How to Extract Instruments from a P&ID PDF.
Turn a P&ID PDF into a reviewed I/O list. Scan resolution, flattening, page orientation, dense-sheet handling, and the rows to check before export.
Every controls engineer has built one. Open a P&ID PDF, find the instrument bubbles, read the tag numbers, classify each signal type, and enter it all into a spreadsheet. The exercise is well understood. The errors are predictable.
What goes into an I/O list
An I/O list maps every field instrument to a signal type for DCS, PLC configuration. Each row captures.
- Tag number. The unique identifier, FIT-101, PSH-200, LCV-301
- Instrument type. What the ISA letter code means, Flow Indicating Transmitter, Pressure Switch High
- Signal class. AI, AO, DI, or DO
- Description. What it measures or controls
- Drawing reference. Which P&ID page it appears on
This list drives procurement, loop wiring, PLC programming, and commissioning. Errors here propagate through the entire project.
Common classification mistakes
| Instrument | Correct class | Common mistake |
|---|---|---|
| FIT, Flow Indicating Transmitter | AI | Classified as AO because "I" is misread as indicating a controller |
| FCV, Flow Control Valve | AO | Classified as DO. It's analog because it modulates position |
| PSH, Pressure Switch High | DI | Classified as AI. Switches are discrete, not analog |
| XV, Shutoff Valve | DO | Classified as AO. On, off valves are digital outputs |
| ZSO, Valve Position Open | DI | Missed entirely. Position feedback is easy to overlook |
These mistakes surface during FAT or commissioning. The most expensive time to find them.
File preparation before extraction
What you upload determines the quality of what you get back. These steps take a few minutes and reduce review time significantly.
Confirm the P&ID revision. Extract instruments from the controlled revision, not a working draft. The controlled revision number should appear in the title block. If you are unsure which revision is controlled, check the drawing index before starting.
Check resolution on scanned drawings. For scanned PDFs and image files, 300 DPI minimum produces readable instrument bubbles. At 150 DPI, small tag text becomes ambiguous. At 72 DPI, web resolution, tag text is too degraded for reliable reading. Check the scan resolution by looking at the pixel dimensions relative to the paper size. An A1 drawing at 300 DPI is approximately 9933 x 7016 pixels. If the image is significantly smaller than that, request a higher-resolution scan.
Flatten markups before uploading. Some PDF viewers save markup annotations as a separate layer. If the annotation layer contains correction information that differs from the drawing, only the base layer will be read. Flatten the PDF, print to PDF, or use Acrobat's flatten annotations function to merge all layers before uploading.
Split very large drawing sets. Drawing sets with more than 30 to 40 pages per PDF process correctly but take longer. If a drawing set is a single multi-hundred-page PDF, splitting it into groups of 20 to 30 pages by process area speeds up review because results come back in smaller, reviewable batches rather than one large output.
Orient the pages correctly. Some scanned PDFs contain pages that are rotated 90 or 180 degrees. Rotate pages to portrait or landscape as intended before uploading so the output positions match the drawing as you read it.
Scanned vs vector PDFs. What changes
Vector PDFs, exported directly from CAD software such as AutoCAD, SmartPlant, or AVEVA contain the drawing geometry as mathematical paths and text as actual character strings. The instrument tag text is machine-readable without optical character recognition. Extraction from vector PDFs is faster and produces fewer garbled tags.
Scanned PDFs are raster images, typically produced by scanning a paper drawing or a physical plotted copy. Small tag text, faded lines, skewed scans, or bleed-through from double-sided originals all produce more tags that need manual review.
For projects where both a scanned PDF and a CAD file exist, always prefer the CAD source. If only the scanned version is available, the quality of the scan is the primary variable. 400 DPI scans of clean drawings with no handwritten annotations produce review queues similar in size to vector PDF extraction. Heavily annotated or low-resolution scans produce larger review queues.
The extraction workflow
Upload
Upload the P&ID PDF or image file through the Tagsight interface. Multi-page PDFs upload as a single job. Each page is processed independently and the results are consolidated. Accepted formats are PDF, PNG, JPG, and TIFF.
Review the extracted output
After processing, the results appear in the review interface. The interface has three useful views.
Drawing mode shows each detected tag overlaid on the original drawing. This is the primary review mode. Use it to verify that the tag text matches the bubble on the drawing and to catch garbled tags, missed bubbles, or non-instrument text that should not appear in the list.
Semantic mode shows the extracted tags in a structured table without the drawing context. Use this for bulk editing. Signal class correction, description editing, and loop number assignment.
Hybrid mode shows both simultaneously. Useful when the drawing is dense enough that tag positions are ambiguous.
Manage multi-page drawing sets
When a P&ID set spans multiple drawings, PID-001 through PID-045, for example, upload the full set as a single job or as sequential jobs within the same project. The review canvas tracks which page each instrument was found on, so the P&ID reference column in the exported I/O list is populated per page.
For very large drawing sets, a useful workflow is to process the drawings in groups by process area, utility systems, unit 100, unit 200, review each group independently, and then consolidate the exported I/O lists in Excel. Consolidation in Excel requires a duplicate-check step because off-page connectors sometimes cause the same instrument to appear in adjacent drawing sections.
What to do with off-page connectors
Off-page connectors are symbols that indicate a process connection continues on another drawing. An instrument that feeds a signal to a controller on a different page appears as a connection reference on the source page. These references are not instrument tags. They should not appear as rows in the I/O list.
During review, check any tag row that appears to originate from an off-page connector reference rather than an instrument bubble, and delete or correct it as needed.
Export the reviewed I/O list
After review, export the I/O list to the format your team uses. The P&ID digitization guide covers what happens between initial upload and a reviewed I/O list, including how to handle scanned vs vector sources and how to set up drawing sets for best results.
- Excel (.xlsx). The standard format for procurement, engineering, and commissioning
- TIA Portal XML. Direct import into Siemens S7-1200, 1500 tag databases
- Rockwell L5X. Import into Studio 5000 for Allen-Bradley systems
- PLCCreator CSV. I/O count estimation by PLC architecture
- DEXPI XML. ISO 15926-based format for process industry data exchange
When this matters most
- Brownfield projects. Legacy P&IDs exist only as scanned PDFs, no CAD source
- Control system migrations. DCS-to-DCS or DCS-to-PLC, need fresh I/O counts
- Bid estimation. Count instruments to size a control system proposal
- Documentation audits. Verify that as-built I/O lists match current P&IDs
After extraction, validate the resulting I/O list against the P&IDs using the structured checklist in how to validate an I/O list against P&IDs.
Troubleshooting common issues
Missing bubbles on dense drawings. Some drawings pack instrument bubbles tightly, especially in compressor station or metering skid drawings. If the instrument count from the list is significantly lower than expected, check the original drawing at full resolution and look for bubbles that are very small or that overlap each other. Splitting the drawing into cropped sections and uploading each section as a separate image is worth trying on extremely dense areas.
Garbled tag text on scanned drawings. If tags in the output contain substituted characters, FT-1O1 instead of FT-101, where O is the letter O instead of the digit 0, the scan resolution is likely insufficient for the text size on that drawing. Request a higher-resolution scan, or correct the misread characters manually during review.
Instruments appearing twice. If the same instrument appears on two consecutive pages because it straddles a sheet boundary, one of the two rows should be deleted. Keep the row from the page where the instrument bubble, the circle with the tag inside appears, not the page where only the line continuation appears.
Tags from drawing notes or legends. Drawing legend boxes often contain example tag formats to illustrate the tagging convention. These are not instruments. During review, delete any row whose source bubble is clearly in a legend or notes block rather than in the process flow. Legend boxes are usually in a corner or along the border of the drawing.
Working example on a real drawing
Upload a P&ID and Tagsight returns a classified I/O list ready to review and export. The best test is your own drawings, because every facility has its own conventions.
For format-specific export guidance, see the I/O list template reference and the instrument index vs I/O list vs line list comparison.
FAQ
What resolution should I scan P&ID drawings.
300 DPI is the practical minimum for reliable tag text. 400 DPI is preferred for drawings with small text or fine linework. Anything below 200 DPI produces ambiguous tag numbers that require significant manual correction. For A0 or A1 drawings with many instruments per page, scan at 400 DPI. For A3 or A4 drawings with larger text, 300 DPI is sufficient. Always scan in grayscale or color, not pure black-and-white, 1-bit, which loses anti-aliasing and makes curved text harder to read.
How do I handle instruments with non-ISA tag formats.
Tag formats that do not follow ISA 5.1, such as Dutch HVAC tags like AB2WV05, German DIN formats, or KKS plant coding like 10LAC01CT001 appear in the output as tag numbers, but signal class cannot be read from the ISA second-letter rule because the letter positions mean different things. Assign the signal class manually based on the instrument type and the facility's tagging convention. If the facility has a tagging legend or standard document, refer to it for the function letter mapping.
Can I upload a drawing that has both P&ID and non-P&ID content on the same page.
Yes. If a page contains a P&ID section and a legend section, P&ID instruments appear in the results and legend text is typically filtered out during review. If a page is a mixed document type, for example, a combined P&ID and cause-and-effect matrix on the same sheet, instrument tags from the P&ID portion are included. Review the output to confirm that content from the non-P&ID section is not included.
How do I deal with revision clouds on P&IDs.
Revision clouds, the curved lines that highlight changed areas on a drawing do not affect the instrument list. All instrument bubbles on the page are included regardless of whether they are inside or outside a revision cloud. If a revision cloud indicates a deleted instrument, shown with an X through the bubble, delete that row from the list during review. Deletions in revision clouds require manual review of the affected area.
What is the best way to handle instruments that appear on a vendor package drawing rather than the main P&IDs.
Vendor package drawings, skid P&IDs supplied by the equipment vendor should be processed separately from the main facility P&IDs. The tag numbering convention may differ, and the SIS, BPCS classification for vendor package instruments must be confirmed against the package instrument index supplied by the vendor. After processing, the vendor package I/O list is merged into the main project I/O list with a clear source flag, for example, a "Package" column noting the vendor package reference. Duplicate tag numbers between the vendor package and the main facility list must be resolved before merging.