Get modal analysis right on your very first day with AI assistance

AI and four decades of expertise reliably automate Simcenter Testlab Neo’s modal analysis workflow

We recently announced that Simcenter Testlab Neo is introducing an entirely new modal analysis workflow, enhanced with artificial intelligence (AI) and a new modal dashboard, to significantly accelerate structural dynamics testing and improve its efficiency and accuracy.

We’re now set to take a closer look at how our new AI-assisted modal analysis automates and speeds your analysis workflow up to 7 times faster and enables novice and intermediate users to seamlessly achieve results typically reached by experts at the end of the analysis, with a remarkable 97.8% agreement and 0% contradiction!

Let’s first raise a toast to the beat of progress, listen to the strike of the hammer, and the excitement of what’s just around the corner in Simcenter Testlab Neo new release 2506!

AI-assisted modal analysis: what’s under the hood?

With the measured FRFs at hand, an experimental modal analysis starts with a detailed evaluation of the stabilization diagram and validation of the selected modes. Such analysis primarily includes the following stages:

1. Pole Selection: Selecting poles from the stabilization diagram to build a comprehensive modal model that accurately fits the measured data
2. Mode Validation: Verifying that the calculated modes correspond to actual physical resonances of the structure.

The challenge lies in selecting a sufficient number of poles to accurately represent the system, without including too many, which could introduce spurious numerical poles caused by noisy data segments. When performed manually, this process is time-consuming and error-prone, especially as structural complexity increases. Automating this can significantly accelerate the process, provided it maintains high accuracy and efficiency.

Our new AI-assisted modal analysis helps engineers of all experience levels perform modal analysis as quickly and accurately as possible. It integrates two distinct strategies to handle pole selection and validation:

• Step 1: Machine Learning (ML)-based pole selection: It employs an unsupervised ML method called Density-Based Spatial Clustering of Applications with Noise (DBSCAN). DBSCAN organizes poles into meaningful clusters based on inherent patterns or similarities beyond regular distributions, while effectively identifying noisy points in the dataset.
• Step 2: knowledge-based mode validation: It applies a two-stage cluster verification based on domain knowledge of modal parameter selection. This enables the system to effectively distinguish between physical and numerical poles. The first stage examines representative poles within clusters to ensure that a single mode is not fragmented into multiple clusters. The second stage relies on a novel metric to analyze the impact of each calculated mode on the final ’mode set’ and the FRF synthesis to identify and eliminate potential numerical modes.

More details on the methodology, and how it was integrated and validated can be found in [1]. An overview of the new AI-assisted modal analysis is depicted in figure 1 below:

Efficiency, accuracy and consistency of the new AI-assisted modal analysis

To evaluate the performance of the new AI-assisted modal analysis, we conducted a jury test involving 24 engineers: 8 novice users (with general knowledge), 8 intermediate users (with occasional hands-on experience), and 8 experts (with deep expertise in modal analysis). Eight diverse, industry-relevant datasets, ranging in complexity from level 2 to 5 on a 5-point scale, were used. This ensured a thorough assessment of the method’s versatility and reliability across a broad spectrum of scenarios, confirming its applicability under varying experimental conditions [1]. Let us share our conclusions derived from these comprehensive and detailed investigations:

1. Time saved: On average, users spent approximately 1 hour and 30 minutes to complete the modal analysis across all 8 datasets, starting from the calculated stabilization diagrams. This equates to approximately 15 minutes per dataset, measured from the selection of the FRFs to the final analysis result. This number of course changes with the complexity of the dataset.  With the new AI-assisted modal analysis in Simcenter Testlab Neo, this time was reduced to just under 2 minutes primarily for selecting the FRFs and generating the stabilization diagram, as the actual modal calculations are completed in a fraction of a second. The analysis time is also nearly irrespective of the complexity of the dataset. This results in up to 7 times faster total analysis time, with the most time-consuming step, i.e., modal analysis, is now performed automatically and almost instantly.
   
   
2. Accuracy and consistency achieved: To evaluate performance, we benchmarked the new AI-assisted method in Simcenter Testlab Neo against the standard Automatic Modal Pole Selection (AMPS) approach previously integrated into Simcenter Testlab. Additionally, results from novice and intermediate users were included for comparison. The evaluation focused on the total number of modes selected, relative to those identified by domain experts. As shown in Figure 2, the results clearly demonstrate that:
   1. Both AMPS and the new AI-assisted modal analysis achieved results very close to expert performance
   2. While AMPS tended to select more modes overall (than what experts selected on average), the AI-assisted method was more accurate, avoiding the selection of modes not confirmed by experts. Notably, every mode selected by the AI-assisted method was also selected by the experts, which indicates that this new method emulates expert behavior more accurately than the AMPS approach.

Here’s what our investigations reveal: the new AI-assisted method automatically delivers accurate and consistent results, matching expert selections with 97.8% agreement and 0% contradiction. Remarkably, and thanks to the automation, these results are achieved 7 times faster than manual analysis. This performance clearly demonstrates the potential of AI, further reinforced by automatic result verification grounded in four decades of proven domain expertise.

Whether you’re working with your own FRF datasets or looking to take advantage of CAD-based instrumentation and measurement with standout features like Smart Hit Selection, it’s time to bring Simcenter Testlab Neo into your workflow and unlock its powerful testing capabilities. Don’t wait any further, just visit Support Center and download the official 2506 release to witness the progress yourself.  We’re committed to making structural dynamics testing faster, smarter and more efficient!

[1] André Tavares, Emilio Di Lorenzo, Bram Cornelis, Simone Manzato, Bart Peeters, Wim Desmet, Konstantinos Gryllias, “Automated experimental modal analysis with density-based clustering: A benchmark validation through jury testing” Mechanical Systems and Signal Processing, Volume 232, 2025, 112714, ISSN 0888-3270.