Multi-Shaker Shock Control: Ideas That Did Not Work in Simulation, and Why

A shock test with several shakers is controlled from a model of the test article measured at a low level. Bolted joints soften as the level rises, so a drive computed from that model can over-test the article on the way to full level. This article reports, with numbers, the control ideas tried in simulation against that problem that did not work. A neural network, as the law's model or as the setter of how much softening to allow for, never beat a plain law built on a family of models, and where its prediction was low it put the article 5 to 8 dB above the full-level target. A modal model that changes inside each pulse passed seven of the eight structures it was developed on and failed held-out structures with modeled bolted joints that the plain law passes. A corrector that steps all controls together was not the main reason its law passed, and it fired full-level pulses 6 to 9 dB over target between the pulses that a judgment of the first, third and last full-level pulses looks at. A step rule did not replace the limit on softening, and two cures for failures under heavy measurement noise were worse, or only partly better, on the strict criteria. Twelve further ideas are listed with their evidence. The failures share causes: low-level pulses do not show the softening that matters, a law that fires a predicted drive is only as safe as its lowest prediction, and judging some pulses hides over-tests. Simulation only; nothing here has been run on shakers. Hertzforge Journal 1, 2026-004, version 1. The version of record is the article page at https://hertzforge.com/journal/2026-004/v1/. This deposit holds that page, the article in one file (it opens without a network connection), the PDF print copy, and the bundle of the data behind every figure and the code that draws it. The article is reviewed in public after publication; its reviews, its review status and any later version are at https://hertzforge.com/journal/2026-004/.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-15
DOI
https://doi.org/10.5281/zenodo.24254181
Primary Topic
Mechanical Engineering Research and Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Multi-Shaker Shock Control: Ideas That Did Not Work in Simulation, and Why

Kevin Cross
Zenodo (CERN European Organization for Nuclear Research)
Mechanical Engineering Research and Applications
article

Multi-Shaker Shock Control: Ideas That Did Not Work in Simulation, and Why

Kevin Cross
article en

Abstract

A shock test with several shakers is controlled from a model of the test article measured at a low level. Bolted joints soften as the level rises, so a drive computed from that model can over-test the article on the way to full level. This article reports, with numbers, the control ideas tried in simulation against that problem that did not work. A neural network, as the law's model or as the setter of how much softening to allow for, never beat a plain law built on a family of models, and where its prediction was low it put the article 5 to 8 dB above the full-level target. A modal model that changes inside each pulse passed seven of the eight structures it was developed on and failed held-out structures with modeled bolted joints that the plain law passes. A corrector that steps all controls together was not the main reason its law passed, and it fired full-level pulses 6 to 9 dB over target between the pulses that a judgment of the first, third and last full-level pulses looks at. A step rule did not replace the limit on softening, and two cures for failures under heavy measurement noise were worse, or only partly better, on the strict criteria. Twelve further ideas are listed with their evidence. The failures share causes: low-level pulses do not show the softening that matters, a law that fires a predicted drive is only as safe as its lowest prediction, and judging some pulses hides over-tests. Simulation only; nothing here has been run on shakers. Hertzforge Journal 1, 2026-004, version 1. The version of record is the article page at https://hertzforge.com/journal/2026-004/v1/. This deposit holds that page, the article in one file (it opens without a network connection), the PDF print copy, and the bundle of the data behind every figure and the code that draws it. The article is reviewed in public after publication; its reviews, its review status and any later version are at https://hertzforge.com/journal/2026-004/.

Zenodo (CERN European Organization for Nuclear Research)
Mechanical Engineering Research and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon