Mold flow analysis is usually bought as reassurance, and read as a picture of a part filling neatly from the gate. What it actually is, is a calculation of where the melt goes, how fast, at what pressure, and where it meets itself — and its real value is that those four things are decided by geometry that is about to be cut into steel. It is not a proof that a part is good. It is a way of finding, before the cavity exists, the two or three decisions that cannot be corrected afterwards without welding.
What the analysis actually decides
A simulation does not answer “is this part good”. It answers a narrower and much more useful set of questions: where does the melt enter, in what order does the cavity fill, at what pressure does the last point fill, where do two flow fronts meet, where does the gas end up, and how unevenly does the part shrink. Each of those has a physical answer that follows from geometry, and each of them constrains a decision that is expensive to reverse.
Four decisions are effectively made by a fill analysis, whether or not anyone runs one:
Gate position and gate count. The gate is the first thing cut and the last thing you can move. Move a gate after T1 and you weld the old gate, cut a new one, re-polish the area, and re-run the trial; on a textured or class A surface the repair is visible forever. A fill analysis costs less than that repair and takes a fraction of the time.
Wall thickness in one specific place. Not the nominal wall, which is a design decision, but the local thickness where the melt has to travel furthest. If the flow ratio is marginal, the fix is often 0.2 mm in one rib rather than a bigger press — and 0.2 mm is free before the cavity is cut and impossible after.
Cooling layout, and whether the core needs a baffle, a bubbler or an insert. Cooling is 60 to 70 % of the cycle, and the core side is almost always the bottleneck. A cooling analysis is the only way to see that before the core block is drilled.
Which press the tool needs. Clamp force and injection pressure both come out of the same run. A tool specified for the wrong tonnage is a tool that runs on someone else’s machine, or runs slowly on the right one.
The commercial value sits in the asymmetry: an analysis run at design stage changes a drawing, which costs an hour of engineering. The same finding at T1 changes steel, which costs weeks. That ratio, not the accuracy of the software, is the argument for doing it.
Four analyses, four different questions
“Mold flow analysis” is used loosely to cover four separate calculations, and they do not have the same reliability. Buying “an analysis” without saying which of the four you need is how projects end up with a beautiful fill animation and no answer about warpage.
| Analysis | The question it answers | What comes out | Its weakest link |
|---|---|---|---|
| Fill | Will it fill at all, at what pressure, and where does the melt meet itself | Fill time, pressure at switchover, flow front temperature, weld line position, air traps, clamp force | The viscosity data for the actual grade |
| Pack | Will the part be dense and uniform, and where will it sink | Volumetric shrinkage map, sink depth estimate, gate freeze time, weight curve | PVT data, and the assumption about how the machine switches over |
| Cool | Will it cool evenly, and roughly how long will it take | Temperature difference between the two faces, hot spots, cycle estimate | How the circuit is represented: ideal pipes, not a real tool |
| Warp | Will it bend, and in which direction | Deflection plot, and a breakdown by cause | Shrinkage data and the stress model behind it |
The order matters, and it is not the order most reports are presented in. Fill comes first because it establishes the gate, the fill pattern and the pressure; packing and cooling both inherit that geometry; warpage needs the output of all three. A warpage result produced without a cooling analysis is a shrinkage calculation wearing a different label, and it will be wrong in a way that is hard to see.
It is also worth knowing which of the four is usually skipped. In practice, fill is almost always run, packing sometimes, cooling rarely in full, and warpage often with a simplified cooling model. The skipped one is usually the one that would have caught the problem: a part that fills perfectly and comes out bent is a cooling and shrinkage problem, and no amount of fill analysis will show it.
The inputs decide whether the answer is worth anything
Simulation is a model, and a model answers the question you asked with the data you gave it. The uncomfortable part is that the data is usually the weakest element in the chain, and errors in it are systematic rather than random: they all push the result in the flattering direction. Optimistic viscosity gives lower pressure; averaged wall thickness gives later gate freeze; ideal cooling gives a shorter cycle. A report that looks good is therefore not evidence that the tool will run well.
| Input | What actually gets entered | What happens when it is wrong |
|---|---|---|
| Resin grade | A viscosity curve, PVT data, shrinkage data, and for filled grades a fiber orientation model | Every pressure number moves by 20 to 40 %, in the optimistic direction |
| Wall thickness | The CAD, after mid-surface extraction | Thin ribs get averaged into the wall, gate freeze time comes out too long, packing looks better than it will be |
| Runner and gate | Runner sections, gate land, gate angle | Shear rate at the gate is wrong, and in a multi-cavity tool the balance is wrong with it |
| Process settings | Melt temperature, mold temperature, fill time, switchover point | The analysis reproduces your assumption rather than your machine |
| Mesh | Element type and size, elements across the thickness | Thick-to-thin transitions are smoothed; pressure peaks disappear |
The mesh is where most of the error enters
Two families of mesh are in common use. A shell or mid-surface model represents the part as a surface with a thickness attached, and it is fast: minutes rather than hours, which is why it dominates early-stage work. A three-dimensional tetrahedral model fills the volume and captures through-thickness behavior, at a cost of an order of magnitude more computing time.
The practical rule for a 3D mesh is at least six to ten elements across the wall thickness, because the temperature and shear gradients that decide skin thickness and packing live inside that dimension. Halve the element count and the frozen layer gets smeared into an average, which makes the part look easier to fill than it is. For a shell model the constraint is different: it assumes the thickness is locally uniform, so a part with a 1 mm wall running into a 3 mm boss is exactly the case where the model is least trustworthy.
Runners and gates are usually modeled as one-dimensional beam or cone elements. That is acceptable for a round runner, and wrong for a trapezoidal cold runner with a full-round gate land, which is where the shear rate calculation is most sensitive. If the report shows a gate shear rate that looks comfortable, check what gate geometry was entered before believing it.
Generic material data is the quiet failure
Every package ships with a database, and the entries in it are not the same thing as your grade. A viscosity curve is measured on a capillary rheometer at three or more temperatures and several shear rates; a database entry may be a single grade measured decades ago, or an average across a family. Polypropylene alone spans homopolymer, copolymer, high-impact and filled grades whose viscosities at the same temperature differ by a factor of two.
Two data sets are worth paying attention to. PVT data — pressure, volume, temperature — is what packing calculations run on, and it is the input most often defaulted. Shrinkage data is worse: it is normally measured on a standard plaque, often 2 mm thick, molded under standard conditions. If the part is 3.5 mm thick, or molded at a different mold temperature, the number does not transfer, and the warpage result inherits the error.
For filled grades there is a third: fiber orientation. The orientation tensor is what makes a glass-filled part behave differently along and across the flow, and it is the only reason a warpage prediction for a glass-filled part is worth anything at all.
A quick sanity check before reading any result: ask what melt temperature and mold temperature were used, and compare them with the datasheet for the grade. If the analysis was run at 200 °C for a resin that processes at 260 °C, every pressure in the report is understated and the fill is easier than reality.
Reading a fill result: the seven numbers that matter
A fill result is usually presented as an animation, which is the least useful form of it. The numbers underneath the animation are what decide the tool, and there are seven of them worth checking every time.
| Result | What it is telling you | Working limit to judge against |
|---|---|---|
| Pressure at end of fill | How hard the machine has to push | Below roughly 70 to 80 % of the pressure the press can actually deliver at that shot size |
| Pressure at switchover | The number that sizes the press and the tool structure | Cavity pressure typically 30 to 80 MPa; above 100 MPa the tool needs a different grade of steel, not a different setting |
| Flow front temperature range | Whether the melt is arriving at the same temperature everywhere | Spread within about 20 °C across the part; more than that and the two ends of the part will not behave alike |
| Shear rate at the gate | Whether the gate is being pushed too hard | Material-specific, broadly 10,000 to 100,000 s−1; the upper end risks chain scission and gate blush |
| Shear stress | Whether the surface is being damaged | Under 0.5 MPa for a technical part, under 0.3 MPa for an optical or class A surface |
| Weld line temperature | How well the two fronts will bond | Fronts meeting within about 20 °C of each other; a bigger gap means a visible, weak line |
| Air trap position | Where the gas ends up | Every trap needs a vent; a trap in the middle of a face cannot be vented at the parting line |
Read together, those seven describe a window rather than a point. The part fills, the pressure is inside the machine, the melt arrives at a consistent temperature, the surface is not being over-sheared, and every place the gas ends up can be vented. A result that satisfies six and breaks one is not a good result; it is a warning with a nice picture attached.
Pressure at switchover is the number that chooses the press
Switchover — the moment the machine stops filling at velocity and starts packing at pressure — happens at 95 to 98 % of the cavity volume. The pressure at that instant is the highest the cavity sees, and it is the one to compare against machine capability and against the tool’s structure.
Two mistakes are common. The first is comparing the wrong pressure: packages can report pressure at the nozzle, at the sprue, at the gate, or averaged over the cavity, and the four differ substantially because of the pressure drop in the runner. On a long runner the gate pressure can be half the nozzle pressure, so a report showing a comfortable number at the nozzle may be describing a gate that is running at the material’s limit.
The second is forgetting what the number is for. Cavity pressure multiplied by projected area gives clamp force, and that is the calculation that sizes the press — the same one set out in clamp force and tonnage. A fill analysis and a tonnage calculation arriving at different numbers means one of them used the wrong area or the wrong pressure, and it is worth finding out which before the tool is booked onto a machine.
Above roughly 100 MPa of cavity pressure the constraint stops being the press and becomes the tool: plate deflection, support pillars, and the steel grade of the cavity insert, which is a subject in its own right in mold steel selection.
A predicted weld line is a location, not a strength
Fill analysis reports where two flow fronts meet and at what temperature each arrives. That is genuinely useful: it tells you whether the line will be visible and roughly how well it will bond. What it does not tell you is whether the part will break there, because strength at a weld line depends on things the model does not contain — whether the front carried degraded material, whether the vent at that point was open, whether the resin is filled, and whether there is moisture in it.
The working threshold is the temperature difference between the two fronts. Within about 20 °C they merge into a largely homogeneous melt and the line is cosmetic. Beyond that a cold skin forms before the bond does, and the line is a structural feature. Placing a weld line in a non-critical area is a design achievement; placing it where the part is loaded and hoping is not.
For parts where the weld line matters, the sequence is: move the gate if the geometry allows, add a flow leader or a overflow well if it does not, and only then accept the line and design around it. The related geometry constraints are collected in the DFM checklist.
Packing: why weight stops rising
Packing is where most of the confusion in molding lives, and the analysis makes it visible. After switchover the cavity is geometrically full but the material in it is still shrinking as it cools; pack pressure pushes more melt in through the gate to compensate. That flow continues only while the gate is open. Once the gate freezes, no further material can enter, and the part’s weight is fixed no matter what the machine does next.
This produces the single most useful curve in the whole analysis: part weight against pack time. It rises, flattens, and stops. The point where it flattens is the gate seal time, and packing beyond it adds cycle time and no weight. A tool running a 12 s pack on a 6 s gate seal is losing six seconds on every shot for nothing, which at a million shots is the difference between one machine and two.
The packing result also shows the volumetric shrinkage map, and this is where sink is decided. Sink appears where the local shrinkage cannot be fed, which means thick sections far from the gate. That is a geometry and gate-position statement, not a process statement, and no amount of hold pressure fixes a thick boss that has no gate nearby.
Two assumptions in every packing analysis are worth knowing about. The first is that switchover happens by volume at a defined percentage; real machines switch over by screw position, by time, or by cavity pressure, and each gives a slightly different result. The second is that the PVT relationship is correct for the grade, which brings the discussion back to material data. The effect of getting packing wrong on the finished part is covered in tolerances and surface finish, because shrinkage variation is what a tolerance band actually has to absorb.
Cooling: the least reliable of the four
Cooling is 60 to 70 % of the cycle, which makes it the most valuable thing to analyze and the hardest to model honestly. The standard approach represents each circuit as a pipe with a fixed coolant temperature and a heat transfer coefficient derived from flow rate. That is a reasonable abstraction of a new, clean, correctly plumbed tool, and it is not what the tool will be after a year of production.
Three things are routinely missing from a cooling model. Fouling is the first: scale on the channel wall reduces heat transfer by 20 to 40 % without changing the channel diameter measurably, and nothing in a standard analysis accounts for it. Flow distribution is the second: a model with parallel circuits assumes each gets its share of the flow, when in reality the shortest path with the least resistance takes most of it and the far circuit runs nearly stagnant. Insert and slide interfaces are the third, and they are thermal resistances that the model usually represents as perfect contact.
What the analysis is genuinely good at is the comparison it enables: this layout against that layout, with the same assumptions on both sides. A core with a baffle against a core with a bubbler, measured by the temperature difference across the part rather than by an absolute cycle time, is a reliable comparison even when neither absolute number is right. The design rules that the analysis is testing are set out in mold cooling design.
The output to look at is not the cycle time estimate. It is the temperature difference between the cavity face and the core face at the same point in the part. Above roughly 10 °C the two sides shrink at different rates and the part bends away from the hot side; that single number predicts the direction of most warpage before the warpage analysis is run.
Warpage: right about direction, loose about size
Warpage analysis is the one people most want and trust least, and both instincts are reasonable. The prediction of direction is usually right: a part bends away from the hotter face, towards the side with more shrinkage, and away from the region with higher orientation. The prediction of magnitude is typically within 20 to 50 %, which is fine for choosing between two designs and not fine for promising a flatness tolerance to a customer.
Warpage comes from four causes, and the analysis separates them, which is the most useful thing it does:
Differential shrinkage from orientation. The dominant cause in filled materials and in long thin flow paths. Glass fibers align with the flow and shrink far less along it than across it; the part then bends as the two directions try to reach different dimensions.
Differential cooling. One face hotter than the other, so one skin solidifies later and shrinks more. This is a cooling design problem, and its size is the temperature difference described above.
Differential pressure. A part packed from one end has a pressure gradient along it, and the far end is less dense than the near end. Long parts in a single gate show this as a bow.
Corner and rib effects. Material at a corner cools from two directions and shrinks differently from material in a flat wall, which is why box-shaped parts pull their sides in.
What to do with the result is a hierarchy rather than a tweak. Uniform wall comes first, because no gate or cooling arrangement saves a part whose thick and thin sections shrink differently. Gate position comes second, because it sets the pressure and orientation pattern. Cooling balance comes third, and packing profile last. Adjusting the packing profile to chase a warpage number is the classic wasted week: it moves the number without changing the mechanism.
Eight things simulation will not tell you
Knowing the boundary of the method is more useful than knowing its accuracy claims, because the boundary is where projects get into trouble.
1. Cosmetic defects below the eye’s threshold. Flow marks, gloss bands, silver streaks, gate blush and jetting are surface phenomena at a scale finer than a standard mesh. A model can say the shear stress is high at the gate; it cannot say whether the customer will reject the part.
2. Steel deflection under clamp force. The model assumes a rigid tool. A large cavity plate with insufficient support bows, the parting line opens in the middle, and the flash that results is a structural problem no flow analysis contains.
3. Vent condition. The model will place air traps precisely, and cannot know that the vent at that point is blocked, polished shut, or nonexistent. Venting is a separate discipline with its own failure modes, covered in mold venting.
4. Machine dynamics. Screw recovery time, check valve leakage, accumulator response and injection unit limits are outside the model. A pressure curve that is achievable in theory can be unachievable on a specific press with a worn check ring.
5. Resin variation. Lot-to-lot differences, regrind content, masterbatch carrier, and moisture level all shift viscosity. A model run on virgin dried pellets describes the first hour of a production run, not the third shift.
6. Ejection and sticking. Draft, ejection force and part distortion on ejection are mechanical, not rheological, and are absent from the result.
7. The tool as built. A cavity cut 0.05 mm off nominal changes the wall thickness and therefore the fill. Simulation describes the drawing, not the steel.
8. The process as run. Every result is conditional on the settings entered. If the molder runs 15 °C hotter to fix a surface defect, the pressure picture changes and the analysis no longer applies.
When it is worth paying for
Analysis is not free, and running it on every part is as unprofessional as running it on none. The decision is not about part complexity but about whether a finding can still change something cheap.
| Part or tool situation | Worth running | Why |
|---|---|---|
| Visible housing, class A surface | Fill and pack, always | Weld lines, gate blush and sink are the defects a customer sees first, and they are set by gate position |
| Wall under 1 mm or flow ratio above 200 | Fill, always | The part is pressure-limited: either it fills inside the machine’s capability or it does not |
| Section thicker than 4 mm | Pack and cool | Sink and cycle time both come from the thick section, and both are decided by cooling layout |
| Glass or mineral filled, structural | Warp, always | Orientation drives differential shrinkage, and differential shrinkage drives the bend |
| More than four cavities | Fill, for balance | Runner balance is a calculation, not a guess, and an unbalanced tool runs below its cavity count |
| Family tool, different parts in one shot | Fill, always | Parts of different volume reach their gates at different pressures and fill at different rates |
| Simple bracket, two cavities, PP, 2 mm wall | Usually not | The rules of thumb are reliable in this range, and a bracket is cheap to iterate on parts |
| Aluminum bridge tool for a few thousand parts | Fill only, if anything | The tool is meant to be changed; analysis is better spent on the production tool that follows |
The pattern behind the table is that analysis earns its place where the number of things that can go wrong is larger than the number of trials you can afford. A visible housing has gate position, weld lines, sink and gloss all interacting; a structural filled part has orientation and warpage; a 16-cavity tool has balance. A two-cavity bracket has none of those interactions, and a competent toolmaker’s judgment plus one trial is faster and better.
There is a second case that the table does not show: a part that has already failed. When a tool is producing out-of-tolerance parts and the cause is not obvious, a fill and packing analysis run against the real process settings is often the cheapest way to distinguish a tool problem from a process problem, because it shows whether the observed behavior is geometrically possible. That diagnostic use is underrated, and it usually takes days rather than weeks.
How to brief it so the report is usable
Most disappointing analyses are disappointing because of the brief, not the software. A request for “a mold flow analysis” gets a fill animation. A request that specifies the decisions the analysis must support gets a document that changes a drawing.
The brief should state four things before anything is run: which resin grade with which data source; which of the four analyses are required; the process window to be used, including the range rather than a single point; and the specific question the report has to answer, in one sentence. “Will this fill on a 250-tonne press with a single gate at the center” is a brief. “Check the molding” is not.
| Ask for | Why it matters | Red flag |
|---|---|---|
| The exact grade name and the source of its data | Pressure results are only as good as the viscosity curve | “Generic PP” or “similar to PC” |
| Mesh type, element count across the wall | Thick and thin features are where meshes fail | A shell model of a part with a 3 mm boss on a 1 mm wall |
| The four process inputs written down | Every result is conditional on them | A report with no melt temperature stated |
| Numbers as well as plots | A picture cannot be checked or compared | Only rainbow fill images |
| Section cuts through the thickness | Temperature and shrinkage gradients are through-thickness | Only surface contour plots |
| A rerun at plus or minus 10 °C mold temperature | Shows whether the conclusion is robust or borderline | A single run presented as fact |
| What was assumed about the runner and gate | It is often simplified or ignored | Runner geometry missing entirely |
| The recommended change and its tool consequence | An analysis that does not change a drawing has no value | “The part will fill” as the only conclusion |
One item on that list deserves emphasis: the sensitivity rerun. A result that holds when the mold temperature moves 10 °C is a design that works in a real factory; a result that flips from acceptable to marginal over that range is a design that will work in one plant and fail in another. Asking for the rerun costs very little at the analysis stage and reveals more about robustness than any single result.
Reconciling the prediction with the trial
An analysis is a hypothesis, and the mold trial is the experiment. The value of running both is that the comparison tells you which of your assumptions was wrong, and that knowledge transfers to the next tool. Most shops run both and compare neither, which means they pay for the analysis twice.
| What was predicted | What T1 shows | What the gap means |
|---|---|---|
| Fill time 1.2 s | 1.2 s | The process window agrees; the model is calibrated and the rest of the report can be trusted more |
| Switchover pressure 85 MPa | 110 MPa | Viscosity data was optimistic, or the runner and gate are undersized. Check the gate first: it is cheaper to open than to change the press |
| Part weight 42.3 g | 41.1 g | Packing is not reaching the cavity, or switchover happens earlier than modeled |
| Warpage 0.6 mm | 1.1 mm | Cooling is less balanced than modeled, or shrinkage data does not describe this grade. Measure both faces before changing the steel |
| Weld line at the boss | There, and it breaks | Location was right, strength was not. Venting or degraded melt, not geometry |
| No air traps reported | Burns at the end of a rib | The trap was inside the mesh, not on the surface. Vent the last place to fill, not the last place on the drawing |
The comparison needs four numbers recorded at T1: fill time, pressure at switchover, part weight, and the dimensions that matter most on the drawing — plus, if warpage was predicted, the actual deflection measured on a flat plate or a surface table. Those six values turn a report into a calibration, and a calibrated model is worth considerably more on the second tool than the first.
One caution about reading the gap. A pressure that comes out 25 % high does not mean the analysis was useless; it means the model was optimistic in a specific, identifiable way, and the same optimism will apply to the next part in the same family. Consistency of error is more useful than a small error, because it can be corrected for.
What each result changes in the tool
The reason to read a report carefully is that each of its outputs maps onto a physical feature that is about to be made. Seeing that mapping explicitly is what turns an analysis into a revision before the steel is cut.
A high pressure at end of fill changes the wall thickness, the flow length, or the machine — in that order of preference, because thickness and gate count are cheap at design stage and a bigger press is a permanent constraint on where the tool can run. If the answer is a bigger press, the tonnage calculation should be redone rather than assumed.
A weld line in a structural area changes the gate position, and in a multi-cavity tool it changes the runner layout, which is the same decision as how many cavities the tool should have and how they are fed. Runner configuration itself is covered in hot runner versus cold runner, because a hot runner changes both the temperature at the gate and where the balance can be adjusted.
An air trap in the middle of a face changes the venting plan, and sometimes the ejection layout, because an ejector pin doubles as a vent and a pin in the wrong place is not free.
A temperature difference above 10 °C across the part changes the cooling circuit: baffles, bubblers, a beryllium copper insert, or in the extreme a conformal core. All of those are drilling decisions, and all of them are decisions that the tool cost and the lead time both depend on.
A shrinkage map that is not uniform changes what tolerance can be promised. If the prediction says the part will vary by 0.15 mm across the cavity and the drawing asks for 0.05, the drawing is the thing that has to change, and it is far better to find that out before the customer’s quality group does. A change of resin grade at that point is also possible, which is why material selection and analysis belong in the same conversation rather than in sequence.
The overall flow of decisions, from part design through to tool build, is the subject of the injection molding process overview, and the point at which a supplier should be brought in is covered in choosing an injection molding supplier.
Ten questions before you accept a report
This list is short enough to use in a meeting and specific enough to catch the reports that are decoration.
1. Which resin grade, and where did the data come from? A family name is not an answer.
2. Which analyses were run, and which were skipped? A warpage result with no cooling model is a shrinkage guess.
3. What mesh, and how many elements across the wall? Fewer than six in a 3D model and the through-thickness behavior is missing.
4. What melt and mold temperatures, and what fill time? Every result is conditional on these three.
5. Where was the switchover point set, and by what method? Volume, screw position and pressure give different answers.
6. Is the pressure reported at the nozzle, the gate, or the cavity? The difference is the runner loss, and it is often half the number.
7. Where are the air traps, and can each one be vented? A trap in the middle of a face needs a pin or an insert, not a groove.
8. What is the temperature difference across the part? Above 10 °C, warpage direction is already decided.
9. Was anything rerun at the edge of the process window? If not, the conclusion is a point estimate with no margin.
10. What should change in the drawing, and what does it cost in the tool? If the answer is “nothing”, either the part was already right or the analysis was not read.
Frequently asked questions
How accurate is mold flow analysis?
For fill pattern and weld line position, usually within a few percent, because those follow from geometry and are robust to data error. For pressure, typically within 10 to 20 % when the grade data is real and the runner is modeled properly, and much worse when a generic grade is used. For temperature difference, reasonable. For warpage magnitude, 20 to 50 % is a realistic expectation, with the direction of the bend far more reliable than its size. The practical implication is to trust it for choosing between designs and not for promising a number to a customer.
Do I need mold flow analysis for a simple part?
Not usually. Simple means the interactions that analysis exists to resolve are absent: one gate, uniform wall, no cosmetic requirement, a material whose behavior is well known, and a cavity count low enough that balance is not a question. A two-cavity PP bracket at 2 mm falls in that category, and a trial is cheaper than a study. The threshold is not complexity in the abstract but whether a finding could still change a decision before the steel is cut.
How long does an analysis take?
A single fill analysis on a mid-surface model of a moderate part is a matter of days including setup and interpretation, and the computing itself is often minutes. A full package with 3D mesh, cooling and warpage is typically one to two weeks, with most of that being geometry cleanup and interpretation rather than solving. The long pole is usually waiting for the customer to answer the four questions the analysis needs: which grade, which process window, which press, and what has to be true for the part to be accepted.
Can mold flow analysis predict cycle time?
It can predict the cooling portion, which is 60 to 70 % of the cycle, and it will usually be optimistic because it assumes a clean circuit and perfect thermal contact at inserts. What it does not model is the rest: plastication time, ejection, any robot motion inside the open time, and the safety margin an operator adds. Treat a predicted cycle as a floor and add the non-cooling elements from experience. A tool specified against a predicted cycle with no allowance runs late from the first shift.
Is 3D analysis better than 2.5D?
More accurate and slower, and the difference matters mainly where the thickness varies a lot or where through-thickness gradients decide the result. For a uniform-wall part with a straightforward gate, a mid-surface model gives essentially the same answer in a fraction of the time. For a part with a thick boss on a thin wall, a gas-assisted section, or a large thickness step, the shell assumption fails exactly where the problem is, and 3D is worth the wait. The wrong choice in either direction wastes either accuracy or time.
Can I skip the trial if the warpage prediction is within tolerance?
No. The prediction assumes the tool is made to the drawing, the resin is dry and within specification, the cooling circuit is clean and balanced, and the process runs at the modeled settings. None of those is guaranteed, and all four are things the trial exists to establish. What a good warpage prediction buys is a lower probability of a surprise, and sometimes the confidence to cut steel without a pre-hardened interim tool. It does not replace the first-off inspection.
Should the molder or the toolmaker run the analysis?
Whoever has the better material data and the more honest process settings. The molder usually knows the real process window and the machine that will run the job, which makes their analysis more realistic; the toolmaker usually knows the runner and cooling geometry that will actually be built, which makes their model more representative. The workable arrangement is a joint one with the assumptions written down, so that when the trial disagrees with the prediction there is a record of which input was wrong.
What is the single most common mistake in a flow analysis?
Using a generic material entry. It is the cheapest thing to get wrong and it contaminates every number downstream: pressure, shear, gate freeze, shrinkage and warpage all derive from the viscosity and PVT behavior of the actual grade. The second most common is not modeling the runner, which makes the gate pressure and the multi-cavity balance meaningless. Both are avoidable with one question each, which is why they are on the checklist above.
