of predicting failure. The conceptual framework of component and subcomponent level tests typically included in the building block approach can be adapted to the seat structure. The large quantity of tests needed to provide a statistical basis comes from the lowest levels, coupons, and elements (which are used to calibrate the numerical model), and the performance of structural details are validated in a lesser number of subcomponent and component tests. Detail and subcomponent tests may be used to validate the ability of analysis methods to predict local strains and failure modes. Added statistical considerations—for example, 13
repetitive point design testing and/or component overload factors to cover material and process variability—will be needed when analysis validation is not achieved. The static strength substantiation program should also consider all critical loading conditions for all critical structure. Refer to AC 20-107B, CHG 1 for further information.
Figure 2. Schematic Diagram of Building Block Approach
8.3.2.4 Joints and Fittings.
Joints and fittings are typically highly loaded seat structural elements. In general, they possess indeterminate load paths and contact (free-play) nonlinearities and might be difficult to model mathematically. Changes in the load path or material properties of these elements can affect the structural integrity and performance of the seat. Therefore, these parts of a seat structure should be modeled with care, paying particular attention to all possible failure mechanisms. Some changes to fittings and joints are acceptable to substantiate with modeling. However, significant changes to the material or mechanism of load transfer require tests.
AC 20-146A CHG 1
8.3.3 Restraint System.
8.3.3.1 With few exceptions, such as aft-facing seats and side-facing seats bounded by a wall or divider, the restraint system contributes significantly to the retention of the occupants and is part of the primary load path from the occupant to the seat. In this case, the test-simulation correlation should be evaluated for the restraint loading time history and the maximum value.
Maximum values that correlate to within 10 percent, unless a different level of correlation was determined prior to the initiation of any program, will ensure the computer model will predict the inertial force transfer from the occupant to the seat. For a seating configuration with a shoulder belt, conservatism is not always straightforward. Over-estimation of the shoulder belt load can lead to under-estimation of the ATD head x-motion and vice versa.
8.3.3.2 Additional parameters, such as belt payout or permanent elongation, when present, are not required for seat certification, but should be taken into account in order to provide additional confidence that the model is capturing the pertinent physics. Another factor to consider is any belt tightening that can occur after an inertial reel lock.
8.3.3.3 Belt payout is a term used to describe how much of the shoulder harness restraint is released before locking of the inertia reel. In a sudden deceleration, it is unlikely the shoulder harness instantly locks in place.
There is a finite length of time where the harness is free to release from the reel. The amount of restraint released from the reel is the belt payout.
8.3.3.4 Occupant trajectory and restraint system loads are closely related functions. The necessity of validating the restraint system loads does not negate the necessity of validating the occupant trajectory in situations where the restraint system loads are required. It is not acceptable to show compliance to occupant trajectory instead of restraint system performance or to validate restraint system performance at the expense of occupant trajectory.
8.3.4 Spine Load.
8.3.4.1 Section 2X.562 and TSO-C127x defines the certification requirements for lumbar spinal loading. The maximum allowable load is 1,500 pounds (lbs) of compression. The computed spine load should be correlated with the test data when appropriate. This includes, but is not limited to, the initial validation of a test scenario or situations where a design change could affect this parameter such as a seat cushion change. The spine load time-history and maximum spine load obtained in the analysis should correlate to within 10 percent of the dynamic test data unless a different level of correlation was determined prior to the initiation of any program and the numerical spine load should not exceed 1,500 lbs.
AC 20-146A CHG 1
8.3.4.2 To account for the testing uncertainty, conservatism can be incorporated into validation and model use via a factor of safety. Repeated testing of seat cushions shows a typical test variability in lumbar load of about ±125 lbs, which gives a data spread of 250 lbs in lumbar load when testing parameters are tightly controlled. Assuming uncertainty is normally distributed, this data spread gives a standard deviation of approximately 42 lbs Based on this standard deviation, there is a 95 percent confidence that the true load is below the regulatory limit of 1,500 lbs, if the measured or simulated load is no greater than 1,430 lbs.
Note: TSO-C127x references SAE AS8049 and SAE ARP 5526C as the source of minimum performance standards (MPS) for seating systems.
SAE ARP5765A defines a means of assessing the credibility of computer models of aircraft systems.
8.3.4.3 Setting the cumulative probability of 95 percent equal to 1,500 lbs for normally distributed data using a standard deviation of 42 lbs results in a mean value of 1,430 lbs for the distribution. Therefore, only seat configurations with dynamic test data that yield spine loads below 1,430 lbs should be used for validation. Likewise, for model use, it is recommended that only models that produce a lumbar load below 1,430 lbs be used.
8.3.4.4 In the validation phase, models can exceed 1,430 lbs. In situations where the numerical spine load under-predicts the dynamic test data, the numerical spine load should not exceed 1,430 lbs minus the magnitude of the under-prediction for the Hybrid II; there is no adjustment made for under-prediction for the FAA Hybrid III.
8.3.4.5 Table 1 below provides an example of validation under-prediction and validation over-prediction and the corresponding model use recommended limits. In the first case, the model under predicts by 50 lbs, so the model use is limited to 1,380 lbs (1,430 lbs minus 50 lbs) for the Hybrid II, and there is no adjustment made for the FAA Hybrid III. In the second case, the model over predicts and the model use is simply limited to 1,430 lbs.
AC 20-146A CHG 1 Table 1. Example Peak Lumbar Loads Example Peak Lumbar Loads Validation Hybrid II Model Use FAA Hybrid III Model Use Model under-predicts Test = 1,400 lbs, Model = 1,350 lbs Model = 1,380 lbs or less Model = 1,430 lbs or less Model over-predicts Test = 1,400 lbs, Model = 1,450 lbs Model = 1,430 lbs or less Model = 1,430 lbs or less
8.3.5 Head Injury Criteria (HIC).
8.3.5.1 Section 2X.562 or TSO-C127x defines the certification requirements for HIC. The applicant can use the results of computer modeling to show compliance with these requirements, within the limitations summarized below. However, an installation change that results in a significantly higher head strike velocity will likely require testing.
8.3.5.2 The regulation specifies calculating HIC during the duration of the head impact, with a maximum allowable HIC limit of 1000 units. The selected time interval should correspond to the duration of the major head impact on aircraft interior features.
8.3.5.3 The profile, that is, the shape and peak “g” of the acceleration time-history plot, as well as the average “g” loading for resultant head accelerations obtained in the analysis, should correlate to the results of the dynamic test.
The acceleration is measured at the head center of gravity for an ATD.
Occupant simulation software should be checked for the reference location of the head acceleration output. Sensor locations should be documented in the v-ATD calibration report.
8.3.5.4 Given two dynamic tests with the same desired deceleration profile, the maximum HIC values will likely vary. Therefore, a precise match between the test derived HIC and the analytical HIC is not realistic. However, the maximum analytical HIC value should correlate to within 100 HIC units of the maximum test derived HIC value unless a different level of correlation was determined prior to the initiation of any program. The FAA encourages generation of conservative HIC prediction models. One method to add conservatism to the process is to incorporate test uncertainty as a factor of safety in validation and model use. Using the same process as in paragraph 8.3.4.2 of this AC, and assuming a typical data spread of ±200 HIC units, the 95 percent confidence HIC value is 890 HIC units. Therefore, the FAA recommends that only seat configurations with dynamic test data that produce a HIC value below 890 HIC units 17
AC 20-146A CHG 1
8.3.5.5
8.3.5.6 should be used for validation. Likewise, for model use, the FAA recommends that only models that produce a HIC value below 890 HIC units be used.
Note: Models can exceed 890 HIC units in the validation phase.
It is also unlikely that the analytical head deceleration time history function will perfectly match the test-generated head deceleration time history function. Therefore, the initial and final integration times, t1 and t2, as defined and used in § 2X.562 or TSO-C127x, will likely vary between test and analysis. These time values should differ by no more than 5 ms (milliseconds). In addition, the contact velocity should match within 10 percent and location of the contact should match.
Note: Regardless of the validation of the model and the specific configuration, the predicted HIC must always remain below 1,000 HIC units for the data to comply with the § 2X.562.
The following is not an exhaustive list, but the applicant may choose to use computer modeling under the following circumstances:
1. The predicted occupant head strike envelope will satisfy the above stated requirements by showing that no contact with adjacent seats, structure, or other items in the cabin will occur.
2. To evaluate a modified seat installation where the potential head impact surfaces are identical, only the geometric strike envelope has changed. Original HIC values exceeding 890 HIC units will typically not support numerical substantiation.
3. The applicant has performed dynamic testing in the presence of a rigid structure. The applicant would then reposition the seat in the aircraft where the head strike will be on a less rigid structure but with equivalent head strike velocities.
4. For scenarios that meet the guidelines in table 2 below. This includes situations where the head impact surface has changed.
Table 2. Example HIC Values Example HIC Values Validation Hybrid II Model Use FAA Hybrid III Model Use Model under-predicts (in HIC units) Test = 850, Model = 800 Model = 840 or less Model = 890 or less Model over-predicts (in HIC units) Test = 850, Model = 900 Model = 890 or less Model = 890 or less 18
AC 20-146A CHG 1
8.3.6 Femur Compressive Load (Part 25 Airplanes Only).
Section 25.562(c)(6) or TSO-C127x defines the certification requirements for axial compressive loading of the femur. The maximum allowable limit is 2,250 lbs. The femur compressive load is usually not an issue in the testing of part 25 seats. However, if the certification branch or the applicant determines it should be evaluated, the load time-history profile for the compressive femur load obtained in the analysis should correlate to the dynamic test data. The peak load value, as determined by the analysis, should correlate to within 10 percent of the dynamic test results unless a different level of correlation was determined prior to the initiation of any program.
8.3.7 Non-Critical Criteria.
8.3.7.1 The applicant should validate parameters that are important to the particular application of the analysis. Depending on the purpose of the analysis, it may not be necessary to meet all validation criteria in this AC.
However, the FAA cautions that gross discrepancies in the model (such as unrealistic load paths or failure modes) may impact the ability of the model to predict parameters of interest. The following examples are offered to illustrate non-critical criteria for the specified model use condition:
8.3.7.1.1 Lumbar loads for the horizontal test, refer to § 2X.562(b)(2) or TSO-C127x, with a two-point restraint are usually not critical. It is unlikely the applicant will have to evaluate the ATD lumbar load in this test condition.
8.3.7.1.2 The upper torso restraint for a side-facing seat, where the occupant is adjacent to a structural barrier, typically carries small loads. It may be of little value for the applicant to correlate the analytical upper torso restraint loads to the test data.
8.3.7.1.3 For the horizontal test required by § 2X.562(b)(2) or TSO-C127x, lateral (y-axis) floor reaction loads are small compared to the vertical or horizontal reaction loads. It is not reasonable to expect the applicant to correlate the model to all three loads (that is, vertical, lateral, and horizontal) reported by each load cell. It is more reasonable to require validation for those loads critical to the application of the model.
8.3.7.1.4 Restraint loads in a § 25.562(b)(1) or TSO-C127x test are essentially zero throughout the critical portion of the test. It is unlikely the applicant will have to evaluate these loads.
8.3.7.2 These examples do not constitute an exhaustive list. They are simply meant to illustrate that engineering judgment and the particular application of the model should guide the applicant and the certification branch to the proper validation criteria.
AC 20-146A CHG 1
8.4
8.4.1
8.4.2
8.5 9
9.1
9.1.1
9.1.2
9.2 Discrepancies.
Failure to satisfy all validation criteria does not automatically preclude the model from being validated. The applicant and the certification branch engineer should evaluate whether the deviations impact the ability of the model to predict credible results and determine if deviations from the validation criteria are acceptable.
In addition, the applicant may present evidence to show the deviation is within the inherent reliability and statistical accuracy of the test measurements. The applicant should quantify any discrepancies between the results obtained from analysis and the dynamic test data for those parameters that are critical to the application of the analysis.
Computer Hardware and Software.
Any analysis should be conducted on a platform, which has been shown to be reliable and acceptable. Certification data produced by a computer model should be performed on the same hardware, compute environment (operating system, auxiliary software, components, etc.), and software version on which the validation was conducted. If a different software version, hardware platform, and/or compute environment is used, the applicant should revalidate the model using the new configuration.
APPLICATION OF COMPUTER MODELING IN SUPPORT OF DYNAMIC TESTING.
General.
There will be occasions when the applicant wants to determine the critical loading scenario for a particular seating system. This paragraph provides guidance on those items to consider when the applicant performs trade studies to identify the most critical configuration/installation. A final certification test is required to certify the critical configuration/installation to the requirements of § 2X.562 or TSO-C127x.
Paragraphs 9.2 through 9.4 of this AC specify the conditions when a computer model can be used to provide engineering analysis and rationale in support of dynamic testing.
These conditions do not form an exhaustive list of items to consider, but they are the most common.
Determination of Worst-Case Scenario for a Seat Design.
After the computer analysis is complete, the results from the simulation can be used to determine the worst-case or critical loading scenario for a particular seating system.
This may include the following:
• Identifying components of seat structure that are critically loaded.
• The selection of the critical seat tracking positions, such as seat adjustment positions.
• An evaluation of the restraint system, such as critical attachment location.
Note: The restraint system is not limited to the actual belts; it also includes the required anchoring attachments. Computer modeling may be used to analyze the effect of anchoring the restraints at different locations on the seat frame or in the aircraft. The restraint system would also include any inflatable devices used for restraint or occupant protection. This AC does not address the issues that may be necessary to correctly model and validate their operation; however, the general principles discussed here apply.
• An evaluation of the yaw condition to address loading on the seat frame and movement of the occupant out of the restraint system.
• The number of seat places occupied.
• The selection of the worst-case seat cushion build-up.
9.3 Determination of Worst-Case Scenario for Seat Installation.
Results of a validated computer model may be used to select the worst-case seat system installation as a candidate for dynamic testing. In determining the most critical seat installation, each seating system should be analyzed in its production installation configuration. For example, an analysis to determine a worst-case seating system may include seating systems installed at different positions in the fuselage, which will result in various restraint anchor positions relative to the occupant and seat structure.
9.4 Determination of Occupant Strike Envelope.
The results of the computer analysis can be used to determine the occupant strike envelope with aircraft interior components. Each seating system should be analyzed in its production installation configuration. The occupant strike envelope will determine if a potential for head strike exists and, if so, which items are required in the test setup during the HIC evaluation tests.
10 APPLICATION OF COMPUTER MODELING INSTEAD OF DYNAMIC TESTING.
There will be occasions when the applicant wants to certify a seat that is based on a certificated design concept, such as a family seat design, that differs from the certificated design. If the applicant intends to use the results of computer modeling to provide engineering/certification data instead of dynamic testing for a modified design, then the results from this validated model can be applied to the modifications specified in paragraphs 10.1 and 10.2 of this AC.
10.1 Seat System Modification.
10.1.1 Analysis based on a validated computer simulation may be used to substantiate seat designs or installations that have been modified from a certificated configuration. These modifications may include changes to primary and non-primary load path structural members.