Retrofitting Vision onto a Legacy Line Without Stopping It

Adding vision to old equipment is rarely blocked by the camera or the algorithm. It is blocked by three shopfloor constraints: no access to internal PLC signals, so you cannot know whether the workpiece has arrived; no spare mounting position on the machine; and a shutdown window that may be a few hours, once or twice a year. Treating these as premises rather than obstacles splits the work naturally into two stages. Stage one runs entirely without stopping the line and only watches, never interlocking — the system starts running and starts accumulating data. Stage two takes an existing shutdown window for the fixed install and the interlock. The practical benefit of this ordering is that by the end of stage one you have real data, so thresholds and interlock logic in stage two rest on evidence instead of guesswork at the machine.

Stage one: four things you can do without stopping

• Temporary mounting: magnetic bases, C-clamps, cable ties or a free-standing post. Do not aim for appearance or long-term stability, only for a usable view. Keep it out of the operator's working envelope.
• Independent power and network: do not open the machine cabinet. Use a nearby socket with a small switch or wireless backhaul, so no existing cabling is touched.
• Software or timing triggers: without PLC signals, either sample at fixed intervals, trigger from a standalone sensor such as a photoelectric or proximity switch, or record continuously and segment on image change.
• Record only, never control: output goes to a database and a dashboard, drives no actuator, issues no stop command. Production risk is zero, and acceptance on the floor is far easier.

After stage one has run for a while, it produces valuable material: real miss and false alarm behaviour, an imaging quality assessment per station, which check items genuinely work and which need a different angle or additional light. The same data is also the strongest possible input when asking management to fund stage two — several weeks of real statistics persuade far better than a proposal document. For how to determine that an action occurred without a PLC signal, vision versus sensors compares the triggering options.

Stage two: what must be finished inside the window

A shutdown window is usually hours, so site work has to be planned to the minute, and anything left to "work out on site" will consume the window without finishing. The workable approach is to prefabricate everything that can be prefabricated: brackets cut and welded to measured dimensions, cables terminated to length, camera focus and exposure pre-set on a bench using a stand-in part. On site only three things remain — fix, wire, calibrate. Calibration in particular needs generous time, because it is the one step that cannot be prefabricated at all: the camera has to run against real workpieces to confirm field of view, focus and criteria. As a rule, allow at least two hours per station for on-site mounting and calibration, scale it by station count, and keep at least a hundred percent contingency on top.

Extra care on old equipment

The same camera and bracket may last several times longer on new equipment than on old. Vibration is usually more pronounced, and sustained vibration loosens brackets and slowly shifts camera angles, which shows up as verdicts that gradually stop being accurate. Dust and oil mist are heavier, so a film forms on the front surface quickly and overall contrast drops. Internal temperatures are higher, so focus can drift after hours of running. All three are slow variables: they do not appear immediately, which is exactly why they get missed. Choosing brackets for the harsher case — thicker plate, anti-loosening features, a replaceable protective window — and writing cleaning and angle checks into the maintenance routine is far cheaper than fixing it afterwards. A fuller list appears in shopfloor effects on vision.