Using AI to prepare press felt parameters for an offer is a significant shift from “experience-based” guessing to Data-Driven Engineering. In this context, the AI acts as a bridge between the paper machine‘s mechanical constraints and the felt’s structural design.
Here is how you can use AI to automate and optimize the preparation of those parameters:
1. Automated Machine Data Extraction (OCR & NLP)
When you receive machine specifications (PDFs, handwritten maintenance logs, or technical drawings), you can use AI to extract parameters instantly.
Computer Vision (OCR): Extracts nip pressure, roll diameters, machine speed, and vacuum box positions from legacy documents.
Natural Language Processing (NLP): Categorizes the “pain points” mentioned in customer emails (e.g., “vibration issues,” “marking,” or “short life”) to prioritize specific felt characteristics.
2. Predictive Parameter Modeling
Instead of looking up a table, you can use Supervised Learning models (like Random Forest or XGBoost) trained on your company’s historical sales and performance data.
Input Variables: Paper grade (e.g., Tissue, Board), machine speed ($m/min$), nip load ($kN/m$), and press configuration (e.g., Shoe press vs. Roll press).
AI Output: It calculates the optimal Target Parameters for your offer:
Base Cloth Weight: To handle the mechanical load.
Void Volume: To ensure sufficient water handling at high speeds.
Air Permeability ($cfm$): Calculated based on the vacuum capacity of the customer’s machine.
3. The “Optimal Match” Algorithm
Understanding the Role of the Paper Machine in Press Felt Performance
You can use a Recommendation Engine (similar to what Netflix uses, but for industrial specs) to compare the current machine information against a database of successful “runs.”
Similarity Scoring: The AI finds the “closest match” machine in your global database. If a felt with Parameter X worked perfectly on a similar machine in another mill, the AI suggests that specific design for the new offer.
Gap Analysis: If the new machine is 10% faster than your previous best-performing reference, the AI uses Regression Analysis to adjust the felt density and batt fineness accordingly.
4. Cost-to-Performance Optimization
AI can help you “engineer to a price point” during the offer phase.
Multi-Objective Optimization: The AI balances the cost of raw materials (e.g., expensive specialized polyamides) against the performance guarantees you are offering.
Benefit Simulation: You can use the AI to generate a report for the customer showing the ROI—predicting exactly how much energy ($steam$) they will save due to the improved dewatering of your proposed felt design.
Workflow for Preparing an Offer with AI
| Step | Action | AI Tool/Technique |
| 1. Intake | Input machine speed, press type, and paper grade. | Data Integration Layer |
| 2. Analysis | Compare data against historical “Success Stories.” | K-Nearest Neighbors (KNN) |
| 3. Design | Calculate GSM, Permeability, and Compressibility. | Neural Networks / Regression |
| 4. Validation | Check if the design exceeds machine safety limits. | Physics-Informed AI |
| 5. Output | Generate the technical data sheet for the offer. | Automated Report Generation |
How to Start Building This
To do this, you don’t need a “general” AI like ChatGPT; you need a Custom ML Model:
Clean your data: Gather 5–10 years of your previous felt designs and the machine data they ran on.
Label the results: Mark which designs were “Great,” “Average,” or “Failed.”
Train the model: Use this data to teach the AI the relationship between Machine Info and Felt Success.
