BAM–IWM PyIron Knowledge Graph
Integration Workflow

This tutorial documents the demo_gui_use_case_trell.ipynb notebook, which implements a complete end-to-end pipeline:

Required packages

Step-by-step setup

# Use Miniforge Prompt or CMD (not PowerShell)
conda create -n pyiron_env python=3.12
conda activate pyiron_env
pip install pyiron-workflow pyironflow openpyxl requests rdflib pandas

conda activate pyiron_env
jupyter lab "C:\path\to\Pyiron_Windows_06_10_2026\demo_gui_use_case_trell.ipynb"

The SetupProject node prepares all simulation inputs and pre-processes experimental data before the optimizer can run.

Notebook cell

import importlib
import setup_project_nodes as spn
from pyiron_workflow import Workflow
from pyironflow import PyironFlow

wf = Workflow("setup", delete_existing_savefiles=True)
wf.Setup = spn.SetupProject(
    project_dir = 'C:/path/to/Pyiron_Windows_06_10_2026'
)
wf.run()  # or: pf = PyironFlow([wf]); pf.gui  for visual modePYTHON

Three-stage parameter identification strategy: a global search finds a good basin, then two local refinement stages converge to the optimum.

Key configuration in Main.txt

MODEL_NAME  = SDCHABOX
MATERIAL_ID = IN718
SIMUL_PATH  = C:/path/to/simulop.exeMAIN.TXT

Key configuration in Input.py

# Set material and model — determines output filenames
material_id = 'IN718'
model_name  = 'SDCHABOX'

directories = Directories(
    parameterFile    = f'{material_id}_{model_name}.mat',
    bestParameterFile= f'{material_id}_{model_name}.mat.best',
    templateFile     = f'{model_name}.mat.tmpl',
    ... # optimizer stages, log files, etc.
)PYTHON

Runs independent validation against held-out experimental data (not used during calibration) to confirm the calibrated parameters generalise.

Notebook cell

from validation_node import run_validation_workflow
import os; os.chdir("Validation/")

results = run_validation_workflow("Validation.txt", verbose=True)
# Returns dict with: num_tests, num_passed, report path, metricsPYTHON

Configure Validation/Validation.txt

MATERIAL_ID             = IN718
MODEL_NAME              = SDCHABOX
OPTIMIZED_PARAMS_FILE   = ../sim/IN718_SDCHABOX.mat.best
TEMPLATE_FILE           = ../sim/SDCHABOX.mat.tmpl
SIMUL_PATH              = C:/path/to/simulop.exeVALIDATION.TXT

A 7-node PyIron workflow that uploads all artefacts to CKAN first (to obtain stable URLs), then generates the knowledge-graph Excel template with URLs embedded, converts to Turtle RDF, and uploads the final TTL.

Notebook cell

project_dir  = 'C:/path/to/Pyiron_Windows_06_10_2026'
MATERIAL_ID  = 'IN718'
MODEL_NAME   = 'SDCHABOX'

wf_kg = Workflow("kg_generation")

# STEP 1 — Zip experimental data
wf_kg.ZipExp = zipn.ZipExperimentalData(project_dir=project_dir)

# STEP 2 — Upload zip to CKAN → URL
wf_kg.UploadZip = ckan.UploadToCKAN(
    ttl_file    = wf_kg.ZipExp.outputs["str(zip_path)"],
    description = "Experimental creep test data for calibration"
)

# STEP 3 — Upload material card → URL
wf_kg.UploadMatCard = ckan.UploadToCKAN(
    ttl_file = f"{project_dir}/sim/{MATERIAL_ID}_{MODEL_NAME}.mat.best"
)

# STEP 4 — Upload validation report → URL
wf_kg.UploadReport = ckan.UploadToCKAN(
    ttl_file = f"{project_dir}/Validation/results/{MATERIAL_ID}_{MODEL_NAME}_validation_report.txt"
)

# STEP 5 — Generate Excel KG template with URLs embedded
wf_kg.Excel = aen.PopulateExcel(
    project_dir               = project_dir,
    exp_container_url         = wf_kg.UploadZip.outputs["resource_url"],
    material_card_url         = wf_kg.UploadMatCard.outputs["resource_url"],
    material_card_filename    = f'{MATERIAL_ID}_{MODEL_NAME}.mat.best',
    validation_report_url     = wf_kg.UploadReport.outputs["resource_url"],
    validation_report_filename= f'{MATERIAL_ID}_{MODEL_NAME}_validation_report.txt'
)

# STEP 6 — Convert Excel to Turtle RDF
wf_kg.TTL = ettn.ExcelToTTL(
    excel_file = wf_kg.Excel.outputs["str(output_excel)"]
)

# STEP 7 — Upload TTL to CKAN
wf_kg.UploadTTL = ckan.UploadToCKAN(
    ttl_file = wf_kg.TTL.outputs["str(ttl_path)"]
)

wf_kg.run()PYTHON

Queries the Fuseki-backed knowledge graph to retrieve all calibrated SDCHABOX parameters for a material and exports them as a pivot-table CSV.

Notebook cell

MATERIAL_NAME = "IN718"

wf_sparql = Workflow("material_card_query")
wf_sparql.Query = sqn.QueryMaterialCard(
    material_name = MATERIAL_NAME,
    project_dir   = PROJECT_DIR,
)
wf_sparql.run()

# Output channel name = the return expression string
csv_path = wf_sparql.Query.outputs["str(output_csv)"].valuePYTHON

Example output (pivot table)

SPARQL endpoint

https://dataportal.material-digital.de/dataset/bam-sdchabox-workflow-kg/fuseki/$/sparql

# Pattern: {ckan_base}/dataset/{dataset_slug}/fuseki/$/sparql
# Method:  POST  (Content-Type: application/x-www-form-urlencoded)
# Auth:    Authorization: <raw_jwt>   (no "Bearer " prefix)SPARQL ENDPOINT

Generates a self-contained, browser-ready HTML Material Card Pass by running 6 SPARQL queries against the Fuseki knowledge graph. The result is published to CKAN and the download URL is returned.

Notebook cell

IDENTIFIER  = "IN718"  # material name OR full MaterialCard ID
PROJECT_DIR = 'C:/path/to/Pyiron_Windows_06_10_2026'

wf_mcp = Workflow("material_card_mcp")
wf_mcp.MCP = mcp.GenerateMaterialCardMCP(
    identifier  = IDENTIFIER,
    project_dir = PROJECT_DIR,
)
wf_mcp.run()

ckan_url = wf_mcp.MCP.outputs["str(resource_url)"].value
print(f"MCP published: {ckan_url}")PYTHON

MCP HTML sections

All output filenames are generated from MATERIAL_ID and MODEL_NAME. Changing the material or model only requires updating these values in Main.txt and Input.py — no hardcoded path edits elsewhere.

Where to change identifiers