Behaviour FG_05

FG_05 — the behaviour contract

Published from _docs/reference/fg-05-behaviour.md — edit it there, not here. Changes made in the wiki are overwritten on the next sync.

Capping. Fits a cap from the bunker onto the part carried on the pallet. Written from the shipped ST; the PLC is master for behaviour.

Source: Beckhoff/TwinCAT_1/PLC/PLC_1/Modules/15 FG_05/FG_05.TcPOU.

The reset and resume model is in gripper-station-reset.md §5.

---

Components

Ten symbols. Six cylinders, all FB_DoubleSolenoidFeedback.

Group	Instances	FB
Cap handling unit	Y_AxisX, Y_AxisZ1, Y_AxisZ2, Y_AxisR, Y_Gripper	FB_DoubleSolenoidFeedback
	B_Detect	FB_DigitalSensor
Cap feed	Y_CapsSourceStopper	FB_DoubleSolenoidFeedback
	Y_CapsSource (ControlBunker)	FB_AbstractBunker
	B_Part	FB_DigitalSensor
Vision	P_Camera	FB_AbstractCamera

B_Part and B_Detect answer different questions

The names are close enough to invite the mistake.

Sensor	Where	Says
B_Part	on the feed	a cap has arrived at the pick position. Step 10 waits for it; once the gripper lifts a cap away it is reporting the next one
B_Detect	inside the jaws	this unit is carrying a cap. Steps 70 and 120 confirm against it, and it decides whether Resetting may open the gripper at all

A cylinder limit cannot stand in for B_Detect: jaws that close on nothing reach .Extended sooner than jaws that close on a cap, so .Extended is true either way.

The bunker is not part of the pallet cycle

Getting this wrong stalls the station permanently.

SequenceBunker.cs turns both bits on in OnEnable and off in OnDisable — it is a continuously running feed, not a step. Driven from inside Execute it would stop between pallets, no cap would ever be waiting when the next pallet arrived, and step 10 would block on B_Part for ever.

Y_CapsSource is therefore commanded from CyclicLogic, through the DriveBunker action, on mode and state only.

Y_CapsSourceStopper is held, never indexed. An accumulating feed needs nothing more: the queue pushes the next cap against the engaged stopper as soon as the gripper takes the one in front. It is commanded to its blocking position and left there for the whole cycle.

It is commanded at Resetting step 0, before anything homes — not at the end of the reset. Its de-energised position does not block, and the feed runs throughout Resetting and Idle, so a late command would let caps spill past the pick position while the axes travel. Execute step 10 re-asserts it.

EXTENDED blocks the path. Settled 2026-08-22 by refactoring the device in the twin, and the PLC follows: both call sites are Extend() and both waits are on .Extended.

This is the opposite sense to the nine transport stoppers, where retracted holds — see the correction register in transport-behaviour.md §6. That register was derived from the TS Index Unit and Stopper Unit prefabs, and this device comes from neither, so the two facts do not contradict each other. Do not "correct" one to match the other.

The camera reads the assembly, not the cap

P_Camera is mounted on the feed side but looks across at the pallet. The key it reads is the assembly's CapsID, so the read answers "which cap does this assembly want", and the station fits a cap only to an assembly of the type it is set up for. An empty pallet reads 0.

FG_01, FG_02 and FG_05 run the identical FB_AbstractCamera under the identical name over the identical four status bits. The only difference between the three stations is which payload key their twin reads and what value it has to be — here CapsID, expected 810. The comparison is made in SIM_1's FB_Camera against a value set in this group's Mapping action, not in the control PLC; the shared mechanism and the full datum table are in fg-02-behaviour.md.

The read is first, and that placement is load-bearing. Index05 has already stopped and lifted the pallet before this group is triggered, so the assembly is in view from the first scan and there is nothing to wait for. Reading here means a NOK aborts before the gripper has taken anything — cap still in the feed, handling unit empty and home. Read after the pick instead and an abort strands a cap in the air above the part.

Production cycle

Execute, mode Production.

Step	Action	Advances on	Timeout
0	derive entry from B_Detect; camera on, trigger cleared	—	wrong park position → fault 4
10	Y_CapsSourceStopper.Extend (hold — extended blocks)	B_Part.Active and SettleTimer 500 ms
20	wait for camera	P_Camera.Ready	ReadyTimer 2 s → fault 2
30	Trigger	DeviceBusy or ResultValid	ShotTimer 3 s → fault 3
40	read verdict	ResultOK — else fault 1	ShotTimer → fault 3
50	ClearTrigger	—
60	Y_AxisZ2.Extend — down on the bunker side	.Extended
70	Y_Gripper.Extend — grip the cap	.Extended and B_Detect.Active	GripTimer 500 ms → fault 5
80	Y_AxisZ2.Retract — lift clear	.Retracted
90	Y_AxisX.Extend — traverse to the pallet	.Extended
100	Y_AxisR.Extend — rotate, in the air and clear	.Extended
110	Y_AxisZ1.Extend — lower onto the part	.Extended
120	Y_Gripper.Retract — release	.Retracted and NOT B_Detect.Active	ReleaseTimer 500 ms → fault 6
130	Y_AxisZ1.Retract — up, clear of the pallet	.Retracted
140	Y_AxisX.Retract — back over the feed	.Retracted
150	Y_AxisR.Retract — rotate home	.Retracted
160	camera off, ChangeState(Complete)	—

Where the cycle starts is derived, not remembered — step 0 reads B_Detect and the park position rather than trusting a stored flag.

Both grip confirmations are two-part: the cylinder limit and the sensor. GripTimer and ReleaseTimer are armed on the limit being reached, so the 500 ms budget covers the sensor answering, not the stroke.

Fault codes

ST_FG_05_Capping_Status, published as HMI.Capping.FaultCode. Codes 1–3 are the camera's, shared in shape with FG_01 and FG_02.

Code	Meaning
1	this station does not cap that assembly (ResultOK false)
2	the camera never became Ready
3	the shot never completed
4	started with a cap held by a unit that is not at its safe position
5	pick failed — jaws closed but B_Detect never confirmed
6	release failed — jaws opened but B_Detect never cleared

An abort means Index05 never receives OperationDone and raises its own fault 3 on the operation timeout. There is no reject station on this line.

Safety

No authored Safety section for this group. The cell chain (Machine.SafetyOk → AbortImmediate) applies.

A de-energised double solenoid does not hold a gripped cap in the twin. CyclicLogic re-asserts the grip every scan in every state but Execute. Aborting and Stopping leave the axes alone — not the grip.

Related

• gripper-station-reset.md — the reset model, safe positions, resume
• fg-03-behaviour.md — the same model with two arms
• fg-01-behaviour.md, fg-02-behaviour.md — the same camera protocol