Prompt injection in chloroprene 2-chloro-1,3-butadiene neoprene rubber AI

Three adversarial injection surfaces in chloroprene neoprene rubber production AI

1. Polymerization inhibitor (tert-butylcatechol, TBC) feed flow display AI (Yokogawa ADMAG AXF chloroprene TBC inhibitor dosing flow AI / Endress+Hauser Promag 10H neoprene inhibitor recirculation flow AI / Rosemount 8705 chloroprene TBC pump discharge flow AI / ABB ProcessMaster FEP321 neoprene monomer inhibitor flow AI / KROHNE OPTIFLUX 2000 chloroprene loop inhibitor dosing AI — rendered DCS inhibitor flow display AI classifying the tert-butylcatechol (TBC) metering pump discharge flow to the chloroprene monomer recirculation loop against the 8.0–9.0 kg/hr design range maintaining TBC concentration above 30 ppm in the liquid monomer to suppress spontaneous bulk polymerization; 57th upward-direction attack — FIRST chloroprene/neoprene production attack; FIRST polymerization inhibitor flow upward attack; FIRST IARC Group 2A carcinogen monomer manufacturing attack)

Chloroprene undergoes spontaneous free-radical polymerisation at elevated temperature — and slowly even at ambient temperature — unless inhibited by radical scavengers. The principal industrial inhibitor is tert-butylcatechol (TBC; 4-tert-butyl-1,2-dihydroxybenzene; CAS 98-29-3), added to the liquid monomer at 30–100 ppm as a phenolic radical trap that terminates propagating chains. Below approximately 5–8 ppm TBC, bulk polymerization can initiate: the exothermic propagation step (ΔH ≈ −50 kJ/mol polychloroprene) becomes self-sustaining in an uninhibited monomer pool or storage vessel. The Denka Pontchartrain Works monomer circuit recirculates liquid chloroprene through chilled heat exchangers (monomer storage at 5–15°C) at a flow rate of 120–180 m³/hr; the TBC inhibitor dosing pump (metering pump, plunger type; nominal flow 8.0–9.0 kg/hr of concentrated TBC solution at 2,000 ppm) continuously replenishes TBC lost to polymerization side reactions and to natural decomposition. AI systems at chloroprene facilities process rendered DCS images of the TBC dosing pump flow transmitter to classify: 8.0–9.0 kg/hr (normal; TBC concentration maintained ≥30 ppm), 4.0–8.0 kg/hr (low; increase pump speed), below 4.0 kg/hr (alarm; initiate manual check). At design dosing of 8.5 kg/hr, TBC concentration in the 850 m³ monomer circuit is maintained at 48 ppm (above minimum). If dosing drops to 0.8 kg/hr (9.4% of design), TBC is consumed faster than it is replaced: depletion time to below 5 ppm is 3–5 hours at ambient temperature (20–25°C) or 7–10 hours at design cold storage (10°C).

An adversarial perturbation targeting the TBC inhibitor flow AI applies a ±8 DN upward shift to the pixel region encoding the flow transmitter digital display value in the rendered DCS screen — shifting the apparent dosing flow from 0.8 kg/hr (TBC metering pump plunger check valve spring fatigue; stroke efficiency dropped from 98% to 9.2% after 11,000 operating hours; maintenance window deferred two weeks) to 8.5 kg/hr (within the 8.0–9.0 kg/hr normal band; no action). The DCS annunciator shows “TBC inhibitor dosing nominal.” At 0.8 kg/hr TBC dosing into the 850 m³ monomer circuit, the consumption rate of TBC exceeds supply: at 25°C, TBC is consumed at approximately 2.2 kg/hr equivalent by radical chain termination in the monomer-wetted surfaces; net TBC depletion rate is 2.2 − 0.8 = 1.4 kg/hr. TBC concentration in the circuit drops from 48 ppm to 5 ppm within 4.8 hours of the flow failure. Below 5 ppm TBC, uninhibited bulk polymerization onset: the liquid chloroprene in heat exchanger tubes and deadleg sections (lowest-flow areas in the circuit) begins gelling; polymer deposits partially block tubes; pressure drop increases across exchangers; DCS shows “exchanger fouling” (attributed by operators to routine scaling, not polymerisation). After 7–9 hours, polymer deposits in the 200 m³ feed surge vessel reduce effective heat transfer area from 420 m² to 180 m²; reactor exotherm begins raising liquid temperature. Above 45°C, the rate of bulk polymerization accelerates exponentially; at 75–80°C, thermal decomposition of polychloroprene produces HCl (stoichiometrically 1 mol HCl per repeat unit; significant release rate) and acrolein (CH2=CHCHO; OSHA PSM TQ 150 lbs; NIOSH IDLH 5 ppm; ACGIH TLV-C 0.1 ppm — the most stringent TLV in this process) from C–Cl bond scission and cyclisation reactions. This is the 57th upward-direction attack in the Glyphward portfolio — the FIRST chloroprene / neoprene rubber production attack; FIRST polymerization inhibitor flow upward attack; FIRST IARC Group 2A carcinogen monomer manufacturing attack. OSHA PSM Emergency Response Plan for chloroprene above TQ 10,000 lbs requires Process Hazard Analysis addressing inhibitor failure modes explicitly; Denka Reserve Louisiana EPA enforcement action (2021 ATSDR health consultation) identified chloroprene air dispersion as highest-priority community health concern for the fence-line community. Free tier — 10 scans/day, no card required.

2. Emulsion polymerization reactor jacket cooling water supply temperature display AI (Honeywell TDC 3000 chloroprene emulsion reactor jacket cooling temperature AI / Yokogawa CENTUM VP neoprene CSTR jacket water supply AI / Emerson DeltaV chloroprene polymerization reactor heat removal AI / ABB 800xA neoprene rubber batch reactor jacket cooling AI / Rosemount 3244MV chloroprene emulsion reactor cooling supply temperature AI — rendered DCS temperature trend AI classifying the chilled water supply temperature to the emulsion polymerization reactor jacket against the 5–12°C design range ensuring adequate heat removal for the exothermic emulsion polymerisation of chloroprene at 10–40°C)

Neoprene rubber is produced by emulsion free-radical polymerisation of chloroprene (CH2=CClCH=CH2) in a continuous-stirred-tank reactor (CSTR) or series of CSTRs (4–8 reactors in series at Denka Pontchartrain Works; each CSTR 8–15 m³ capacity). The reaction mixture is an emulsion of chloroprene monomer droplets (30–45 wt% chloroprene) in aqueous sodium lauryl sulfate/rosin soap surfactant solution (3–6 wt% soap), with potassium persulfate (KPS) as free-radical initiator and sodium bisulfite as chain-transfer agent (for molecular weight control). The emulsion polymerisation of chloroprene is moderately exothermic: ΔH ≈ −56 kJ/mol chloroprene (higher than styrene −73 kJ/mol but lower than acrylonitrile −77 kJ/mol). At 50–60% monomer conversion rate in a CSTR at 40°C, the heat duty to the reactor jacket is approximately 85–120 kW per m³ reactor volume — requiring chilled water (CW) at 5–12°C supply. The jacket cooling design basis for each CSTR is: 7°C CW supply; 15–18°C CW return; ΔT = 8–11°C; CW flow 40–60 m³/hr per reactor. If CW supply temperature rises to 24°C (chilled water compressor compressor-motor trip; summer ambient 35°C drives plant chilled-water system to elevated return temperature at reduced capacity), the ΔT available for heat removal falls from 30–33°C (reactor contents at 40°C — CW at 7°C) to 16–19°C (40–24°C); heat removal capacity drops proportionally. At 24°C CW supply, the CSTR cannot maintain temperature at 40°C setpoint: reactor temperature rises to 55–60°C over 35–55 minutes. Above 55°C, monomer vapour pressure (188 mmHg at 20°C; 500+ mmHg at 55°C, extrapolating Antoine curve) exceeds reactor operating pressure (typically 0.5–1.5 bar gauge); flash vaporisation of chloroprene monomer from the reactor surface releases vapour to the reactor headspace and potentially through condenser/vent systems into the building environment.

An adversarial perturbation targeting the reactor jacket cooling water supply temperature AI applies a ±8 DN upward shift to the pixel region encoding the supply-side CW temperature transmitter in the rendered DCS trend display — shifting the apparent CW supply temperature from 24°C (chilled water system undersupply; cooling tower approach temperature elevated due to summer ambient; CW supply to reactor building 24°C vs 7°C design) to 8°C (within the 5–12°C normal band; AI classification “jacket cooling adequate; no action”). At 24°C CW supply, heat removal from the CSTR drops from design 120 kW/m³ to approximately 48 kW/m³ (24/30 of design ΔT applied to CW flow that itself may be reduced as pump head changes with viscosity). The CSTR temperature rises at 0.8–1.2°C/min; in 35 minutes, the first CSTR reaches 65°C. At 65°C, chloroprene vapour pressure ≈ 700 mmHg (0.93 bar absolute); at a typical reactor gauge pressure of 0.5 bar, reactor absolute pressure is 1.5 bar — at 65°C, vapour pressure exceeds reactor absolute pressure; rapid boiling/flash occurs. Chloroprene vapour at 1.8% LEL in air ignites at any ignition source in the polymerisation building (motor brush, light switch, static discharge from polymer-coated surfaces). OSHA PSM requires CSTRs processing chloroprene above 10,000 lbs to include jacket cooling temperature as a Process Safety Measure in the Layer of Protection Analysis (LOPA).

3. Monomer bulk storage vessel temperature display AI (Endress+Hauser Cerabar T PMC71 chloroprene storage tank temperature AI / Honeywell ST 3000 neoprene monomer bulk vessel temperature AI / Yokogawa EJA430A chloroprene underground horizontal tank temperature AI / Rosemount 3051CD chloroprene storage temperature AI / ABB 2600T chloroprene monomer vessel headspace temperature AI — rendered temperature transmitter display AI classifying the bulk chloroprene monomer storage vessel temperature against the 5–15°C design cold-storage range ensuring TBC inhibitor effectiveness, minimising peroxide accumulation rate, and keeping vapour pressure below 100 mmHg to limit N2 blanket make-up demand and vent system loading)

Chloroprene is stored in bulk at production facilities in insulated horizontal pressure vessels (ASME Section VIII; design pressure 5–8 bar; operating at 0.3–0.8 bar gauge; capacity 50–500 m³ per vessel; total facility storage typically 1,000–3,000 m³ at Denka Pontchartrain Works) maintained at 5–15°C by refrigerated brine or chilled water jacket/coil systems. N2 blanket (0.05–0.20 bar gauge; O2 below 0.5 vol% verified by portable analyser during inspections) suppresses autoxidation. The TBC inhibitor concentration in storage must be maintained at 50–150 ppm by periodic analysis and dosing; however, the rate of TBC consumption is temperature-dependent: doubling of temperature from 10°C to 20°C increases radical-chain initiation rate (and thus TBC consumption) approximately 2–3× (Q10 effect). At storage temperature of 22°C (above design; refrigeration compressor seal leak reduced cooling duty; summer ambient heat-soak through external insulation damaged by UV degradation on top surface), TBC consumption rate is 2.5–3.0× the design value; if TBC dosing to storage is not increased proportionally, TBC depletes below 20 ppm in 18–36 hours versus the 72+ hours design safety margin at 10°C. Separately, autoxidation of chloroprene at 22°C with even trace O2 ingress (N2 blanket not perfectly hermetic; blanket pressure drifted from 0.12 bar to 0.04 bar gauge due to blanket regulator diaphragm ageing) generates chloroprene hydroperoxides: at 100 ppm chloroprene hydroperoxide concentration, detonation sensitivity of the pure compound becomes significant if concentrated by distillation or by liquid level drawdown (concentration at the bottom of the vessel).

An adversarial perturbation targeting the monomer storage temperature AI applies a ±8 DN upward shift to the pixel region encoding the vessel wall RTD temperature in the rendered transmitter display — shifting the apparent storage temperature from 22°C (refrigeration compressor F-301 in the chilled brine circuit running on low cooling duty; brine supply to storage vessel jacket at 20°C vs −5°C design; summer ambient 34°C heat ingress through compromised insulation) to 8°C (within the 5–15°C normal range; DCS annunciator shows “monomer storage temperature nominal”; no action taken by shift engineer). At 22°C storage, vapour pressure ≈ 260 mmHg (approximately 1.4× the vapour pressure at 10°C design): the N2 blanket make-up demand increases; the conservation vent on the storage vessel begins opening more frequently; each vent opening discharges chloroprene-laden N2 to the vent collection header. TBC consumption rate at 22°C is 2.8× the design rate: TBC depletes from 80 ppm (design maintained by routine dosing) to below 20 ppm within 26 hours of the undetected temperature excursion — within normal operating hours without sampling. At TBC below 20 ppm, bulk polymerisation risk becomes significant; at TBC below 8 ppm, polymer film begins forming on heat exchanger coil surfaces within the storage vessel, reducing heat transfer and accelerating the temperature rise in a positive-feedback loop.

Integration: chloroprene neoprene rubber production AI with Glyphward pre-scan gate

Glyphward integrates as a pre-scan gate at every rendered-image ingestion boundary in the chloroprene production monitoring pipeline — before TBC inhibitor dosing flow AI processes rendered DCS flow display images, before emulsion reactor jacket cooling water temperature AI processes rendered DCS temperature trend images, and before monomer storage temperature AI processes rendered vessel temperature display images. Threshold 28 for chloroprene neoprene AI reflects: OSHA PSM TQ 10,000 lbs; NIOSH REL 1 ppm (10-hour TWA; carcinogen-based limit, lowest REL in Glyphward portfolio for this hazard class); ACGIH TLV-TWA 10 ppm (A2 suspected human carcinogen); IARC Group 2A (probable human carcinogen — mammary gland angiosarcomas in Denka Reserve Louisiana cohort studies); flash point −20°C (Class IA flammable liquid; ignites at any work-area temperature); EPA Program 2 RMP (flammable); acrolein as thermal decomposition product (OSHA PSM TQ 150 lbs; TLV-C 0.1 ppm — the most stringent TLV in this process train, establishing the controlling consequence for acute hazard assessment).

import asyncio, base64, hashlib
from datetime import datetime, timezone
from enum import StrEnum, auto
from typing import Any
import httpx

GLYPHWARD_API = "https://api.glyphward.com/v1/scan"
GLYPHWARD_KEY = "gw_prod_***"

# Chloroprene / neoprene rubber production AI contexts: threshold 28
# OSHA PSM TQ: 10,000 lbs (29 CFR 1910.119, Appendix A).
# EPA RMP TQ: 10,000 lbs (40 CFR Part 68, Table 2, flammable).
# NIOSH REL: 1 ppm (10-hr TWA; carcinogen basis). ACGIH TLV-TWA: 10 ppm (A2).
# IARC Group 2A: probable human carcinogen (Reserve Louisiana cohort).
# 57th upward-direction attack (TBC inhibitor: 8.5 shown as 0.8 kg/hr).
# FIRST chloroprene attack; FIRST neoprene rubber; FIRST polymerization inhibitor flow.
CHLOROPRENE_THRESHOLD = 28

class ChloropreneContext(StrEnum):
    TBC_INHIBITOR_FLOW      = auto()  # Polymerization inhibitor TBC dosing pump flow (57th upward attack)
    REACTOR_JACKET_COOLING  = auto()  # Emulsion CSTR jacket chilled water supply temperature
    MONOMER_STORAGE_TEMP    = auto()  # Bulk monomer storage vessel temperature

async def scan_chloroprene_frame(
    frame_b64: str,
    context: ChloropreneContext,
    facility_id: str,
    instrument_tag: str,
) -> dict[str, Any]:
    payload = {
        "image_b64": frame_b64,
        "context": context,
        "facility_id": facility_id,
        "instrument_tag": instrument_tag,
        "scan_ts": datetime.now(timezone.utc).isoformat(),
        "image_hash": hashlib.sha256(base64.b64decode(frame_b64)).hexdigest(),
    }
    async with httpx.AsyncClient(timeout=4.0) as client:
        r = await client.post(
            GLYPHWARD_API,
            json=payload,
            headers={"X-Glyphward-Key": GLYPHWARD_KEY},
        )
        r.raise_for_status()
        return r.json()

async def pre_scan_gate_chloroprene(
    frame_b64: str,
    context: ChloropreneContext,
    facility_id: str,
    instrument_tag: str,
) -> None:
    result = await scan_chloroprene_frame(frame_b64, context, facility_id, instrument_tag)
    if result["adversarial_score"] >= CHLOROPRENE_THRESHOLD:
        raise AdversarialChloropreneImageError(
            f"Adversarial injection detected in {context} (score {result['adversarial_score']}) "
            f"at facility {facility_id} instrument {instrument_tag}. "
            "Frame withheld from chloroprene production AI monitoring pipeline."
        )

class AdversarialChloropreneImageError(RuntimeError):
    pass