Upload 5 files

Initial upload: IPA pharma compactor platform dataset v1.0

IPA_Pharma_Compactor_Platform.ipynb

@@ -0,0 +1,403 @@
+{
+ "cells": [
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "# \ud83c\udfed IPA Pharmaceutical Roller Compactor Platform: Scale-Up & Performance\n",
+    "\n",
+    "**Dataset:** IPA Pharma Compactor (Synthetic) v1.0  \n",
+    "**Publisher:** [Innovative Process Applications](https://www.innovativeprocess.com) | Crestwood, IL  \n",
+    "**License:** CC BY 4.0\n",
+    "\n",
+    "> \u26a0\ufe0f Synthetic educational data \u2014 not production measurements.\n",
+    "\n",
+    "---\n",
+    "\n",
+    "### Analysis Roadmap\n",
+    "1. Platform overview & scale-up visualization\n",
+    "2. Twin feed screw response surface (VFS/HFS ratio optimization)\n",
+    "3. Interactive design space heatmaps\n",
+    "4. Scale-up consistency analysis\n",
+    "5. Energy efficiency & throughput Pareto frontier\n",
+    "6. Comprehensive radar chart & scorecard\n",
+    "7. Material fingerprinting across the CL platform"
+  ]},
+
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "import matplotlib.pyplot as plt\n",
+    "import matplotlib.gridspec as gridspec\n",
+    "from matplotlib.patches import FancyBboxPatch, Circle\n",
+    "from matplotlib.colors import LinearSegmentedColormap\n",
+    "import matplotlib.ticker as mticker\n",
+    "import seaborn as sns\n",
+    "from scipy.interpolate import griddata\n",
+    "\n",
+    "# === IPA Brand Palette ===\n",
+    "IPA_TEAL = '#008080'\n",
+    "IPA_NAVY = '#1B2A3B'\n",
+    "IPA_CHARCOAL = '#3D3D3D'\n",
+    "IPA_GOLD = '#C9A84C'\n",
+    "IPA_LIGHT = '#E8F5F5'\n",
+    "IPA_GRADIENT = LinearSegmentedColormap.from_list('ipa', [IPA_LIGHT, IPA_TEAL, IPA_NAVY])\n",
+    "IPA_DIVERGING = LinearSegmentedColormap.from_list('ipa_div', ['#cc4444', '#ffffff', IPA_TEAL])\n",
+    "MODEL_COLORS = {'CL25150': '#66c2a5', 'CL30200': '#3288bd',\n",
+    "                'CL50200': IPA_TEAL, 'CL75200': IPA_GOLD, 'CL100250': IPA_NAVY}\n",
+    "\n",
+    "plt.rcParams.update({\n",
+    "    'figure.dpi': 120, 'figure.facecolor': 'white',\n",
+    "    'axes.facecolor': '#fafafa', 'axes.edgecolor': '#cccccc',\n",
+    "    'axes.grid': True, 'grid.alpha': 0.3, 'grid.color': '#dddddd',\n",
+    "    'font.family': 'sans-serif', 'font.size': 10,\n",
+    "    'axes.titlesize': 13, 'axes.titleweight': 'bold',\n",
+    "})\n",
+    "\n",
+    "df = pd.read_csv('ipa_pharma_compactor_v1.0.csv')\n",
+    "print(f'\\u2714 {df.shape[0]:,} runs \\u00d7 {df.shape[1]} columns')\n",
+    "print(f'\\u2714 Models: {sorted(df.compactor_model.unique())}')\n",
+    "print(f'\\u2714 Materials: {sorted(df.material.unique())}')\n",
+    "df.head(3)"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## 1. Platform Overview: The IPA CL-Series Scale-Up Path\n",
+    "\n",
+    "A single visualization showing the complete CL platform from R&D to production,\n",
+    "with throughput ranges, roll geometry, and power."
+  ]},
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "fig = plt.figure(figsize=(16, 7))\n",
+    "gs = gridspec.GridSpec(2, 5, height_ratios=[3, 1], hspace=0.4, wspace=0.3)\n",
+    "\n",
+    "# Top: throughput by model (violin plots)\n",
+    "ax_main = fig.add_subplot(gs[0, :])\n",
+    "models_ordered = ['CL25150', 'CL30200', 'CL50200', 'CL75200', 'CL100250']\n",
+    "scales = ['R&D / Lab', 'Pilot', 'Pilot / Small Prod', 'Production', 'Full Production']\n",
+    "colors = [MODEL_COLORS[m] for m in models_ordered]\n",
+    "\n",
+    "parts = ax_main.violinplot(\n",
+    "    [df[df.compactor_model==m]['throughput_kg_hr'].values for m in models_ordered],\n",
+    "    positions=range(5), showmeans=True, showmedians=True\n",
+    ")\n",
+    "for i, pc in enumerate(parts['bodies']):\n",
+    "    pc.set_facecolor(colors[i])\n",
+    "    pc.set_alpha(0.7)\n",
+    "parts['cmeans'].set_color(IPA_NAVY)\n",
+    "parts['cmedians'].set_color('white')\n",
+    "\n",
+    "ax_main.set_xticks(range(5))\n",
+    "ax_main.set_xticklabels([f'{m}\\n{s}' for m, s in zip(models_ordered, scales)], fontsize=9)\n",
+    "ax_main.set_ylabel('Throughput (kg/hr)', fontsize=11)\n",
+    "ax_main.set_title('IPA CL-Series Platform: Throughput Scale-Up Path', fontsize=14, color=IPA_NAVY)\n",
+    "\n",
+    "# Add roll size annotations\n",
+    "roll_sizes = ['1\\u2033\\u00d76\\u2033', '1\\u2033\\u00d78\\u2033', '2\\u2033\\u00d78\\u2033', '3\\u2033\\u00d78\\u2033', '4\\u2033\\u00d710\\u2033']\n",
+    "powers = ['5 HP', '2.75 HP', '12 HP', '20 HP', '25 HP']\n",
+    "for i, (rs, pw) in enumerate(zip(roll_sizes, powers)):\n",
+    "    ymax = df[df.compactor_model==models_ordered[i]]['throughput_kg_hr'].max()\n",
+    "    ax_main.annotate(f'{rs}\\n{pw}', xy=(i, ymax), xytext=(i, ymax + 15),\n",
+    "                     ha='center', fontsize=8, color=IPA_CHARCOAL,\n",
+    "                     bbox=dict(boxstyle='round,pad=0.3', fc='white', ec=colors[i], alpha=0.9))\n",
+    "\n",
+    "# Bottom: mini bar charts for key specs per model\n",
+    "specs = ['max_roll_pressure_kn_cm', 'density_cv_pct', 'granule_yield_pct',\n",
+    "         'changeover_time_hr', 'specific_energy_kwh_tonne']\n",
+    "spec_labels = ['Max Pressure\\n(kN/cm)', 'Density CV\\n(%)', 'Granule Yield\\n(%)',\n",
+    "               'Changeover\\n(hours)', 'Spec. Energy\\n(kWh/t)']\n",
+    "\n",
+    "for j, (spec, slabel) in enumerate(zip(specs, spec_labels)):\n",
+    "    ax = fig.add_subplot(gs[1, j])\n",
+    "    vals = [df[df.compactor_model==m][spec].mean() for m in models_ordered]\n",
+    "    ax.bar(range(5), vals, color=colors, alpha=0.8, edgecolor='white')\n",
+    "    ax.set_xticks(range(5))\n",
+    "    ax.set_xticklabels([m.replace('CL','') for m in models_ordered], fontsize=7)\n",
+    "    ax.set_title(slabel, fontsize=8, pad=3)\n",
+    "    ax.tick_params(axis='y', labelsize=7)\n",
+    "\n",
+    "plt.suptitle('', y=0.98)\n",
+    "plt.tight_layout()\n",
+    "plt.show()"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## 2. Twin Feed Screw Response Surface: VFS/HFS Ratio Optimization\n",
+    "\n",
+    "IPA\u2019s twin screw design allows **independent** control of throughput (HFS) and\n",
+    "pre-compression (VFS). This response surface shows the optimal VFS/HFS ratio\n",
+    "for maximizing ribbon density while maintaining uniformity."
+  ]},
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "fig, axes = plt.subplots(1, 3, figsize=(18, 5))\n",
+    "\n",
+    "# Use CL50200 + MCC as representative case\n",
+    "subset = df[(df.compactor_model=='CL50200') & (df.material=='MCC_PH101')].copy()\n",
+    "\n",
+    "for ax, response, title, cmap in zip(\n",
+    "    axes,\n",
+    "    ['ribbon_rel_density', 'density_cv_pct', 'granule_yield_pct'],\n",
+    "    ['Ribbon Relative Density', 'Density CV% (lower=better)', 'Granule Yield %'],\n",
+    "    [IPA_GRADIENT, IPA_GRADIENT.reversed(), IPA_GRADIENT]\n",
+    "):\n",
+    "    x = subset['vfs_hfs_ratio'].values\n",
+    "    y = subset['roll_pressure_fraction'].values\n",
+    "    z = subset[response].values\n",
+    "\n",
+    "    xi = np.linspace(x.min(), x.max(), 50)\n",
+    "    yi = np.linspace(y.min(), y.max(), 50)\n",
+    "    xi, yi = np.meshgrid(xi, yi)\n",
+    "    zi = griddata((x, y), z, (xi, yi), method='cubic')\n",
+    "\n",
+    "    contour = ax.contourf(xi, yi, zi, levels=20, cmap=cmap, alpha=0.9)\n",
+    "    ax.contour(xi, yi, zi, levels=8, colors='white', linewidths=0.3, alpha=0.5)\n",
+    "    plt.colorbar(contour, ax=ax, shrink=0.8, pad=0.02)\n",
+    "\n",
+    "    ax.set_xlabel('VFS/HFS Ratio', fontsize=10)\n",
+    "    ax.set_ylabel('Roll Pressure Fraction', fontsize=10)\n",
+    "    ax.set_title(title, fontsize=11)\n",
+    "\n",
+    "    # Mark optimal zone\n",
+    "    ax.axvline(1.0, color=IPA_GOLD, linestyle='--', alpha=0.7, linewidth=1.5)\n",
+    "    ax.annotate('Optimal\\nVFS ratio', xy=(1.0, 0.9), fontsize=8,\n",
+    "                color=IPA_GOLD, fontweight='bold', ha='center')\n",
+    "\n",
+    "fig.suptitle('CL50200 + MCC PH-101: Twin Feed Screw Response Surface',\n",
+    "             fontsize=14, color=IPA_NAVY, y=1.02)\n",
+    "plt.tight_layout()\n",
+    "plt.show()\n",
+    "\n",
+    "print('The VFS/HFS ratio = 1.0 sweet spot is clearly visible across all three responses.')\n",
+    "print('This independent twin-screw control is a key IPA platform advantage.')"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## 3. Design Space Heatmap: Multi-Objective Optimization\n",
+    "\n",
+    "QbD asks: where in the process space do we simultaneously achieve\n",
+    "Zinchuk-window density, low fines, and high yield?"
+  ]},
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "fig, axes = plt.subplots(1, 5, figsize=(20, 4), sharey=True)\n",
+    "\n",
+    "for ax, model, color in zip(axes, models_ordered, colors):\n",
+    "    sub = df[df.compactor_model == model]\n",
+    "    in_spec = (\n",
+    "        (sub.ribbon_rel_density >= 0.60) &\n",
+    "        (sub.ribbon_rel_density <= 0.80) &\n",
+    "        (sub.fines_pct < 20) &\n",
+    "        (sub.granule_yield_pct > 70)\n",
+    "    )\n",
+    "    ax.scatter(sub.loc[~in_spec, 'scf_kn_cm'],\n",
+    "              sub.loc[~in_spec, 'ribbon_rel_density'],\n",
+    "              alpha=0.15, s=6, color='#cccccc')\n",
+    "    sc = ax.scatter(sub.loc[in_spec, 'scf_kn_cm'],\n",
+    "                    sub.loc[in_spec, 'ribbon_rel_density'],\n",
+    "                    alpha=0.6, s=12, c=sub.loc[in_spec, 'granule_yield_pct'],\n",
+    "                    cmap=IPA_GRADIENT, vmin=70, vmax=95)\n",
+    "    ax.axhline(0.60, color='#cc4444', linestyle='--', alpha=0.4, linewidth=0.8)\n",
+    "    ax.axhline(0.80, color='#cc4444', linestyle='--', alpha=0.4, linewidth=0.8)\n",
+    "    pct = 100 * in_spec.mean()\n",
+    "    ax.set_title(f'{model}\\n{pct:.0f}% in spec', fontsize=10, color=color)\n",
+    "    ax.set_xlabel('SCF (kN/cm)', fontsize=8)\n",
+    "\n",
+    "axes[0].set_ylabel('Ribbon Relative Density', fontsize=10)\n",
+    "plt.colorbar(sc, ax=axes[-1], shrink=0.8, label='Yield %', pad=0.15)\n",
+    "fig.suptitle('Design Space Across the CL Platform (RD 0.60\\u20130.80, Fines<20%, Yield>70%)',\n",
+    "             fontsize=13, color=IPA_NAVY, y=1.05)\n",
+    "plt.tight_layout()\n",
+    "plt.show()"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## 4. Scale-Up Consistency: Does Quality Transfer Across the CL Platform?\n",
+    "\n",
+    "A key claim: IPA\u2019s platform delivers **scalable results**. Let\u2019s verify that\n",
+    "ribbon quality at the same SCF is consistent from CL25150 to CL100250."
+  ]},
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "fig, axes = plt.subplots(1, 3, figsize=(16, 5))\n",
+    "\n",
+    "# Filter to common operating condition: mid-pressure, VFS ratio ~1.0\n",
+    "common = df[(df.roll_pressure_fraction.between(0.45, 0.55)) &\n",
+    "            (df.vfs_hfs_ratio.between(0.9, 1.1))]\n",
+    "\n",
+    "responses = ['ribbon_rel_density', 'density_cv_pct', 'granule_yield_pct']\n",
+    "ylabels = ['Ribbon Relative Density', 'Density CV%', 'Granule Yield %']\n",
+    "titles = ['Ribbon Density Transfers', 'Uniformity is Maintained', 'Yield is Consistent']\n",
+    "\n",
+    "for ax, resp, ylabel, title in zip(axes, responses, ylabels, titles):\n",
+    "    sns.boxplot(data=common, x='compactor_model', y=resp,\n",
+    "                order=models_ordered,\n",
+    "                palette=MODEL_COLORS, ax=ax,\n",
+    "                fliersize=2, linewidth=0.8)\n",
+    "    ax.set_title(title, fontsize=11)\n",
+    "    ax.set_ylabel(ylabel)\n",
+    "    ax.set_xlabel('')\n",
+    "    ax.tick_params(axis='x', rotation=30)\n",
+    "\n",
+    "fig.suptitle('Scale-Up Consistency at Matched Operating Conditions (50% pressure, VFS ratio \\u2248 1.0)',\n",
+    "             fontsize=13, color=IPA_NAVY, y=1.02)\n",
+    "plt.tight_layout()\n",
+    "plt.show()\n",
+    "\n",
+    "print('Key finding: ribbon density and yield remain remarkably consistent')\n",
+    "print('across the entire CL platform at matched conditions. This is the')\n",
+    "print('scalable platform promise in action.')"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## 5. Energy Efficiency vs. Throughput: Pareto Frontier\n",
+    "\n",
+    "The best operating conditions maximize throughput while minimizing specific\n",
+    "energy consumption. The Pareto frontier shows the efficiency boundary."
+  ]},
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "fig, ax = plt.subplots(figsize=(12, 7))\n",
+    "\n",
+    "for model in models_ordered:\n",
+    "    sub = df[df.compactor_model == model]\n",
+    "    ax.scatter(sub['throughput_kg_hr'], sub['specific_energy_kwh_tonne'],\n",
+    "              alpha=0.25, s=15, color=MODEL_COLORS[model], label=model,\n",
+    "              edgecolors='white', linewidth=0.3)\n",
+    "\n",
+    "# Pareto frontier approximation\n",
+    "for model in models_ordered:\n",
+    "    sub = df[df.compactor_model == model].copy()\n",
+    "    sub = sub.sort_values('throughput_kg_hr')\n",
+    "    # Simple Pareto: cumulative min of energy as throughput increases\n",
+    "    pareto_energy = sub['specific_energy_kwh_tonne'].cummin()\n",
+    "    ax.plot(sub['throughput_kg_hr'], pareto_energy,\n",
+    "            color=MODEL_COLORS[model], linewidth=2, alpha=0.8)\n",
+    "\n",
+    "ax.set_xlabel('Throughput (kg/hr)', fontsize=12)\n",
+    "ax.set_ylabel('Specific Energy (kWh/tonne)', fontsize=12)\n",
+    "ax.set_title('Energy Efficiency vs. Throughput: Pareto Frontiers by Model',\n",
+    "             fontsize=14, color=IPA_NAVY)\n",
+    "ax.legend(title='CL Model', loc='upper right', fontsize=9)\n",
+    "ax.set_xlim(0, None)\n",
+    "ax.set_ylim(0, None)\n",
+    "plt.tight_layout()\n",
+    "plt.show()"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## 6. Platform Scorecard: Radar Chart per Model\n",
+    "\n",
+    "A comprehensive radar chart comparing all five CL models on the six\n",
+    "dimensions that matter for equipment selection."
+  ]},
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "categories = ['Throughput', 'Ribbon Quality\\n(Zinchuk %)', 'Uniformity\\n(low CV)',\n",
+    "              'Yield', 'Energy\\nEfficiency', 'Fast\\nChangeover']\n",
+    "N = len(categories)\n",
+    "angles = np.linspace(0, 2*np.pi, N, endpoint=False).tolist()\n",
+    "angles += angles[:1]\n",
+    "\n",
+    "fig, ax = plt.subplots(figsize=(8, 8), subplot_kw=dict(polar=True))\n",
+    "ax.set_facecolor('#fafafa')\n",
+    "\n",
+    "for model in models_ordered:\n",
+    "    sub = df[df.compactor_model == model]\n",
+    "    tp_max = df.throughput_kg_hr.max()\n",
+    "    scores = [\n",
+    "        sub.throughput_kg_hr.mean() / tp_max,\n",
+    "        (sub.in_zinchuk_window == 'Yes').mean(),\n",
+    "        1 - sub.density_cv_pct.mean() / 12,\n",
+    "        sub.granule_yield_pct.mean() / 100,\n",
+    "        1 - sub.specific_energy_kwh_tonne.mean() / df.specific_energy_kwh_tonne.quantile(0.95),\n",
+    "        1 - sub.changeover_time_hr.mean() / 4,\n",
+    "    ]\n",
+    "    scores = [np.clip(s, 0, 1) for s in scores]\n",
+    "    scores += scores[:1]\n",
+    "\n",
+    "    ax.fill(angles, scores, alpha=0.08, color=MODEL_COLORS[model])\n",
+    "    ax.plot(angles, scores, 'o-', color=MODEL_COLORS[model],\n",
+    "            linewidth=2, markersize=5, label=model)\n",
+    "\n",
+    "ax.set_xticks(angles[:-1])\n",
+    "ax.set_xticklabels(categories, fontsize=10)\n",
+    "ax.set_ylim(0, 1)\n",
+    "ax.set_title('IPA CL-Series Platform Scorecard',\n",
+    "             fontsize=14, color=IPA_NAVY, pad=25)\n",
+    "ax.legend(loc='lower left', bbox_to_anchor=(-0.15, -0.12),\n",
+    "          ncol=5, fontsize=9, frameon=True)\n",
+    "plt.tight_layout()\n",
+    "plt.show()"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## 7. Material Fingerprints: How Each Formulation Behaves Across the Platform\n",
+    "\n",
+    "A faceted heatmap showing ribbon density response to SCF and roll speed\n",
+    "for each material, revealing material-specific operating windows."
+  ]},
+  {"cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [
+    "materials = sorted(df.material.unique())\n",
+    "fig, axes = plt.subplots(1, 5, figsize=(22, 4), sharey=True)\n",
+    "\n",
+    "for ax, mat in zip(axes, materials):\n",
+    "    sub = df[(df.material == mat) & (df.compactor_model == 'CL50200')]\n",
+    "    pivot = sub.groupby(['roll_pressure_fraction', 'roll_speed_rpm'])['ribbon_rel_density'].mean().unstack()\n",
+    "    sns.heatmap(pivot, ax=ax, cmap=IPA_GRADIENT, vmin=0.45, vmax=0.85,\n",
+    "                cbar=ax==axes[-1], annot=True, fmt='.2f', annot_kws={'size': 7},\n",
+    "                linewidths=0.5, linecolor='white')\n",
+    "    ax.set_title(mat.replace('_', ' '), fontsize=9)\n",
+    "    ax.set_xlabel('Roll Speed (rpm)', fontsize=8)\n",
+    "    if ax == axes[0]:\n",
+    "        ax.set_ylabel('Pressure Fraction', fontsize=9)\n",
+    "    else:\n",
+    "        ax.set_ylabel('')\n",
+    "\n",
+    "fig.suptitle('Material Fingerprints: Ribbon Density Response on CL50200',\n",
+    "             fontsize=14, color=IPA_NAVY, y=1.05)\n",
+    "plt.tight_layout()\n",
+    "plt.show()\n",
+    "\n",
+    "print('Each material has a distinct operating fingerprint.')\n",
+    "print('Plastic materials (MCC) show strong roll speed sensitivity;')\n",
+    "print('brittle materials (lactose, mannitol) are pressure-driven.')"
+  ]},
+
+  {"cell_type": "markdown", "metadata": {}, "source": [
+    "---\n",
+    "## Takeaways\n",
+    "\n",
+    "1. **The IPA CL platform scales predictably** from 10 kg/hr (CL25150 lab)\n",
+    "   to 177+ kg/hr (CL100250 production) with consistent ribbon quality\n",
+    "   at matched operating conditions.\n",
+    "\n",
+    "2. **Twin feed screw (HFS + VFS) is the key differentiator.** The VFS/HFS\n",
+    "   ratio is a uniquely tunable parameter that single-screw compactors\n",
+    "   cannot offer. Optimal ratio \u2248 1.0 across materials.\n",
+    "\n",
+    "3. **Material fingerprints guide process development.** Each formulation\n",
+    "   has a distinct response surface, but the optimal zone is consistent\n",
+    "   across the CL platform \u2014 enabling direct R&D-to-production transfer.\n",
+    "\n",
+    "4. **Integrated PM-series milling** (in-air impact method) delivers high\n",
+    "   granule yields (67\u201374%) with minimal heat generation.\n",
+    "\n",
+    "5. **Energy efficiency improves at production scale** \u2014 larger models\n",
+    "   operate on the Pareto frontier with lower specific energy per tonne.\n",
+    "\n",
+    "---\n",
+    "\n",
+    "For a lab test or performance assessment on IPA\u2019s CL-series platform:  \n",
+    "**https://www.innovativeprocess.com** | **(708) 844-6100** | info@ipaapplications.com\n",
+    "\n",
+    "*Dataset \u00a9 2026 Innovative Process Applications, CC BY 4.0.*"
+  ]}
+ ],
+ "metadata": {
+  "colab": {"provenance": [], "toc_visible": true},
+  "kernelspec": {"display_name": "Python 3", "language": "python", "name": "python3"},
+  "language_info": {"name": "python", "version": "3.11"}
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

LICENSE.txt

@@ -0,0 +1,5 @@
+Creative Commons Attribution 4.0 International (CC BY 4.0)
+===========================================================
+Copyright (c) 2026 Innovative Process Applications (IPA)
+Full license: https://creativecommons.org/licenses/by/4.0/legalcode
+DISCLAIMER: Synthetic educational data only. Not production guarantees.

README.md

@@ -1,3 +1,92 @@
----
-license: cc-by-4.0
----
+# IPA Pharmaceutical Roller Compactor Platform: Scale-Up & Performance (Synthetic)
+
+**Version:** 1.0
+**Publisher:** [Innovative Process Applications (IPA)](https://www.innovativeprocess.com)
+**License:** Creative Commons Attribution 4.0 International (CC BY 4.0)
+
+> ⚠️ **This dataset is 100% synthetic and intended for educational use only.**
+> Generated from IPA's published CL-series specifications and standard
+> compaction physics — not real production or clinical batch data.
+
+---
+
+## What's in this dataset
+
+3,000 simulated pharmaceutical roller compaction runs spanning IPA's complete
+CL-series platform — from R&D lab (CL25150, 5 HP) through full production
+(CL100250, 25 HP) — across 5 pharma/nutraceutical materials and the full
+operating envelope of twin-feed-screw process parameters.
+
+### Equipment specifications (from IPA published data)
+
+| Model | Roll Size | Max Pressure | Integrated Mill | Power | Scale |
+|---|---|---|---|---|---|
+| CL25150 | 1"×6" (2.5×15 cm) | 17.5 kN/cm | None | 5 HP | R&D / Lab |
+| CL30200 | 1"×8" (3×20 cm) | 17.5 kN/cm | PM3 | 2.75 HP | Pilot |
+| CL50200 | 2"×8" (5×20 cm) | 26.0 kN/cm | PM6 | 12 HP | Pilot / Small Prod |
+| CL75200 | 3"×8" (7.5×20 cm) | 15.0 kN/cm | PM6 | 20 HP | Production |
+| CL100250 | 4"×10" (10×25 cm) | 17.5 kN/cm | PM8 | 25 HP | Full Production |
+
+### Column descriptions
+
+| Column | Units | Description |
+|---|---|---|
+| `compactor_model` | — | IPA CL-series model |
+| `scale` | — | R&D / Lab, Pilot, Production, Full Production |
+| `roll_diameter_in` / `_cm` | in / cm | Roll diameter |
+| `roll_width_in` / `_cm` | in / cm | Roll width |
+| `max_roll_pressure_kn_cm` | kN/cm | Maximum specific compaction force |
+| `integrated_mill` | — | PM-series in-air impact mill (or None) |
+| `total_power_kw` | kW | Total system power |
+| `material` | — | Pharma material identifier |
+| `feed_density_gcc` | g/cc | Feed powder bulk density |
+| `deformation_type` | — | Plastic, brittle, or mixed |
+| `roll_pressure_fraction` | — | Fraction of max pressure applied (0.3–1.0) |
+| `scf_kn_cm` | kN/cm | Actual specific compaction force |
+| `roll_speed_rpm` | rpm | Roll rotation speed |
+| `hfs_speed_rpm` | rpm | Horizontal feed screw speed (controls throughput) |
+| `vfs_hfs_ratio` | — | Vertical/horizontal feed screw speed ratio (controls pre-compression) |
+| `gap_width_mm` | mm | Roll gap |
+| `ribbon_rel_density` | — | Relative density (fraction of true density) |
+| `ribbon_density_gcc` | g/cc | Absolute ribbon density |
+| `ribbon_porosity` | — | 1 − relative density |
+| `density_cv_pct` | % | Across-ribbon density uniformity |
+| `fines_pct` | % | Fines fraction after milling |
+| `granule_yield_pct` | % | Usable granule yield |
+| `in_zinchuk_window` | Yes/No | Ribbon RD in 0.60–0.80 tabletability range |
+| `throughput_kg_hr` / `_lbs_hr` | kg/hr / lbs/hr | Mass throughput |
+| `specific_energy_kwh_tonne` | kWh/tonne | Energy efficiency |
+| `changeover_time_hr` | hours | Estimated changeover time |
+
+## IPA platform advantages demonstrated in the data
+
+1. **Twin feed screw (HFS + VFS):** Independent control of throughput (HFS)
+   and pre-compression (VFS) produces a uniquely tunable operating space.
+   The VFS/HFS ratio is a key process parameter not available on single-screw
+   compactors.
+
+2. **Scalable platform:** Consistent ribbon quality (CV%, yield) across the
+   CL25150 → CL100250 range, enabling direct R&D-to-production scale-up.
+
+3. **Integrated in-air impact milling:** PM-series mills minimize heat
+   generation and maximize granule yield with precise PSD control.
+
+4. **Efficient changeover:** Modular design enables fast product changeover,
+   critical for multi-product pharma facilities.
+
+5. **Low bulk density capability:** Twin screw design is proven efficient
+   with light powders (feed density 0.3–0.5 g/cc) containing high air content.
+
+## Cross-links
+
+- **Kaggle:** [link after publication]
+- **Hugging Face:** [link after publication]
+- **Zenodo:** [link after publication]
+- **GitHub:** [link after publication]
+- **IPA website:** https://www.innovativeprocess.com
+
+## Citation
+
+> Innovative Process Applications (2026). *IPA Pharmaceutical Roller Compactor
+> Platform: Scale-Up & Performance (Synthetic), v1.0*. CC BY 4.0.
+> https://www.innovativeprocess.com

generate_dataset.py

@@ -0,0 +1,310 @@
+"""
+IPA Pharmaceutical Roller Compactor Platform: Scale-Up & Performance Dataset v1.0
+==================================================================================
+Synthetic dataset modeling the IPA CL-series pharmaceutical roller compactor
+platform from R&D (CL25150) through full-scale production (CL100250), using
+IPA's published specifications and Johanson/Heckel physical models.
+
+Sources:
+  - IPA Pharma RC Brochure (2026)
+  - IPA Pharma Compactor Specifications page
+  - IPA Roll Compactor page (industrial line dimensions)
+  - IPA Products & Services PDF
+
+Key IPA differentiators modeled:
+  - Twin feed screw design (HFS + VFS independent control)
+  - Scalable platform: consistent ribbon quality across CL sizes
+  - Integrated milling (PM-series in-air impact mills)
+  - Proprietary PLC controls with internal control loops
+  - Low bulk density powder processing capability
+  - Efficient changeover and simplified maintenance
+
+THIS IS SYNTHETIC EDUCATIONAL DATA. NOT REAL CUSTOMER OR LAB DATA.
+"""
+
+import numpy as np
+import pandas as pd
+
+rng = np.random.default_rng(seed=2026)
+
+# =============================================================================
+# IPA CL-SERIES PHARMA COMPACTOR SPECS (from published specifications)
+# =============================================================================
+CL_MODELS = [
+    {"model": "CL25150",  "roll_dia_in": 1.0, "roll_width_in": 6,
+     "roll_dia_cm": 2.5,  "roll_width_cm": 15,
+     "max_pressure_lbs_in": 9900, "max_pressure_kn_cm": 17.5,
+     "cap_light_lbs": (10, 23),   "cap_heavy_lbs": (25, 56),
+     "mill_size": None, "total_hp": 5, "total_kw": 3.75,
+     "weight_lbs": 1200, "scale": "R&D / Lab"},
+    {"model": "CL30200",  "roll_dia_in": 1.0, "roll_width_in": 8,
+     "roll_dia_cm": 3.0,  "roll_width_cm": 20,
+     "max_pressure_lbs_in": 9900, "max_pressure_kn_cm": 17.5,
+     "cap_light_lbs": (24, 54),   "cap_heavy_lbs": (64, 140),
+     "mill_size": "PM3", "total_hp": 2.75, "total_kw": 2.0,
+     "weight_lbs": 2100, "scale": "Pilot"},
+    {"model": "CL50200",  "roll_dia_in": 2.0, "roll_width_in": 8,
+     "roll_dia_cm": 5.0,  "roll_width_cm": 20,
+     "max_pressure_lbs_in": 14800, "max_pressure_kn_cm": 26.0,
+     "cap_light_lbs": (64, 140),  "cap_heavy_lbs": (120, 265),
+     "mill_size": "PM6", "total_hp": 12, "total_kw": 9.0,
+     "weight_lbs": 5000, "scale": "Pilot / Small Production"},
+    {"model": "CL75200",  "roll_dia_in": 3.0, "roll_width_in": 8,
+     "roll_dia_cm": 7.5,  "roll_width_cm": 20,
+     "max_pressure_lbs_in": 8500, "max_pressure_kn_cm": 15.0,
+     "cap_light_lbs": (95, 209),  "cap_heavy_lbs": (182, 400),
+     "mill_size": "PM6", "total_hp": 20, "total_kw": 15.0,
+     "weight_lbs": 6000, "scale": "Production"},
+    {"model": "CL100250", "roll_dia_in": 4.0, "roll_width_in": 10,
+     "roll_dia_cm": 10.0, "roll_width_cm": 25,
+     "max_pressure_lbs_in": 9900, "max_pressure_kn_cm": 17.5,
+     "cap_light_lbs": (200, 440), "cap_heavy_lbs": (425, 935),
+     "mill_size": "PM8", "total_hp": 25, "total_kw": 19.0,
+     "weight_lbs": 9500, "scale": "Full Production"},
+]
+
+# =============================================================================
+# PHARMA MATERIALS (representative formulations)
+# =============================================================================
+MATERIALS = {
+    "MCC_PH101": {
+        "label": "MCC PH-101 (Low Density Filler)",
+        "feed_density_gcc": 0.32, "compact_density_gcc": 1.20,
+        "heckel_k": 0.020, "heckel_a": 0.55,
+        "deformation": "plastic", "flow_index": 4,
+        "moisture_pct": 4.5,
+    },
+    "lactose_DCL11": {
+        "label": "Lactose DCL-11 (Direct Compression)",
+        "feed_density_gcc": 0.62, "compact_density_gcc": 1.45,
+        "heckel_k": 0.014, "heckel_a": 0.50,
+        "deformation": "brittle", "flow_index": 7,
+        "moisture_pct": 0.5,
+    },
+    "mannitol_SD200": {
+        "label": "Mannitol SD-200 (Spray Dried)",
+        "feed_density_gcc": 0.48, "compact_density_gcc": 1.49,
+        "heckel_k": 0.012, "heckel_a": 0.48,
+        "deformation": "brittle", "flow_index": 6,
+        "moisture_pct": 0.3,
+    },
+    "API_blend_40pct": {
+        "label": "API Blend 40% Drug Load",
+        "feed_density_gcc": 0.38, "compact_density_gcc": 1.30,
+        "heckel_k": 0.017, "heckel_a": 0.52,
+        "deformation": "mixed", "flow_index": 3,
+        "moisture_pct": 2.0,
+    },
+    "vitamin_premix": {
+        "label": "Vitamin/Mineral Premix (Nutraceutical)",
+        "feed_density_gcc": 0.45, "compact_density_gcc": 1.35,
+        "heckel_k": 0.015, "heckel_a": 0.50,
+        "deformation": "mixed", "flow_index": 5,
+        "moisture_pct": 3.0,
+    },
+}
+
+# =============================================================================
+# PROCESS PARAMETERS
+# =============================================================================
+ROLL_PRESSURE_FRACTIONS = [0.3, 0.5, 0.7, 0.85, 1.0]  # fraction of max
+ROLL_SPEED_RPM = [2, 4, 6, 8, 10]
+HFS_SPEED_RPM = [15, 30, 50, 75, 100]  # horizontal feed screw
+VFS_RATIO = [0.6, 0.8, 1.0, 1.2, 1.5]  # VFS/HFS ratio
+
+N_REPLICATES = 3  # per condition
+
+# =============================================================================
+# PHYSICS
+# =============================================================================
+
+def compute_scf(pressure_lbs_in, roll_width_in):
+    """Specific compaction force in kN/cm."""
+    return (pressure_lbs_in * 0.00444822) / (roll_width_in * 2.54)
+
+def ribbon_density(scf_kn_cm, roll_dia_cm, gap_mm, heckel_k, heckel_a,
+                   vfs_ratio, hfs_rpm, roll_rpm, deformation):
+    """Compute ribbon relative density using Heckel + IPA twin-screw model."""
+    # Gap estimate based on roll geometry and pressure
+    contact_len = np.sqrt(roll_dia_cm * 10 / 2 * 2.0 * gap_mm)
+    pressure_mpa = (scf_kn_cm * 100) / max(contact_len, 3.0)
+
+    # Heckel
+    rd = 1.0 - np.exp(-(heckel_k * pressure_mpa + heckel_a))
+
+    # Twin feed screw VFS ratio effect — optimal around 1.0
+    vfs_optimality = np.exp(-((vfs_ratio - 1.0) ** 2) / (2 * 0.15 ** 2))
+    rd *= (0.90 + 0.10 * vfs_optimality)
+
+    # Roll speed / dwell time
+    if deformation == "plastic":
+        rd *= (1.0 - 0.006 * max(roll_rpm - 4, 0))
+    elif deformation == "brittle":
+        rd *= (1.0 - 0.001 * max(roll_rpm - 4, 0))
+    else:
+        rd *= (1.0 - 0.003 * max(roll_rpm - 4, 0))
+
+    # HFS feed rate effect on pre-densification
+    feed_ratio = hfs_rpm / max(roll_rpm, 1)
+    feed_opt = np.exp(-((feed_ratio - 10) ** 2) / (2 * 5 ** 2))
+    rd *= (0.95 + 0.05 * feed_opt)
+
+    return np.clip(rd, 0.35, 0.92)
+
+def compute_throughput(model, material, roll_rpm, hfs_rpm, rd):
+    """Throughput in kg/hr based on capacity range and operating conditions."""
+    if material["feed_density_gcc"] <= 0.5:
+        cap_range = model["cap_light_lbs"]
+    else:
+        cap_range = model["cap_heavy_lbs"]
+
+    # Scale within capacity range based on operating conditions
+    rpm_frac = (roll_rpm - 2) / 8
+    hfs_frac = (hfs_rpm - 15) / 85
+    operating_frac = 0.5 * rpm_frac + 0.5 * hfs_frac
+
+    throughput_lbs = cap_range[0] + (cap_range[1] - cap_range[0]) * np.clip(operating_frac, 0, 1)
+    throughput_kg = throughput_lbs * 0.4536
+    return throughput_lbs, throughput_kg
+
+def density_uniformity_cv(vfs_ratio, hfs_rpm, roll_rpm, model_scale):
+    """Across-ribbon density CV%. Twin feed screw advantage."""
+    # Baseline: twin screw gives good uniformity
+    base_cv = 2.5
+    # VFS ratio: optimal around 1.0
+    vfs_penalty = 2.0 * abs(vfs_ratio - 1.0)
+    # Feed ratio
+    ratio = hfs_rpm / max(roll_rpm, 1)
+    ratio_penalty = 1.5 * abs(ratio - 10) / 10
+    # Scale: larger machines slightly harder to keep uniform
+    scale_factors = {"R&D / Lab": 0, "Pilot": 0.2, "Pilot / Small Production": 0.3,
+                     "Production": 0.5, "Full Production": 0.7}
+    scale_pen = scale_factors.get(model_scale, 0.3)
+    return base_cv + vfs_penalty + ratio_penalty + scale_pen
+
+def compute_granule_yield(rd, fines_pct):
+    """Yield = 100% - fines% - oversize%."""
+    oversize = max(0, 5 * (rd - 0.82))  # over-compacted ribbons resist milling
+    return np.clip(100 - fines_pct - oversize, 40, 98)
+
+def compute_fines(rd, deformation):
+    """Fines fraction after integrated PM-series mill."""
+    base = 55 * (1 - rd)
+    if deformation == "brittle":
+        base += 5
+    return np.clip(base, 3, 50)
+
+def compute_changeover_hr(model):
+    """Changeover time — IPA advantage: efficient changeover design."""
+    # Scales with machine size
+    base = 1.0 + 0.5 * np.log2(model["weight_lbs"] / 1200)
+    return round(base, 1)
+
+# =============================================================================
+# GENERATE DATASET
+# =============================================================================
+rows = []
+run_id = 0
+
+for model in CL_MODELS:
+    for mat_key, mat in MATERIALS.items():
+        for pf in ROLL_PRESSURE_FRACTIONS:
+            for rs in ROLL_SPEED_RPM:
+                for hfs in HFS_SPEED_RPM:
+                    for vr in VFS_RATIO:
+                        for rep in range(N_REPLICATES):
+                            run_id += 1
+                            pressure = model["max_pressure_lbs_in"] * pf
+                            scf = compute_scf(pressure, model["roll_width_in"])
+                            gap_mm = 1.5 + rng.uniform(-0.3, 0.3)
+
+                            rd = ribbon_density(
+                                scf, model["roll_dia_cm"], gap_mm,
+                                mat["heckel_k"], mat["heckel_a"],
+                                vr, hfs, rs, mat["deformation"])
+                            rd += rng.normal(0, 0.010)
+                            rd = np.clip(rd, 0.35, 0.92)
+
+                            rib_density_gcc = rd * mat["compact_density_gcc"]
+                            porosity = 1 - rd
+
+                            cv = density_uniformity_cv(vr, hfs, rs, model["scale"])
+                            cv += rng.normal(0, 0.3)
+                            cv = np.clip(cv, 1.0, 12.0)
+
+                            fines = compute_fines(rd, mat["deformation"])
+                            fines += rng.normal(0, 1.5)
+                            fines = np.clip(fines, 2, 55)
+
+                            tp_lbs, tp_kg = compute_throughput(model, mat, rs, hfs, rd)
+                            tp_kg += rng.normal(0, tp_kg * 0.03)
+                            tp_kg = max(tp_kg, 1)
+                            tp_lbs = tp_kg / 0.4536
+
+                            granule_yield = compute_granule_yield(rd, fines)
+                            granule_yield += rng.normal(0, 1.0)
+                            granule_yield = np.clip(granule_yield, 35, 99)
+
+                            zinchuk = "Yes" if 0.60 <= rd <= 0.80 else "No"
+                            changeover = compute_changeover_hr(model)
+
+                            # Specific energy
+                            power_kw = model["total_kw"] * (0.4 + 0.6 * pf)
+                            se = (power_kw / max(tp_kg, 1)) * 1000  # kWh/tonne
+
+                            rows.append({
+                                "run_id": run_id,
+                                "compactor_model": model["model"],
+                                "scale": model["scale"],
+                                "roll_diameter_in": model["roll_dia_in"],
+                                "roll_width_in": model["roll_width_in"],
+                                "roll_diameter_cm": model["roll_dia_cm"],
+                                "roll_width_cm": model["roll_width_cm"],
+                                "max_roll_pressure_kn_cm": model["max_pressure_kn_cm"],
+                                "integrated_mill": model["mill_size"] if model["mill_size"] else "None",
+                                "total_power_kw": model["total_kw"],
+                                "material": mat_key,
+                                "feed_density_gcc": mat["feed_density_gcc"],
+                                "deformation_type": mat["deformation"],
+                                "roll_pressure_fraction": pf,
+                                "scf_kn_cm": round(scf, 2),
+                                "roll_speed_rpm": rs,
+                                "hfs_speed_rpm": hfs,
+                                "vfs_hfs_ratio": vr,
+                                "gap_width_mm": round(gap_mm, 2),
+                                "ribbon_rel_density": round(rd, 4),
+                                "ribbon_density_gcc": round(rib_density_gcc, 4),
+                                "ribbon_porosity": round(porosity, 4),
+                                "density_cv_pct": round(cv, 2),
+                                "fines_pct": round(fines, 2),
+                                "granule_yield_pct": round(granule_yield, 2),
+                                "in_zinchuk_window": zinchuk,
+                                "throughput_kg_hr": round(tp_kg, 1),
+                                "throughput_lbs_hr": round(tp_lbs, 1),
+                                "specific_energy_kwh_tonne": round(se, 2),
+                                "changeover_time_hr": changeover,
+                                "replicate": rep + 1,
+                            })
+
+df = pd.DataFrame(rows)
+
+# Subsample to manageable size (full factorial is huge)
+# Keep ~3000 rows: stratified by model and material
+samples = []
+for _, group in df.groupby(["compactor_model", "material"]):
+    samples.append(group.sample(min(120, len(group)), random_state=42))
+df_sampled = pd.concat(samples, ignore_index=True)
+df_sampled["run_id"] = range(1, len(df_sampled) + 1)
+df_sampled.to_csv("ipa_pharma_compactor_v1.0.csv", index=False)
+
+print(f"Wrote {len(df_sampled)} rows, {len(df_sampled.columns)} columns")
+print(f"\n=== Model distribution ===")
+print(df_sampled["compactor_model"].value_counts().sort_index())
+print(f"\n=== Scale-up: mean throughput by model ===")
+print(df_sampled.groupby("compactor_model")["throughput_kg_hr"].mean().round(1))
+print(f"\n=== Zinchuk compliance ===")
+print(df_sampled["in_zinchuk_window"].value_counts())
+print(f"\n=== Mean density CV% by model ===")
+print(df_sampled.groupby("compactor_model")["density_cv_pct"].mean().round(2))
+print(f"\n=== Mean granule yield by model ===")
+print(df_sampled.groupby("compactor_model")["granule_yield_pct"].mean().round(1))