Extrusion International 5-2026

33 Extrusion International 5/2026 in stability, consistency and scrap reduction across thou- sands of production hours. On a typical HDPE pipe ex- trusion line processing around 900 kg/h, even a 1% reduction in scrap can represent several tonnes of mate- rial saved each month and tens of thousands of pounds in annual savings. The challenge is rarely developing the AI model itself. In most cases, the biggest obstacle is obtaining reliable, high-quality process data from existing equipment, par- ticularly on older extrusion lines where sensor coverage and data collection capabilities may be limited. Lessons from Other Industries The extrusion sector is not the rst manufacturing in - dustry to explore predictive process control. Steel pro- ducers, semiconductor manufacturers and automotive component suppliers have spent years developing ma- chine-learning systems capable of identifying process deviations before defects occur. A useful lesson from these industries is that the great- est bene ts often come from solving relatively small problems. Manufacturers have reported signi cant annual savings simply by reducing process variation enough to lower scrap rates by a small percentage. This highlights an important reality: AI projects do not need to deliver dramatic breakthroughs to generate substan- tial commercial value. In high-volume extrusion operations, a one-percent improvement in ef ciency rarely attracts headlines. Yet over months and years of production, small gains accu- mulate into substantial reductions in material consump- tion, downtime, energy costs and quality-related losses. Understanding Self-Optimising Extrusion Lines One of the most ambitious applications of AI is the development of self-optimising extrusion systems. The term is sometimes misunderstood. It does not mean the extrusion line operates without people. Instead, it re- fers to a manufacturing system capable of continuously adjusting process conditions to achieve speci c perfor - mance objectives. A self-optimising line combines several technologies: • Real-time process data collection • Advanced process models • Machine learning algorithms • Automated control systems • Feedback from quality measurement devices The system continually compares current operating conditions against production targets such as: • Throughput rate • Product dimensions • Surface quality • Mechanical performance • Energy consumption • Scrap generation If performance begins to drift, the AI model deter- mines whether changes to screw speed, temperature pro les, haul-off rates, feeder settings, or cooling pa - rameters could improve the process. Unlike traditional control systems, which generally optimise one parameter at a time, a self-optimising platform evaluates the interaction between multiple variables simultaneously. Consider a pipe extrusion operation producing HDPE pressure pipe. An AI model might detect a gradual increase in die pressure combined with rising motor torque and minor dimensional variation. Historical pro- duction data may show that similar conditions frequent- ly lead to wall-thickness instability within the next 20 to 30 minutes. Rather than waiting for the deviation to occur, the system can recommend or automatically implement small process adjustments. In many cases, opera- tors may never notice that the correction has taken place because the process remains stable throughout production. The objective is not to eliminate human expertise but to provide thousands of small optimisa- tion decisions that would be impractical to perform manually. A Realistic Operational Scenario Consider a polyethylene pipe manufacturer running a continuous 24-hour production schedule. Tradition- ally, an experienced operator might notice a gradual in- crease in die pressure late in a production run and make small adjustments based on experience. An AI-assisted system could recognise that the same pressure pattern has historically led to dimensional instability within the next production hour and recommend corrective action immediately. www.battenfeld-cincinnati.com Automated Wall Thickness Control for Precise Pipe Production The new helix II IOA with electromechanical Intelligent Centering Adjustment (ICA) and Fast Dimension Change (FDC) • Reduces start-up scrap • Significantly lowers material costs IOA =Intelligent Operating Adjustment ICA =Intelligent Centering Adjustment FDC=Fast Dimension Change