Executive Overview + Technical Description

HMR-EPI

Extrusion Process Intelligence - AI-Powered Hybrid Intelligence for Plastic Extrusion

Turning Process Data into Intelligence.

HMR-EPI is a Hybrid Intelligence Platform designed to transform production data, process knowledge and operational experience into actionable intelligence for plastics manufacturing, initially focused on PVC sheet extrusion. Unlike conventional AI systems that rely primarily on statistical models, HMR-EPI combines production data, historical production records, expert correlation intelligence, engineering rules, technical documentation, Lean Six Sigma DMAIC methodologies and artificial intelligence into a unified decision-support framework. This Hybrid Intelligence approach provides explainable, technically grounded recommendations while significantly reducing the risk of unsupported AI conclusions.

The platform enables manufacturers to reduce scrap, improve productivity, accelerate startups, optimize energy consumption, preserve decades of expert knowledge and continuously improve operational performance. By transforming process data into sustainable operational intelligence, HMR-EPI helps organizations make better decisions, achieve faster problem resolution and establish a culture of continuous improvement while maintaining full control over all production data and process know-how.

Contents

This document combines an executive business overview with a technical description for management, production, quality, automation and IT evaluation.

Part A - Executive Overview

1Executive Summary
2Why Customers Invest
3Hybrid Intelligence for PVC Extrusion
4Four Integrated Intelligence Areas
5Typical Customer Journey
6How HMR-EPI Generates Explainable Recommendations
7Example KPI Dashboard & Process Analytics
8Potential Business Impact & ROI
9Commercial Model (Illustrative)
10Why HMR-EPI

Part B - Technical Description

1Formulation Intelligence
2Mixing & Blending Intelligence
3Extrusion Process Intelligence
4Quality & Market Intelligence
5Equipment Selection & Optimization
6Hybrid Learning Approach
7Expert Correlation Intelligence
8Conversational Industrial Intelligence
9Lean Six Sigma DMAIC Integration
10Example Use Case: Startup Optimization of a PVC Free Foam Sheet Line
11Real-Time Dashboard & KPI Analytics
12Example ROI & Savings Scenario
13Integration & Connectivity
14Architecture & Security
15Secure Data Transfer Concept
16Possible Interfaces & Connectivity
17HMR-EPI Intelligence Framework
18Statistical Methods
19Hybrid Intelligence Technologies & Knowledge Framework
20Process Data into Intelligence

Part A

Executive Overview

Executive Summary

HMR-EPI is a modular Industrial AI Platform for plastics extrusion, initially focused on PVC sheet extrusion. While traditional systems collect, store and visualize data, HMR-EPI transforms production and quality data into actionable, explainable operational intelligence.

The platform combines decades of PVC sheet extrusion expertise, Lean Six Sigma methodologies, engineering rules, historical production knowledge and modern Industrial AI technologies. Its purpose is to support operators, process engineers, production managers and management teams in achieving more stable processes, better quality, lower scrap, faster startups and improved productivity.

Executive KPI / Value Driver

Executive KPI / Value Driver: Typical ROI range
Illustrative Potential / Example: 3–12 months

Executive KPI / Value Driver: Annual improvement potential
Illustrative Potential / Example: Up to €1.2 million per line and year

Executive KPI / Value Driver: Productivity improvement
Illustrative Potential / Example: Up to +15%

Executive KPI / Value Driver: Scrap reduction
Illustrative Potential / Example: Up to -22%

Executive KPI / Value Driver: Startup time reduction
Illustrative Potential / Example: Up to -35%

Executive KPI / Value Driver: Energy reduction
Illustrative Potential / Example: Up to -12%

Executive KPI / Value Driver	Illustrative Potential / Example
Typical ROI range	3–12 months
Annual improvement potential	Up to €1.2 million per line and year
Productivity improvement	Up to +15%
Scrap reduction	Up to -22%
Startup time reduction	Up to -35%
Energy reduction	Up to -12%

Why Customers Invest

PVC sheet producers face increasing operational pressure: higher energy costs, tighter quality requirements, limited availability of experienced operators, and the need to reduce material losses and startup scrap. HMR-EPI addresses these challenges by transforming experience-driven process knowledge into repeatable, measurable and scalable operational intelligence.

Customer Challenge: Knowledge loss and retirements
HMR-EPI Response: Capture and preserve process know-how and best practices.

Customer Challenge: Operator dependency
HMR-EPI Response: Benchmark operator strategies and standardize successful operating windows.

Customer Challenge: Process instability
HMR-EPI Response: Detect patterns, drift and recurring causes of variation.

Customer Challenge: High scrap and off-spec production
HMR-EPI Response: Identify root causes and recommend corrective actions.

Customer Challenge: Long startup times
HMR-EPI Response: Analyze successful startups and replicate best-practice procedures.

Customer Challenge: Rising energy costs
HMR-EPI Response: Evaluate energy per kg, process windows and optimization potential.

Customer Challenge	HMR-EPI Response
Knowledge loss and retirements	Capture and preserve process know-how and best practices.
Operator dependency	Benchmark operator strategies and standardize successful operating windows.
Process instability	Detect patterns, drift and recurring causes of variation.
High scrap and off-spec production	Identify root causes and recommend corrective actions.
Long startup times	Analyze successful startups and replicate best-practice procedures.
Rising energy costs	Evaluate energy per kg, process windows and optimization potential.

Hybrid Intelligence for PVC Extrusion

HMR-EPI is intentionally designed as a Hybrid Intelligence Platform rather than a conventional AI system. It does not rely on AI alone. The system combines production data, historical knowledge, expert correlation tables, engineering rules, Lean Six Sigma DMAIC methods, technical documentation, expert experience and AI-based analytics. This reduces the risk of unsupported AI recommendations and helps prevent hallucination-like behavior in industrial decision support.

Input Layer: Production Data
Purpose: Provides real operating conditions, process values, alarms and trends.

Input Layer: Historical Knowledge
Purpose: Preserves previous problem-solving cases and successful operating strategies.

Input Layer: Lean Six Sigma DMAIC
Purpose: Structures improvement work through Define, Measure, Analyze, Improve and Control.

Input Layer: PVC Processing Expertise
Purpose: Defines meaningful correlations based on decades of practical extrusion experience.

Input Layer: Engineering Rules
Purpose: Constrains recommendations to technically plausible and safe operating windows.

Input Layer: Expert Correlation Tables
Purpose: Capture decades of expert knowledge linking product defects, process deviations, equipment conditions, material properties and root causes.

Input Layer: Expert Interviews & Workshops
Purpose: Capture tacit operational knowledge, troubleshooting experience and best practices not available in databases.

Input Layer: Technical Documents & Publications
Purpose: Provide additional engineering knowledge from manuals, service reports, internal documents, presentations and technical publications.

Input Layer: Historical Cases & Proven Solutions
Purpose: Enable reuse of successful solutions from previous projects, production campaigns, customer cases and DMAIC activities.

Input Layer: AI & Machine Learning
Purpose: Detect patterns, correlations, anomalies and optimization potential across large datasets.

Input Layer	Purpose
Production Data	Provides real operating conditions, process values, alarms and trends.
Historical Knowledge	Preserves previous problem-solving cases and successful operating strategies.
Lean Six Sigma DMAIC	Structures improvement work through Define, Measure, Analyze, Improve and Control.
PVC Processing Expertise	Defines meaningful correlations based on decades of practical extrusion experience.
Engineering Rules	Constrains recommendations to technically plausible and safe operating windows.
Expert Correlation Tables	Capture decades of expert knowledge linking product defects, process deviations, equipment conditions, material properties and root causes.
Expert Interviews & Workshops	Capture tacit operational knowledge, troubleshooting experience and best practices not available in databases.
Technical Documents & Publications	Provide additional engineering knowledge from manuals, service reports, internal documents, presentations and technical publications.
Historical Cases & Proven Solutions	Enable reuse of successful solutions from previous projects, production campaigns, customer cases and DMAIC activities.
AI & Machine Learning	Detect patterns, correlations, anomalies and optimization potential across large datasets.

This creates explainable operational intelligence: recommendations are not only generated, but also explained with the data, expert knowledge, expert-defined correlations, historical cases and assumptions supporting them.

Best-of-Knowledge Principle

HMR-EPI follows a Best-of-Knowledge approach. Recommendations are generated and validated using multiple independent knowledge sources, including production data, historical cases, expert correlation tables, engineering rules, technical documentation, expert knowledge, DMAIC project knowledge and artificial intelligence.

Hallucination Risk Mitigation

Unlike conventional AI systems, HMR-EPI does not rely solely on statistical models. Recommendations are cross-checked against engineering knowledge, expert correlations and historical production experience. This multi-layer validation approach improves recommendation quality, enhances transparency, increases user confidence and helps reduce the risk of unsupported AI recommendations.

Four Integrated Intelligence Areas

HMR-EPI is structured around four intelligence areas that reflect the complete PVC sheet value chain: formulation, mixing and blending, extrusion, and quality/market feedback.

Intelligence Area: Formulation Intelligence
Primary Objective: Optimize raw materials, additives, suppliers, batch consistency, formulation stability and material cost.

Intelligence Area: Mixing & Blending Intelligence
Primary Objective: Improve dryblend consistency, mixing sequences, temperature control, cooling behavior and energy input.

Intelligence Area: Extrusion Process Intelligence
Primary Objective: Improve process stability, density control, throughput, OEE, startup performance, scrap reduction and energy efficiency.

Intelligence Area: Quality & Market Intelligence
Primary Objective: Integrate inline QC, laboratory QC, SPC, complaints and market feedback for continuous improvement.

Intelligence Area	Primary Objective
Formulation Intelligence	Optimize raw materials, additives, suppliers, batch consistency, formulation stability and material cost.
Mixing & Blending Intelligence	Improve dryblend consistency, mixing sequences, temperature control, cooling behavior and energy input.
Extrusion Process Intelligence	Improve process stability, density control, throughput, OEE, startup performance, scrap reduction and energy efficiency.
Quality & Market Intelligence	Integrate inline QC, laboratory QC, SPC, complaints and market feedback for continuous improvement.

Typical Customer Journey

HMR-EPI does not require customers to implement the complete platform from day one. The recommended deployment approach is modular and scalable. Customers can start with the highest-value business objective and expand step by step based on proven results and ROI.

Phase: Phase 1
Focus: Extrusion Process Intelligence
Typical Business Objective: Fastest ROI through startup optimization, scrap reduction, density optimization, OEE improvement and process stability.

Phase: Phase 2
Focus: Quality Intelligence
Typical Business Objective: Improved product consistency through SPC analytics, quality prediction, root cause analysis and complaint reduction.

Phase: Phase 3
Focus: Mixing & Blending Intelligence
Typical Business Objective: More stable production conditions through dryblend consistency, mixing optimization and material variability reduction.

Phase: Phase 4
Focus: Formulation Intelligence
Typical Business Objective: Strategic product and cost optimization through formulation development, supplier benchmarking and long-term performance analysis.

Phase: Optional Phase 5
Focus: Equipment Selection & Optimization
Typical Business Objective: Support optimized line configurations, screw design requirements, die concepts, calibration systems and inline measurement concepts.

Phase	Focus	Typical Business Objective
Phase 1	Extrusion Process Intelligence	Fastest ROI through startup optimization, scrap reduction, density optimization, OEE improvement and process stability.
Phase 2	Quality Intelligence	Improved product consistency through SPC analytics, quality prediction, root cause analysis and complaint reduction.
Phase 3	Mixing & Blending Intelligence	More stable production conditions through dryblend consistency, mixing optimization and material variability reduction.
Phase 4	Formulation Intelligence	Strategic product and cost optimization through formulation development, supplier benchmarking and long-term performance analysis.
Optional Phase 5	Equipment Selection & Optimization	Support optimized line configurations, screw design requirements, die concepts, calibration systems and inline measurement concepts.

How HMR-EPI Generates Explainable Recommendations

HMR-EPI transforms raw data into explainable recommendations by combining data sources, expert knowledge and statistical/AI analysis in a structured recommendation workflow.

Step: 1. Capture Data
Description: Collect production, quality, alarm, event, recipe, operator and laboratory data from available systems.

Step: 2. Contextualize
Description: Link process values to product, recipe, line, operator, shift, quality result and production event.

Step: 3. Analyze
Description: Detect correlations, patterns, process drift, root causes and best-practice operating windows.

Step: 4. Validate
Description: Apply engineering rules, Lean Six Sigma logic and expert-defined constraints.

Step: 5. Recommend
Description: Generate actionable recommendations for operators, engineers or management.

Step: 6. Explain
Description: Explain why the recommendation was made, which data supports it and which correlations were found.

Step: 7. Learn
Description: Update the knowledge base with new successful production outcomes and operator feedback.

Step	Description
1. Capture Data	Collect production, quality, alarm, event, recipe, operator and laboratory data from available systems.
2. Contextualize	Link process values to product, recipe, line, operator, shift, quality result and production event.
3. Analyze	Detect correlations, patterns, process drift, root causes and best-practice operating windows.
4. Validate	Apply engineering rules, Lean Six Sigma logic and expert-defined constraints.
5. Recommend	Generate actionable recommendations for operators, engineers or management.
6. Explain	Explain why the recommendation was made, which data supports it and which correlations were found.
7. Learn	Update the knowledge base with new successful production outcomes and operator feedback.

For each recommendation, HMR-EPI can explain what was discovered, why it matters, which variables have the strongest influence, which historical production runs support the conclusion, what improvement potential may be expected and the confidence level of the recommendation.

Additional Validation Layer

To ensure recommendation quality and transparency, HMR-EPI applies a Best-of-Knowledge validation approach. Recommendations are not generated solely from AI models. Instead, findings are validated against expert-defined correlation tables, engineering rules, historical production cases, technical documentation, DMAIC project knowledge, expert feedback and production and quality data.

Validation Source: 1. Expert Correlation Tables
Contribution: Validate whether identified relationships are technically plausible and consistent with practical PVC processing experience.

Validation Source: 2. Engineering Rules
Contribution: Constrain recommendations to safe, meaningful and technically feasible operating windows.

Validation Source: 3. Historical Production Cases
Contribution: Support recommendations with comparable past situations and proven solutions.

Validation Source: 4. Technical Documentation
Contribution: Provide additional context from reports, manuals, presentations and publications.

Validation Source: 5. DMAIC Project Knowledge
Contribution: Use verified root causes, corrective actions and control plans from previous improvement projects.

Validation Source: 6. Production & Quality Data
Contribution: Confirm the recommendation with actual process behavior, quality results and operating history.

Validation Source	Contribution
1. Expert Correlation Tables	Validate whether identified relationships are technically plausible and consistent with practical PVC processing experience.
2. Engineering Rules	Constrain recommendations to safe, meaningful and technically feasible operating windows.
3. Historical Production Cases	Support recommendations with comparable past situations and proven solutions.
4. Technical Documentation	Provide additional context from reports, manuals, presentations and publications.
5. DMAIC Project Knowledge	Use verified root causes, corrective actions and control plans from previous improvement projects.
6. Production & Quality Data	Confirm the recommendation with actual process behavior, quality results and operating history.

This additional validation layer improves explainability, increases user confidence and helps reduce the risk of unsupported AI recommendations.

Example KPI Dashboard & Process Analytics

The KPI dashboard is intended to make process performance visible and measurable. The dashboard values below are illustrative examples of key production and quality parameters commonly analyzed by PVC extrusion experts and continuous improvement teams.

KPI: Example OEE
Example / Potential: 92%

KPI: Example Quality Index
Example / Potential: 99.2%

KPI: Potential Scrap Reduction
Example / Potential: Up to -22%

KPI: Potential Startup Optimization
Example / Potential: Up to -35%

KPI: Potential Energy Reduction
Example / Potential: Up to -12%

KPI: Example Thickness Stability
Example / Potential: ±0.03 mm

KPI: Density Optimization Example
Example / Potential: 0.44 g/cm³, with trend direction toward lower density where technically feasible

KPI: Example Surface Quality
Example / Potential: 96.7%

KPI: Process Stability Monitoring
Example / Potential: Trend, drift and run-rule monitoring

KPI	Example / Potential
Example OEE	92%
Example Quality Index	99.2%
Potential Scrap Reduction	Up to -22%
Potential Startup Optimization	Up to -35%
Potential Energy Reduction	Up to -12%
Example Thickness Stability	±0.03 mm
Density Optimization Example	0.44 g/cm³, with trend direction toward lower density where technically feasible
Example Surface Quality	96.7%
Process Stability Monitoring	Trend, drift and run-rule monitoring

Potential Business Impact & ROI

HMR-EPI is positioned as a business-impact platform, not as a dashboard-only solution. The system targets measurable improvements in productivity, material efficiency, startup performance, process stability and quality consistency.

Impact Area: Productivity improvement
Potential Improvement Range: Up to +15%

Impact Area: Scrap reduction
Potential Improvement Range: Up to -22%

Impact Area: Startup time reduction
Potential Improvement Range: Up to -35%

Impact Area: Energy consumption
Potential Improvement Range: Up to -12%

Impact Area: Process stability
Potential Improvement Range: Up to +30%

Impact Area: Typical ROI range
Potential Improvement Range: 3–12 months

Impact Area: Annual improvement potential
Potential Improvement Range: Up to €1.2 million per line and year

Impact Area	Potential Improvement Range
Productivity improvement	Up to +15%
Scrap reduction	Up to -22%
Startup time reduction	Up to -35%
Energy consumption	Up to -12%
Process stability	Up to +30%
Typical ROI range	3–12 months
Annual improvement potential	Up to €1.2 million per line and year

Commercial Model (Illustrative)

The following commercial model is an indicative internal discussion basis and may be adapted depending on customer scope, number of lines, data availability, cybersecurity requirements and implementation effort.

Package: Pilot
Typical Scope: 1-2 lines, selected business case, initial data integration and analytics
Illustrative Price Range: €50k–80k
Support / Maintenance: €10k–15k/year

Package: Professional
Typical Scope: 3-5 sheet lines, broader process and quality intelligence
Illustrative Price Range: €100k–250k
Support / Maintenance: 15–20% per year

Package: Enterprise
Typical Scope: 6-10 lines, multi-line comparison, advanced analytics and integration
Illustrative Price Range: €250k–500k
Support / Maintenance: 15–20% per year

Package: Corporate
Typical Scope: Multiple plants, enterprise architecture, corporate knowledge base
Illustrative Price Range: €500k–1M+
Support / Maintenance: Individual agreement

Package	Typical Scope	Illustrative Price Range	Support / Maintenance
Pilot	1-2 lines, selected business case, initial data integration and analytics	€50k–80k	€10k–15k/year
Professional	3-5 sheet lines, broader process and quality intelligence	€100k–250k	15–20% per year
Enterprise	6-10 lines, multi-line comparison, advanced analytics and integration	€250k–500k	15–20% per year
Corporate	Multiple plants, enterprise architecture, corporate knowledge base	€500k–1M+	Individual agreement

Why HMR-EPI

Benefit: Better Quality
Description: More consistent product quality and reduced variation.

Benefit: Higher Productivity
Description: Higher output, reduced downtime and better line utilization.

Benefit: Lower Costs
Description: Less scrap, lower energy consumption and better material efficiency.

Benefit: Faster Time to Market
Description: Faster process stabilization, fewer trials and faster learning cycles.

Benefit: Knowledge Retention
Description: Capture, structure and scale expert knowledge across shifts and lines.

Benefit: Sustainable Production
Description: Reduced material waste, improved energy efficiency and more stable operations.

Benefit	Description
Better Quality	More consistent product quality and reduced variation.
Higher Productivity	Higher output, reduced downtime and better line utilization.
Lower Costs	Less scrap, lower energy consumption and better material efficiency.
Faster Time to Market	Faster process stabilization, fewer trials and faster learning cycles.
Knowledge Retention	Capture, structure and scale expert knowledge across shifts and lines.
Sustainable Production	Reduced material waste, improved energy efficiency and more stable operations.

Part B

Technical Description

Formulation Intelligence

Formulation Intelligence supports the analysis and optimization of PVC resins, additives, suppliers, material specifications, batch-to-batch variations, formulation stability and cost optimization.

Resin consistency and K-value behavior
Additive interactions and stabilizer systems
Supplier comparison and batch stability
Long-term formulation performance
Material cost optimization while maintaining target properties
Feedback of quality and market performance into formulation decisions

Mixing & Blending Intelligence

Mixing & Blending Intelligence focuses on dryblend consistency and stable input conditions for the extrusion process. It monitors and analyzes the mixing sequence, timing, temperatures, energy input, cooling process, intermediate storage and dryblend quality.

Hot/cold mixer temperature profiles
Mixing duration and energy input
Cooling efficiency and dryblend discharge temperature
Dryblend density, moisture and storage behavior
Cycle-time variation and operator practices
Impact of dryblend conditions on extrusion stability and product quality

Extrusion Process Intelligence

Extrusion Process Intelligence is expected to be the most common entry point for many customers, because it directly addresses startup time, scrap reduction, density optimization, OEE improvement and process stability.

Parameter Group: Extruder
Examples: Screw speed, torque, motor load, specific energy, throughput, barrel temperatures, melt temperature, melt pressure

Parameter Group: Vacuum / Devolatilization
Examples: Vacuum levels, filter conditions, condensate behavior, pressure fluctuations

Parameter Group: T-Die / Sheet Die
Examples: Die zone temperatures, side zones, die lips, temperature uniformity, pressure distribution

Parameter Group: Downstream
Examples: Roll stack temperatures, roll gaps, line speed, puller speed, cooling section conditions

Parameter Group: Inline Measurement
Examples: Thickness, weight per area, density, surface quality, color ΔE, gloss, optical defects

Parameter Group	Examples
Extruder	Screw speed, torque, motor load, specific energy, throughput, barrel temperatures, melt temperature, melt pressure
Vacuum / Devolatilization	Vacuum levels, filter conditions, condensate behavior, pressure fluctuations
T-Die / Sheet Die	Die zone temperatures, side zones, die lips, temperature uniformity, pressure distribution
Downstream	Roll stack temperatures, roll gaps, line speed, puller speed, cooling section conditions
Inline Measurement	Thickness, weight per area, density, surface quality, color ΔE, gloss, optical defects

Quality & Market Intelligence

Quality & Market Intelligence connects inline QC data, laboratory data, SPC, customer feedback, complaints and market performance. The objective is continuous improvement across the complete value chain.

SPC and trend analysis
Cp/Cpk capability analysis
Run rules, drift and abnormal pattern detection
Complaint analysis and link to production history
Quality prediction based on process conditions
Feedback of customer and market information into formulation, mixing and extrusion parameters

Equipment Selection & Optimization

Equipment Selection & Optimization supports engineering decisions regarding line configuration, dosing concepts, extruder configuration, screw specification, co-extrusion concepts, die technology, calibration, cooling and inline measurement systems.

Extruder sizing and L/D selection
Screw design requirements
Feedblock or multi-manifold die concepts
Die width and flow concepts
Cooling and calibration dimensions
Inline measurement selection and placement
Comparison of existing and optimized line concepts based on long-term production data

Hybrid Learning Approach

HMR-EPI follows a hybrid industrial learning approach combining structured data and expert rules, AI-based analytics and unstructured knowledge sources. The intent is to create a continuously learning system with a strong engineering foundation and transparent recommendation logic.

Learning Layer: Structured Data Learning
Contribution: Uses SCADA, Historians, MES, ERP, LIMS, inline measurement systems and laboratory quality data to understand process behavior, quality performance, productivity and operational stability.

Learning Layer: Expert Learning
Contribution: Captures practical know-how from expert interviews, workshops, technical discussions, troubleshooting sessions and operational reviews.

Learning Layer: Document Learning
Contribution: Extracts and structures knowledge from technical manuals, service reports, internal reports, training materials, presentations, technical publications and historical project records.

Learning Layer: Expert Correlation Learning
Contribution: Uses expert-defined correlation tables linking defects, deviations, material properties, equipment conditions and root causes.

Learning Layer: DMAIC Learning
Contribution: Converts Define, Measure, Analyze, Improve and Control outputs into reusable knowledge, including root causes, corrective actions, control plans and verified improvements.

Learning Layer: Case-Based Learning
Contribution: Identifies similar historical production cases, comparable defects, similar quality issues and successful corrective actions.

Learning Layer: AI-Based Data Learning
Contribution: Applies pattern recognition, correlation analysis, anomaly detection, predictive analytics and recommendation generation.

Learning Layer: Continuous Learning
Contribution: Updates the knowledge base with validated outcomes, operator feedback, successful improvements and new best practices.

Learning Layer	Contribution
Structured Data Learning	Uses SCADA, Historians, MES, ERP, LIMS, inline measurement systems and laboratory quality data to understand process behavior, quality performance, productivity and operational stability.
Expert Learning	Captures practical know-how from expert interviews, workshops, technical discussions, troubleshooting sessions and operational reviews.
Document Learning	Extracts and structures knowledge from technical manuals, service reports, internal reports, training materials, presentations, technical publications and historical project records.
Expert Correlation Learning	Uses expert-defined correlation tables linking defects, deviations, material properties, equipment conditions and root causes.
DMAIC Learning	Converts Define, Measure, Analyze, Improve and Control outputs into reusable knowledge, including root causes, corrective actions, control plans and verified improvements.
Case-Based Learning	Identifies similar historical production cases, comparable defects, similar quality issues and successful corrective actions.
AI-Based Data Learning	Applies pattern recognition, correlation analysis, anomaly detection, predictive analytics and recommendation generation.
Continuous Learning	Updates the knowledge base with validated outcomes, operator feedback, successful improvements and new best practices.

AI therefore becomes one component of the Hybrid Intelligence ecosystem rather than the sole decision-making mechanism.

Expert Correlation Intelligence

One of the most unique elements of HMR-EPI is the integration of expert-defined correlation tables developed by process engineers and industry specialists with decades of practical PVC processing experience.

Knowledge Element: Expert-Defined Correlation Tables
Contribution: Document known relationships between product defects, process deviations, material properties, equipment conditions, operational practices and root causes.

Knowledge Element: Decades of PVC Processing Experience
Contribution: Capture practical troubleshooting knowledge from commissioning, production support, customer projects and long-term process optimization.

Knowledge Element: Defect-to-Cause Relationships
Contribution: Link typical defects such as density variation, mass pressure fluctuation, shark skin, surface defects, burning, thickness instability and color variation to possible root causes.

Knowledge Element: Engineering Validation Layer
Contribution: Complement statistical AI findings with expert-defined process relationships and technical plausibility checks.

Knowledge Element: Continuous Refinement
Contribution: Allow correlation structures to be refined based on new production cases, DMAIC projects and validated improvement actions.

Knowledge Element	Contribution
Expert-Defined Correlation Tables	Document known relationships between product defects, process deviations, material properties, equipment conditions, operational practices and root causes.
Decades of PVC Processing Experience	Capture practical troubleshooting knowledge from commissioning, production support, customer projects and long-term process optimization.
Defect-to-Cause Relationships	Link typical defects such as density variation, mass pressure fluctuation, shark skin, surface defects, burning, thickness instability and color variation to possible root causes.
Engineering Validation Layer	Complement statistical AI findings with expert-defined process relationships and technical plausibility checks.
Continuous Refinement	Allow correlation structures to be refined based on new production cases, DMAIC projects and validated improvement actions.

Unlike purely statistical AI systems, HMR-EPI combines data-driven findings with expert-defined engineering relationships. This significantly improves recommendation quality, operational relevance and user confidence.

Conversational Industrial Intelligence

HMR-EPI enables users to interact with accumulated organizational knowledge through natural language conversations. Users can communicate with HMR-EPI similarly to consulting an experienced process engineer, quality specialist or continuous improvement expert.

Typical User Question: Why did scrap increase last week?
How HMR-EPI Supports the Discussion: Analyzes production data, quality data, alarms, historical cases and expert correlations.

Typical User Question: Which process variables most likely caused the density variation?
How HMR-EPI Supports the Discussion: Identifies influencing variables using process data, correlations, expert knowledge and historical production behavior.

Typical User Question: Show similar historical production cases.
How HMR-EPI Supports the Discussion: Retrieves comparable cases, previous corrective actions and proven solutions.

Typical User Question: Which root causes are most likely?
How HMR-EPI Supports the Discussion: Combines AI-supported analytics with expert-defined correlation tables and DMAIC knowledge.

Typical User Question: What corrective actions would you recommend?
How HMR-EPI Supports the Discussion: Generates explainable recommendations supported by data, engineering rules and historical experience.

Typical User Question: Which DMAIC projects addressed similar issues?
How HMR-EPI Supports the Discussion: Connects the current issue with previous improvement projects, root causes, actions and control plans.

Typical User Question	How HMR-EPI Supports the Discussion
Why did scrap increase last week?	Analyzes production data, quality data, alarms, historical cases and expert correlations.
Which process variables most likely caused the density variation?	Identifies influencing variables using process data, correlations, expert knowledge and historical production behavior.
Show similar historical production cases.	Retrieves comparable cases, previous corrective actions and proven solutions.
Which root causes are most likely?	Combines AI-supported analytics with expert-defined correlation tables and DMAIC knowledge.
What corrective actions would you recommend?	Generates explainable recommendations supported by data, engineering rules and historical experience.
Which DMAIC projects addressed similar issues?	Connects the current issue with previous improvement projects, root causes, actions and control plans.

This creates a Human-Machine Knowledge Interface that makes decades of organizational expertise instantly accessible across shifts, lines and sites.

Lean Six Sigma DMAIC Integration

HMR-EPI integrates Lean Six Sigma DMAIC principles to support structured continuous improvement and systematic root cause analysis.

DMAIC Phase: DEFINE
Objective: Clearly define the problem, business impact and improvement targets.
How HMR-EPI Supports: Identifies recurring production issues, quality losses, excessive scrap, startup inefficiencies and process instabilities. Helps prioritize improvement opportunities and establish KPI-based success criteria.
Typical Output: Improvement opportunity identification; project charter inputs; KPI definition; business case estimation.

DMAIC Phase: MEASURE
Objective: Understand current process performance and establish a reliable baseline.
How HMR-EPI Supports: Collects and integrates production, process, quality, laboratory, inline measurement, MES and historical data. Provides structured current-state visibility.
Typical Output: Baseline KPI dashboard; process capability assessment; quantified current-state performance; reliable measurement foundation.

DMAIC Phase: ANALYZE
Objective: Determine true root causes of process variation and performance losses.
How HMR-EPI Supports: Combines correlation analysis, root cause analysis, Ishikawa/Fishbone logic, Pareto analysis, historical case comparison, expert correlation tables and AI-supported pattern recognition.
Typical Output: Verified root causes; critical influencing variables; process relationships; improvement priorities.

DMAIC Phase: IMPROVE
Objective: Develop and implement effective solutions.
How HMR-EPI Supports: Generates actionable and explainable recommendations for process parameter optimization, startup improvement, operating window definition, material handling, energy reduction and best practice deployment.
Typical Output: Prioritized action plans; recommended process adjustments; expected business impact; implementation roadmap.

DMAIC Phase: CONTROL
Objective: Sustain improvements and prevent regression.
How HMR-EPI Supports: Monitors KPIs, detects deviations early, maintains operating windows, standardizes best practices and captures newly generated knowledge for reuse.
Typical Output: Sustained performance improvements; continuous KPI monitoring; knowledge retention; long-term operational excellence.

DMAIC Phase	Objective	How HMR-EPI Supports	Typical Output
DEFINE	Clearly define the problem, business impact and improvement targets.	Identifies recurring production issues, quality losses, excessive scrap, startup inefficiencies and process instabilities. Helps prioritize improvement opportunities and establish KPI-based success criteria.	Improvement opportunity identification; project charter inputs; KPI definition; business case estimation.
MEASURE	Understand current process performance and establish a reliable baseline.	Collects and integrates production, process, quality, laboratory, inline measurement, MES and historical data. Provides structured current-state visibility.	Baseline KPI dashboard; process capability assessment; quantified current-state performance; reliable measurement foundation.
ANALYZE	Determine true root causes of process variation and performance losses.	Combines correlation analysis, root cause analysis, Ishikawa/Fishbone logic, Pareto analysis, historical case comparison, expert correlation tables and AI-supported pattern recognition.	Verified root causes; critical influencing variables; process relationships; improvement priorities.
IMPROVE	Develop and implement effective solutions.	Generates actionable and explainable recommendations for process parameter optimization, startup improvement, operating window definition, material handling, energy reduction and best practice deployment.	Prioritized action plans; recommended process adjustments; expected business impact; implementation roadmap.
CONTROL	Sustain improvements and prevent regression.	Monitors KPIs, detects deviations early, maintains operating windows, standardizes best practices and captures newly generated knowledge for reuse.	Sustained performance improvements; continuous KPI monitoring; knowledge retention; long-term operational excellence.

DMAIC Powered by Hybrid Intelligence

Traditional DMAIC projects often depend heavily on the availability of experienced experts and significant manual data analysis. HMR-EPI accelerates DMAIC execution by combining production data, historical knowledge, PVC processing expertise, engineering rules, expert correlation tables, technical documents, Lean Six Sigma methodologies, statistical analysis and artificial intelligence.

This Hybrid Intelligence approach enables faster problem solving, more consistent decision-making and sustainable continuous improvement while maintaining full transparency and explainability.

Example Use Case: Startup Optimization of a PVC Free Foam Sheet Line

The following example demonstrates how HMR-EPI could analyze historical startup procedures and convert the findings into explainable recommendations. Startup optimization is only one example of many possible application areas.

Product / Line Data: Product
Example: Tin-stabilized PVC free foam sheet

Product / Line Data: Finished width
Example: 1220 mm

Product / Line Data: Thickness
Example: 25 mm

Product / Line Data: Target density
Example: 0.45 g/cm³

Product / Line Data: Production environment
Example: 4 identical production lines

Product / Line Data: Extruder
Example: twinEX 135-28S, L/D 28

Product / Line Data: Output
Example: Up to 1,100 kg/h

Product / Line Data: Heating zones
Example: 5 barrel zones, 1 adapter zone, flat sheet die with upper/lower/side/lip heating zones

Product / Line Data: Sensors
Example: Melt pressure and melt temperature sensors

Product / Line Data	Example
Product	Tin-stabilized PVC free foam sheet
Finished width	1220 mm
Thickness	25 mm
Target density	0.45 g/cm³
Production environment	4 identical production lines
Extruder	twinEX 135-28S, L/D 28
Output	Up to 1,100 kg/h
Heating zones	5 barrel zones, 1 adapter zone, flat sheet die with upper/lower/side/lip heating zones
Sensors	Melt pressure and melt temperature sensors

Example data evaluated could include 24 months of production history, 386 startup procedures, 11 operators, SCADA data, density and thickness results, operator logs, downtime reports and scrap records.

Recommended Startup Phase: Thermal stabilization
Example Recommendations: Barrel zones and adapter stabilized; melt temperature 184–186°C stable for 5 minutes.

Recommended Startup Phase: Die stabilization
Example Recommendations: Upper/lower die zones 182–184°C; side zones 183°C; all die zones within ±2°C.

Recommended Startup Phase: Roll stack stabilization
Example Recommendations: Top, middle and bottom roll temperatures stable for at least 5 minutes.

Recommended Startup Phase: Screw speed ramp
Example Recommendations: Initial 8 rpm, pressure build-up 12 rpm, process stabilization 18 rpm, production ramp-up 25 rpm.

Recommended Startup Phase: Pressure stabilization
Example Recommendations: Melt pressure 150–165 bar, stability requirement ±5 bar for 5 minutes.

Recommended Startup Phase	Example Recommendations
Thermal stabilization	Barrel zones and adapter stabilized; melt temperature 184–186°C stable for 5 minutes.
Die stabilization	Upper/lower die zones 182–184°C; side zones 183°C; all die zones within ±2°C.
Roll stack stabilization	Top, middle and bottom roll temperatures stable for at least 5 minutes.
Screw speed ramp	Initial 8 rpm, pressure build-up 12 rpm, process stabilization 18 rpm, production ramp-up 25 rpm.
Pressure stabilization	Melt pressure 150–165 bar, stability requirement ±5 bar for 5 minutes.

HMR-EPI can explain that historical analysis showed increased startup scrap when screw speed was increased before pressure and temperature stabilization. The recommendation is therefore not a black-box AI instruction, but a data-supported and expert-constrained operational recommendation.

Real-Time Dashboard & KPI Analytics

The dashboard layer visualizes relevant process, quality, productivity and sustainability KPIs. Depending on available data sources, dashboards may include OEE, quality index, scrap rate, thickness stability, density trend, surface quality, energy consumption, startup time and process stability.

Real-time monitoring of critical process and quality parameters
SPC visualization including trends, limits, run rules and drift detection
Comparison across lines, shifts, operators, products and recipes
Early warning of quality risks and abnormal operating conditions
Support for daily production meetings and continuous improvement reviews

Example ROI & Savings Scenario

The following scenario illustrates the potential economic impact of AI-supported process optimization. It should be used as an illustrative reference only.

Parameter: Production output
Example Value: 1,500 kg/h

Parameter: Operating hours
Example Value: 7,000 h/year

Parameter: Annual production
Example Value: 10,500,000 kg/year

Parameter: Example material cost
Example Value: €1.60/kg

Parameter: 5% reduction of scrap/startup/process-instability waste
Example Value: 525,000 kg/year saved

Parameter: Annual material savings in this example
Example Value: €840,000/year

Parameter	Example Value
Production output	1,500 kg/h
Operating hours	7,000 h/year
Annual production	10,500,000 kg/year
Example material cost	€1.60/kg
5% reduction of scrap/startup/process-instability waste	525,000 kg/year saved
Annual material savings in this example	€840,000/year

Additional savings may result from energy reduction, increased effective production capacity, lower downtime, reduced troubleshooting time, fewer quality claims and improved product consistency.

Integration & Connectivity

HMR-EPI is SCADA-independent. It can utilize BOOM SCADA as one possible data acquisition and SPC platform, but it can also integrate with existing customer systems and third-party SCADA, Historian, MES, ERP, LIMS and QC systems.

Possible Interface: OPC UA
Purpose: Machine and process data acquisition

Possible Interface: SQL / Historian
Purpose: Historical production and process data access

Possible Interface: REST API
Purpose: Integration with digital platforms and business systems

Possible Interface: MQTT
Purpose: Event-driven industrial data transfer where applicable

Possible Interface: MES / ERP
Purpose: Orders, products, recipes, planning and production context

Possible Interface: LIMS / QC
Purpose: Laboratory results, SPC data and quality release information

Possible Interface: Inline measurement systems
Purpose: Thickness, density, color, gloss and optical defect data

Possible Interface	Purpose
OPC UA	Machine and process data acquisition
SQL / Historian	Historical production and process data access
REST API	Integration with digital platforms and business systems
MQTT	Event-driven industrial data transfer where applicable
MES / ERP	Orders, products, recipes, planning and production context
LIMS / QC	Laboratory results, SPC data and quality release information
Inline measurement systems	Thickness, density, color, gloss and optical defect data

Architecture & Security

HMR-EPI is designed for industrial environments where process know-how and production data are highly sensitive. Deployment concepts can be adapted to customer requirements and IT policies.

Security / Architecture Element: On-Premise Deployment
Description: Local installation within customer-controlled infrastructure.

Security / Architecture Element: Air-Gapped Option
Description: No direct network connection between production environment and AI system where required.

Security / Architecture Element: SCADA Independence
Description: Ability to work with BOOM SCADA, third-party SCADA or existing historians/databases.

Security / Architecture Element: Customer Data Ownership
Description: Customer retains full control over all production data and process know-how.

Security / Architecture Element: No Cloud Dependency
Description: Cloud connection is optional, not mandatory.

Security / Architecture Element: Role-Based Access
Description: Access rights can be separated for operators, engineers, quality, management and IT.

Security / Architecture Element: Audit Trails
Description: Traceability for data imports, recommendations and user interactions.

Security / Architecture Element	Description
On-Premise Deployment	Local installation within customer-controlled infrastructure.
Air-Gapped Option	No direct network connection between production environment and AI system where required.
SCADA Independence	Ability to work with BOOM SCADA, third-party SCADA or existing historians/databases.
Customer Data Ownership	Customer retains full control over all production data and process know-how.
No Cloud Dependency	Cloud connection is optional, not mandatory.
Role-Based Access	Access rights can be separated for operators, engineers, quality, management and IT.
Audit Trails	Traceability for data imports, recommendations and user interactions.

Secure Data Transfer Concept

For maximum cybersecurity and process know-how protection, HMR-EPI can operate without direct network connectivity between the production environment and the AI system.

Daily export of production data from SCADA or Historian
Storage on controlled external media or secure synchronization server
Customer-controlled physical or logical transfer
Automated import into HMR-EPI
One-way controlled data flow where required
No external data transmission and no direct AI access to machine control systems

Possible Interfaces & Connectivity

HMR-EPI is designed to integrate production, process and quality information from different systems and equipment into one centralized process intelligence platform.

Production planning, scheduling and performance data
Formulation and recipe data
Mixing and blending parameters
Extrusion process parameters
Energy consumption
Alarm and event history
SPC and quality data
Inline measurement results
Laboratory QC results
Operator and production logs
Cross-line performance comparison and long-term KPI analysis

HMR-EPI Intelligence Framework

27.1 Statistical Methods

Correlation analysis
Regression analysis
SPC, Cp and Cpk
Trend analysis
Drift detection
Run rules
Root cause analysis
Anomaly detection
Principal component analysis where applicable

27.2 Hybrid Intelligence Technologies & Knowledge Framework

Data Analytics & Machine Learning

Pattern Recognition
Supervised Learning
Unsupervised Learning
Predictive Analytics

Recommendation & Decision Support

Recommendation Engines
Best-Practice Matching
Best-of-Knowledge Recommendation Validation

Similarity & Historical Knowledge Retrieval

Similarity Search Across Historical Production Runs
Similarity Search Across Historical Production Cases
Similarity Search Across DMAIC Projects

Knowledge Intelligence

Expert Correlation Intelligence
Knowledge Retrieval from Technical Documents
Knowledge Retrieval from Technical Reports and Presentations
Historical Case-Based Reasoning

Conversational & Explainable Intelligence

Conversational Industrial Intelligence
Natural-Language Interaction
Natural-Language Explanation of Findings and Recommendations
Explainable Recommendation Generation

Continuous Learning & Knowledge Retention

Hybrid Learning Architecture
DMAIC Knowledge Retention
Continuous Organizational Learning

27.3 Turning Process Data into Intelligence

HMR-EPI can collect, contextualize, analyze and continuously learn from a wide range of process, quality, operational and business-related parameters. The following examples illustrate typical information sources available for process intelligence, root cause analysis, continuous improvement and AI-supported decision support.

27.3.1 Raw Process Parameters

Material Preparation & Dosing

Resin feed rate
Additive dosing rates
Regrind percentage
Dryblend density
Dryblend temperature
Hopper levels
Material consumption rates

Extruder & Drive System

Screw speed
Screw torque
Motor load
Motor current
Throughput
Specific throughput
Specific energy consumption (kWh/kg)
Barrel temperatures (all zones)
Barrel heating output (%)
Barrel cooling output (%)
Barrel temperature deviation from setpoint
Melt temperature
Melt pressure
Melt pressure fluctuations
Adapter temperatures
Adapter pressure
Gearbox oil temperature
Gearbox oil pressure
Gearbox operating status

Vacuum & Devolatilization

Vacuum level
Vacuum stability
Vacuum fluctuations
Condensate temperature
Condensate level
Filter condition
Vacuum pump operating status

Die & Melt Distribution

Die temperatures (all zones)
Upper die temperatures
Lower die temperatures
Side zone temperatures
Die lip temperatures
Die pressure
Die pressure drop
Melt distribution uniformity
Left/right temperature deviation
Edge-to-center temperature deviation

Downstream & Calibration

Roll stack temperatures
Individual roll temperatures
Roll gaps
Roll synchronization
Roll speed
Puller speed
Line speed
Calibration temperatures
Vacuum calibration parameters
Cooling section temperatures
Sheet exit temperature

Product Quality & Inline Measurement

Sheet thickness
Thickness profile
Density
Density profile
Weight per unit area
Surface quality index
Surface temperature
Color (ΔE)
Gloss
Optical defects
Edge quality
Flatness / warpage
Product dimensions

Quality & Laboratory Data

Laboratory QC results
SPC parameters
Cp/Cpk values
Complaint history
Customer feedback
Market performance indicators

27.3.2 Derived KPIs

In addition to raw process data, HMR-EPI automatically generates higher-level performance indicators.

Productivity & Operational Performance

OEE (Overall Equipment Effectiveness)
Availability
Performance efficiency
Quality rate
Production output
Output stability
Throughput fluctuation
Startup duration
Startup scrap
Scrap rate
Downtime duration
Alarm frequency
Mean Time Between Disturbances (MTBD)
Mean Time To Recovery (MTTR)

Process Stability & Quality KPIs

Process Stability Index
Density Stability Index
Thickness Stability Index
Pressure Stability Index
Temperature Stability Index
Quality Consistency Index
Energy Efficiency Index
Material Utilization Efficiency

Operator & Operational Intelligence

Operator performance index
Shift performance comparison
Best startup performance ranking
Manual intervention frequency
Parameter adjustment frequency
Recipe deviation frequency

27.3.3 AI-Generated Intelligence

The true value of HMR-EPI is not the collection of process data itself, but the transformation of data into actionable operational intelligence.

Examples include:

Predictive Intelligence

Quality risk prediction
Scrap risk prediction
Process instability prediction
Startup success prediction
Energy optimization recommendations

Root Cause Intelligence

Root cause probability ranking
Defect-to-cause correlation analysis
Similarity search across historical production runs
Similarity search across historical production cases
Similarity search across DMAIC projects

Knowledge Intelligence

Expert Correlation Intelligence
Knowledge retrieval from technical documents
Knowledge retrieval from service reports and presentations
Historical case-based reasoning
Best-practice matching

Conversational & Explainable Intelligence

Conversational Industrial Intelligence
Natural-language interaction
Natural-language explanation of findings and recommendations
Explainable recommendation generation
Best-of-Knowledge recommendation validation

Optimization Intelligence

Recommended operating windows
Process optimization recommendations
Startup optimization recommendations
Parameter prioritization
Continuous improvement opportunities

Closing

Closing Statement

HMR-EPI is a Hybrid Intelligence Platform for PVC manufacturing excellence. By combining production data, historical knowledge, expert correlation intelligence, technical documentation, Lean Six Sigma methodologies and artificial intelligence, HMR-EPI transforms operational information into actionable knowledge, better decisions and measurable business results.

The platform enables organizations to preserve decades of expertise, accelerate continuous improvement, reduce operational variability and make accumulated knowledge accessible through natural language interaction.

The result is explainable, engineering-grounded operational intelligence that supports sustainable improvements while maintaining maximum customer control over all production data and process know-how.

Table of Contents

Part A - Executive Overview

Part B - Technical Description

Executive Overview

Executive Summary

Executive KPI / Value Driver

Why Customers Invest

Hybrid Intelligence for PVC Extrusion

Best-of-Knowledge Principle

Hallucination Risk Mitigation

Four Integrated Intelligence Areas

Typical Customer Journey

How HMR-EPI Generates Explainable Recommendations

Additional Validation Layer

Example KPI Dashboard & Process Analytics

Potential Business Impact & ROI

Commercial Model (Illustrative)

Why HMR-EPI

Technical Description

Formulation Intelligence

Mixing & Blending Intelligence

Extrusion Process Intelligence

Quality & Market Intelligence

Equipment Selection & Optimization

Hybrid Learning Approach

Expert Correlation Intelligence

Conversational Industrial Intelligence

Lean Six Sigma DMAIC Integration

DMAIC Powered by Hybrid Intelligence

Example Use Case: Startup Optimization of a PVC Free Foam Sheet Line

Real-Time Dashboard & KPI Analytics

Example ROI & Savings Scenario

Integration & Connectivity

Architecture & Security

Secure Data Transfer Concept

Possible Interfaces & Connectivity

HMR-EPI Intelligence Framework

27.1 Statistical Methods

27.2 Hybrid Intelligence Technologies & Knowledge Framework

27.3 Turning Process Data into Intelligence

27.3.1 Raw Process Parameters

27.3.2 Derived KPIs

27.3.3 AI-Generated Intelligence

Closing Statement