Mobile: 18688976300 (Mr. Peng) 15222276387 (Mr. Bai)
Email: pengwl@honred.com
QQ： 176496342
Address：Building A1, Runteng Industrial Park, Shiyan Street, Bao'an District, Shenzhen
At a micro-screw removal station, a torque reading on its own carries very little information. The same 0.35 N·m may be a normal threadlocker breakaway, a stripped screw with the bit spinning free, or a galled thread where torque was driven up by force. Separating these cases requires watching torque across the whole rotation. That is the role of the torque-angle curve on the removal side: it converts "how much did we turn" into "what happened, and in which segment".
On the tightening side the torque-angle curve has long been standard: snug point, elastic slope, yield point, final angle. Four features settle whether a fastening is good. Removal is harder because the starting state is unknown. At assembly the initial preload is a design value; at removal it may have become anything through threadlocker cure, thermal cycling or galvanic corrosion. Look only at peak torque and a healthy screw and a galled screw can report the same number. Add angle and the two separate immediately: a healthy screw reaches peak only after several degrees and then drops sharply, while a galled screw climbs steeply within a very small rotation and does not drop at all.
The physics is direct. Once the bit starts turning, the thread pair must first overcome static friction, and torque rises steeply in that phase at a slope reflecting the contact stiffness of thread and bearing face. The peak arrives at the instant static friction breaks, which is the breakaway torque. Past the peak, torque falls to a lower plateau: the running torque, governed mainly by thread friction and residual axial load. When the screw finally leaves the thread, torque returns to zero while angle keeps advancing with no response from the workpiece. Four phases, climb, peak, plateau and zero, form the complete signature of a normal removal.
No elaborate pattern recognition is required. Five features cover nearly everything the floor produces.
The first is the ratio of breakaway torque to assembly torque. Without threadlocker and soon after assembly, the ratio typically sits at 1.1 to 1.4. With pre-applied microencapsulated threadlocker it runs 1.5 to 2.5 and correlates with cure time. Above 3, galling or corrosion should be suspected, and the probability of thread damage from forcing the screw rises markedly in that band.
The second is the angle before the peak. A healthy screw peaks within 5 to 30 degrees. Too little angle means almost no elastic reserve in the thread pair, usually galling or an already-stripped screw. Too much angle points to a thick threadlocker layer or actual bonding.
The third is the depth of the drop after the peak. A healthy screw falls to forty to sixty percent of its peak. A shallow drop means abnormal thread friction; a drop to nearly zero suggests the bit is spinning free.
The fourth is the torque ripple in the run-out segment. Normal running torque is steady within plus or minus twenty percent. Regular oscillation indicates local thread damage or debris; a continuous climb indicates a galling point that is propagating.
The fifth is the total angle at zero crossing. Effective run-out angle is approximately thread length divided by pitch, multiplied by 360 degrees. An M1.2 screw at 0.25 mm pitch with 2 mm engagement should turn roughly 2880 degrees. Measured angle well below theory means the screw did not fully exit or the thread has broken inside the hole; well above theory means it spun without advancing.
Every criterion above assumes fine data. Sample rate should be at least 1 kHz, because the breakaway peak often lasts only a few milliseconds and a slow sampler flattens it, which destroys the ratio criterion outright. Angle resolution comes from the encoder: a 1024-line encoder with quadrature decoding and a typical reduction ratio reaches the 0.1 degree order, enough to resolve differences before the peak. Torque sensor range should follow the largest variant but without excessive headroom, since an oversized range loses resolution in the low-torque region and distorts the plateau ripple criterion.
One engineering detail is easily missed: the angle zero. Angle must be counted from the moment the bit actually loads, not from motor start, otherwise idle travel is accumulated into the angle and every pre-peak criterion shifts. A workable method is to reset on a torque threshold set at two to three percent of full scale. On a desktop small screw removal equipment platform, mainstream configurations time-align the torque threshold trigger, encoder angle and servo current, and compute the five features inside the controller rather than uploading raw data for later analysis, precisely because the verdict has to land within one cycle.
The value of a criterion lies in the action it triggers, not in the report it produces. Practice maps the curve onto four outcomes: a normal removal proceeds to the next screw; breakaway above the ceiling stops the axis with an alarm for human confirmation, protecting the thread from stripping; abnormal run-out ripple flags that hole for thread inspection; insufficient zero-crossing angle classifies the screw as not removed and triggers a retry or a vision recheck. Four outcomes, four responses, and the operator never reads a curve, only handles the flagged exceptions.
On a station with an automatic vision screw removal machine this closes into a loop: vision supplies target position and head style, the torque curve supplies the removal verdict, and the two are cross-checked before advancing. Cross-checking solves what neither can alone. A screw may be out while a broken thread stub remains in the hole: vision sees an empty hole, the curve sees insufficient rotation, and only the comparison reveals that the position cannot be counted as complete.
The curve is not universal. It assumes synchronised sampling and time alignment, and torsional deflection between sensor and bit introduces phase lag while spring collets and floating joints store elastic energy, so peak readings sit slightly below true breakaway torque. Those offsets must be characterised during calibration. It also does not suit short-engagement self-tapping screws in thin sheet, where rotation features compress into a few degrees and the five features stop separating.
Three practical cautions. First, thresholds cannot be shared across sizes, since M0.8 and M3 peaks differ by an order of magnitude. Second, after any bit change the angle zero must be re-confirmed, because edge wear alters the initial meshing state. Third, raw samples should be retained rather than only the computed features, since process changes usually require re-reading old curves to re-derive thresholds. In an M0.8-M2.0 screw removal solution these criteria are normally delivered with the torque ceiling and the cycle counter as one parameter group, because all three share a single acquisition and logging chain.
One point deserves emphasis. The real purpose of the curve is to make tacit skill explicit. A veteran operator can hear whether a screw is galled; the curve turns that judgement into a reproducible rule. Experience does not lose value in the process. It becomes writable into a program, and therefore consistent across new variants, new operators and new lines.
Related Articles
/ RECOMMENDED NEWS
