Published research / software

Perforated-beam inverse design.

Search from a performance target towards possible parent sections.

Supplied comparison of numerical response and reference-test evidence for perforated beams. Reference tests are not attributed to Sina. Publication · 10.1016/j.engappai.2025.113275

Role: Research co-author · credited developer
Period: 2026
Methods: Multi-output regression
Status: Published research / software

Engineering software / Streamlit + desktop

A defined input.
An inspectable decision.

For steel-beam designers

Search from a performance target towards possible parent sections.

01 / INPUT

Set the engineering case

Target web-post resistance and opening descriptors

02 / OUTPUT

Inspect the result

Parent-section dimensions and material yield strength

03 / DECISION

Compare the next options

Shortlist section/material combinations for detailed design checks.

Technical basis

Multi-output regression

Validation evidence

Published inverse-model comparisons and personally credited interfaces.

Current scope

Inverse candidates need geometric, fabrication and independent resistance checks.

Public deployment has not been verified. The supplied interface and source establish the implementation shown.

Engineering record

Scope, method and evidence.

Published predictive research and credited application. The folder name does not establish neural-network implementation; all designs still require verification.

01Engineering question

Search from a performance target towards possible parent sections.

Context: Structural steel.

02Role and documented scope

Research co-author · credited developer

03Methods and evidence

Multi-output regression links performance/opening descriptors to parent-section depth, flange width, web/flange thickness and yield strength. Model comparisons, interpretation plots and a credited GUI support that inverse task. The MORNN folder label is not used as proof of a neural-network implementation.

Supplied research files support the scope shown. Authorship is distinguished from personal ownership of every dataset, test or model; an interface is not proof of current deployment.

Publication record · 10.1016/j.engappai.2025.113275

04Interpretation and limitations

No universal accuracy, engineering approval or field benefit is inferred from a selected output. Model domains and independent checks must be established for a new assignment.

Research supporting this project