[{"data":1,"prerenderedAt":155},["ShallowReactive",2],{"page-content:en:blog\u002Fspringback-prediction-experiment":3},{"meta":4,"hero":10,"body":13,"article":118,"references":127,"related":135,"cta":151},{"path":5,"title":6,"seoTitle":7,"description":8,"updated":9},"\u002Fblog\u002Fspringback-prediction-experiment\u002F","What a three-point bending experiment shows about springback prediction","Springback Prediction: What an Air-Bending Experiment Found | JS Ragos","A 2023 open-access study of three-point air bending measured springback and load on clad AW-2024 at 1.0 and 2.0 mm. What its numbers say about the ram-depth table on a press brake.","2026-11-13",{"kicker":11,"title":6,"subtitle":12},"Blog \u002F Paper reading","The angle formula every shop uses ignores tool radius and thickness. A measured correction exists, and the model you pick matters more than the numbers you type in.",[14,18,21,23,26,47,49,52,54,56,59,61,63,66,68,70,73,75,77,80,88,102,105,107],{"type":15,"text":16,"id":17},"h2","The short version","section-1",{"type":19,"text":20},"p","A press brake control predicts the angle from ram depth using a geometric relation, a material model and a springback correction. All three are assumptions, and a 2023 experiment on three-point air bending put numbers on how wrong each one gets. The headline results: the classic angle-from-deflection formula drifts as deflection grows; the cladding on the sheet doubled the load-prediction error when its share of thickness rose from 5.5% to 10%; and among five numerical models applied to the same measured data, the best one for load was not the best one for springback.",{"type":19,"text":22},"That last point is the one to carry into a purchase decision. A control that predicts load well is not automatically a control that predicts angle well. They are different models fitted to different measurements.",{"type":15,"text":24,"id":25},"What was tested","section-3",{"type":27,"items":28},"key",[29,32,35,38,41,44],{"label":30,"value":31},"Material","AW-2024 aluminium sheet clad both sides with AW-1050A",{"label":33,"value":34},"Thickness","1.0 mm and 2.0 mm as received, after softening annealing",{"label":36,"value":37},"Cladding","50 µm on 1.0 mm (10.0% of thickness); 55 µm on 2.0 mm (5.5%)",{"label":39,"value":40},"Specimens","40 × 100 mm, cut parallel to the rolling direction",{"label":42,"value":43},"Tooling","Three-point bending, tool radius r = 5 mm, support spacing w = 50 and 100 mm",{"label":45,"value":46},"Cycle","Deflection 0–20 mm, run as one load cycle and as eight cycles",{"type":19,"text":48},"The setup is close to a press brake in the only way that matters: a punch pushes a strip between two supports, the strip bends over a radius, and the angle after unloading is not the angle under load. A three-point lab fixture is not a V-die, but the measurement problem is the same one a ram-depth table is trying to solve.",{"type":15,"text":50,"id":51},"Result 1: the angle formula drifts with deflection","section-5",{"type":19,"text":53},"The standard way to get the bend angle from a three-point test is the arctangent of deflection over half-span. It assumes the strip stays in contact with the supports at a fixed point and ignores the radius of the tool and the thickness of the strip. Both assumptions fail as the deflection grows, because the contact point migrates along the radius.",{"type":19,"text":55},"The authors replaced it with a form that carries a coefficient A depending on the thickness-to-radius ratio t\u002Fr and the support spacing w: the angle is the arctangent of 2f divided by (w − A·f). The measured angles matched this form noticeably better than the classic one, and the gap widened at larger deflection. For a press brake this is the same warning in a different geometry: the further the ram travels into the stroke, the less a simple depth-to-angle relation describes what the sheet is doing.",{"type":15,"text":57,"id":58},"Result 2: the coating changed the answer more than the model did","section-7",{"type":19,"text":60},"Half the point of the paper is the cladding. At 1.0 mm the cladding is 10% of the thickness; at 2.0 mm it is 5.5%. A model that ignores the clad layers predicted bending load with an error below 1% only when the layers were included; excluding them pushed the error up. Increasing the cladding share from 5.5% to 10% roughly doubled the relative error of the maximum bending load across the models tested, and quadrupled it for one of them.",{"type":19,"text":62},"The shop translation is direct. A plated, clad, galvanised or laminated sheet is not the base alloy with decoration on the surface. Its outer fibres are a different material with a different yield point, and the bend reads that. If the control is set up for bare 2024 and the job runs Alclad, the angle will be off in a way no amount of backgauge adjustment fixes.",{"type":15,"text":64,"id":65},"Result 3: the best model for load was not the best for springback","section-9",{"type":19,"text":67},"Five numerical models were compared against the measurements. For bending load, the two-dimensional model that included the cladding layers came closest, under 1% relative error for both thicknesses. For springback, the best performer was a three-dimensional shell model using the Huber–von Mises isotropic condition, with relative error between 0.96% and 3.24% for the 1.0 mm sheet across the bending-curvature range.",{"type":19,"text":69},"One to three percent sounds small until it is placed on a flange. A 90° bend with a 3% angle error is roughly 2.7° out, which is far outside any general tolerance on a formed part. The experiment is careful, the model is the best of five, and the residual is still larger than the tolerance. That is what prediction means in practice: a starting point for the first trial bend, not a substitute for measuring it.",{"type":15,"text":71,"id":72},"Result 4: springback was linear in relative deflection","section-11",{"type":19,"text":74},"For both thicknesses, the springback factor K plotted against the ratio of support spacing to deflection came out as a straight line with a high correlation. That is the mechanical justification for the way shops actually work: a correction table built from a few trial bends is valid across a window, because the underlying relation is close to linear inside that window.",{"type":19,"text":76},"The experiment also ran one cycle and eight cycles. The load curves nearly coincided, which says the elastic unloading is repeatable — the material is not drifting between strokes. If a press brake loses angle across a shift on the same program and the same coil, the cause is not this kind of instability; look at the tooling, the material batch or the machine.",{"type":15,"text":78,"id":79},"What to do with the ram-depth table","section-13",{"type":81,"items":82},"ul",[83,84,85,86,87],"Treat the control's table as a calibration, not a law. It was fitted on some material, at some thickness, with some tooling. Change any of the three and the table is extrapolating.","Record the material state on the drawing, not just the grade. Clad, coated and laminated sheets bend differently from the bare alloy, and the paper quantifies how differently a model reacts to a 5.5% versus 10% clad share.","Keep the tolerance and the prediction error in the same conversation. A model with a 1–3% angle residual cannot be the acceptance criterion for a part with a tighter angle tolerance.","Use trial bends to build a local table, then verify it at the ends of the range you actually run. Linearity holds inside the measured window; nobody has promised it holds outside.","When a supplier quotes an accuracy figure, ask which quantity it refers to — load, angle before springback, or final angle. The paper shows those are predicted with different models and different errors.",{"type":19,"text":89},[90,91,96,97,101],"How springback is compensated on a press brake, and why the compensation needs measurement, is covered in ",{"key":92,"href":93,"role":94,"label":95},"springback","\u002Fblog\u002Fspringback-and-angle-compensation\u002F","supplement","the springback and angle compensation article",". The tool side of the same problem is in ",{"key":98,"href":99,"role":94,"label":100},"vdie","\u002Fknowledge\u002Ftooling\u002Fv-die\u002F","the V-die knowledge page",".",{"type":15,"text":103,"id":104},"The source","section-15",{"type":19,"text":106},"The study is open access and the data are reported in tables, which is why it is worth citing rather than summarising loosely. If you use this reading, cite the paper for the measurements and the model comparisons; the interpretation above is ours.",{"type":27,"items":108},[109,112,115],{"label":110,"value":111},"Source","Materials 2023, 16(8), 2945 — three-point air bending of clad AW-2024",{"label":113,"value":114},"DOI","10.3390\u002Fma16082945",{"label":116,"value":117},"Measured","Bending load vs deflection; springback factor vs relative deflection; 1.0 and 2.0 mm",{"publishedAt":9,"updatedAt":9,"author":119,"authors":120},"RAGOS Engineering Team & Yaxiio Engineering Alliance",[121,124],{"name":122,"href":123},"RAGOS Engineering Team","\u002Fteam\u002F",{"name":125,"url":126},"Yaxiio Engineering Alliance","https:\u002F\u002Fwww.yaxiio.com\u002Fen\u002Feditorial-policy",{"title":128,"items":129},"Sources",[130],{"id":131,"title":132,"url":133,"note":134},"A1","On the Springback and Load in Three-Point Air Bending of the AW-2024 Aluminium Alloy Sheet with Rolled AW-1050A Cladding","https:\u002F\u002Fwww.mdpi.com\u002F1996-1944\u002F16\u002F8\u002F2945","Open-access article, Materials 16(8):2945, 2023. Read 2026-11-13: abstract, experimental procedure, results and conclusions. Values quoted above are from the reported tables and text.",[136,139,142,145,148],{"label":137,"path":138},"Bend deduction calculator","\u002Fcalculators\u002Fbend-deduction\u002F",{"label":140,"path":141},"Press brake tooling","\u002Fknowledge\u002Ftooling\u002F",{"label":143,"path":144},"Press brakes","\u002Fpress-brake\u002F",{"label":146,"path":147},"Machine acceptance","\u002Fguides\u002Fmachine-acceptance\u002F",{"label":149,"path":150},"Back to the blog","\u002Fblog\u002F",{"heading":152,"text":153,"button":154},"Chasing an angle across a material change?","Send the material, the coating and the target angle. We will tell you where the ram-depth table stops being valid and what the trial bend should measure.","Contact us",1794847192546]