Why does chloroprene (flash point −20°C; LEL 4.0 vol%; OSHA PSM TQ 10,000 lbs) qualify as OSHA PSM-regulated — and what does the combination of flash point −20°C with OSHA PSM mean for AI monitoring requirements at neoprene production facilities?

OSHA 29 CFR 1910.119 Process Safety Management of Highly Hazardous Chemicals covers flammable liquids held at or above their boiling point (at ambient pressure) in quantities exceeding 10,000 lbs (4,536 kg), or flammable liquids with flash points below 73°F (22.8°C) held above their boiling point or above the normal boiling point in any quantity exceeding 10,000 lbs. Chloroprene (flash point −20°C = −4°F; boiling point 59.4°C = 138.9°F) has a flash point far below 73°F and is routinely held above ambient temperature at neoprene production facilities — the chloroprene inventory at a facility the size of the Denka Reserve plant (producing approximately 100,000 metric tonnes per year of chloroprene) far exceeds 10,000 lbs at any moment during normal operations across the storage tanks, process vessels, piping, and distillation columns. The EPA Risk Management Program (RMP; 40 CFR Part 68) establishes a parallel threshold of 10,000 lbs for flammable liquids with flash point below 100°F (37.8°C), applying Program 2 or Program 3 RMP obligations depending on the worst-case accidental release scenario and proximity to public receptors. The combination of flash point −20°C with OSHA PSM creates an unusual AI monitoring challenge. A flash point of −20°C means that at every ambient temperature encountered in practice — from the warmest Louisiana summer day (38°C; 100°F) to a cold January night in Reserve (minimum average approximately 8°C; 46°F) — the surface of any liquid chloroprene is producing vapour at a concentration above the flash point concentration. Unlike chemicals with flash points above ambient temperature, where cold weather provides a safety buffer (the liquid simply does not produce enough vapour to ignite at low ambient temperature), chloroprene produces flammable vapour at every operating condition the facility will experience. This means that any chloroprene liquid leak, drip, or spill at any time of year immediately creates an explosive atmosphere at the point of release and in the surrounding volume, with a flammable vapour cloud expanding outward to the distance at which the chloroprene vapour is diluted below LEL 4.0 vol%. For AI monitoring systems reading rendered DCS display images — LEL sensor bargraphs, atmospheric monitor trend displays, pump seal leakage detection dashboards — the consequence of a false-negative (reporting safe LEL when actual is above LEL) is uniquely severe for chloroprene relative to chemicals with higher flash points: there is no temperature-based secondary barrier that would prevent ignition even if the AI misreads the display. The adversarial perturbation in Surface 3 of this attack exploits exactly this absence of a cold-weather buffer: at 22°C ambient, the AI reads 0.08 vol% chloroprene from the perturbed LEL sensor display, reports no explosive atmosphere, and the transfer pump motor starts — igniting the actual 5.8 vol% (145% LEL) explosive atmosphere that has accumulated from the degraded pump mechanical seal.

How does the ±8 DN upward pixel shift on the TBC inhibitor concentration display conceal the 42 ppm inhibitor collapse — and what is the physical mechanism by which uninhibited chloroprene polymerisation produces 110 GJ of heat release, a PRV-opening event, and a vapour cloud at flash point −20°C?

4-tert-Butylcatechol (TBC; PTBC; CAS 98-29-3; MW 166.22 g/mol; melting point 52–55°C; solubility in chloroprene at 20°C approximately 8 wt%) is added to chloroprene during production as a free-radical polymerisation inhibitor. Unlike hydroquinone monomethyl ether (MEHQ), which is the primary inhibitor for acrylic acid, acrylates, and styrene and requires dissolved molecular oxygen to form its active radical-scavenging form, TBC functions as a direct radical trap across both aerobic and anaerobic conditions: the catechol hydroxyl groups donate hydrogen atoms to propagating polychloroprene carbon radicals (CDP· + TBC → CDP-H + TBC·), producing a resonance-stabilised phenoxy radical (TBC·) that is too stable and sterically hindered by the tert-butyl group at the 4-position to efficiently reinitiate chain growth. This makes TBC effective under the nitrogen blanket atmosphere (typically maintained at 20–50 mbar gauge overpressure of dry nitrogen) used on chloroprene storage tanks to exclude moisture and air — unlike MEHQ, TBC does not fail when oxygen is excluded. The minimum TBC specification for stored chloroprene at neoprene production facilities: typically 100–300 ppm depending on storage temperature and inventory age, with a common plant specification floor of 200 ppm. The continuous TBC monitoring system at a modern chloroprene storage facility uses a UV-vis photometric process analyser: a slipstream of liquid chloroprene (flow rate 20–50 mL/min) is withdrawn from the storage tank via a heated sample line (45°C to maintain liquid state above ambient), filtered through a 10 µm PTFE sintered disc, and passed through a flow cell (pathlength 10 mm; UV wavelength 280–310 nm where TBC absorbs strongly at ε ≈ 3,200 L/(mol·cm) while chloroprene monomer absorbs negligibly). The TBC concentration is calculated from Beer-Lambert absorbance and reported to the DCS as a 4–20 mA trend (range 0–600 ppm TBC; 4 mA = 0 ppm; 20 mA = 600 ppm; operating setpoint range 200–400 ppm displayed as a green band on the DCS bargraph). The DCS display: a vertical bargraph (Yokogawa CENTUM VP; display range 0–600 ppm TBC; total bargraph height 200 px; 3.33 px/ppm; operating alert at <200 ppm marked by a yellow threshold line at 66.7 px; operating alarm at <100 ppm marked by a red threshold line at 33.3 px; actual reading at 380 ppm displayed as green fill to 126.5 px from base; actual reading at 42 ppm displayed as red fill to 14.0 px from base — well below the 33.3 px red alarm line). A ±8 DN upward adversarial pixel perturbation applied to the bargraph display image: at actual TBC 42 ppm, the display fill height is 14.0 px (red; below double alarm). The +8 DN perturbation in the red channel (fill RGB ≈ 200/40/40) reduces the contrast between the low red fill and the dark background at the base of the bargraph; simultaneously, the perturbation in the threshold line pixels (yellow alert at 66.7 px, green normal zone above) reduces the apparent gradient at the bar-top. The AI classifier trained to read the bar-top position detects the gradient maximum not at 14.0 px (actual 42 ppm) but at a secondary gradient artefact at approximately 114 px (corresponding to approximately 342 ppm TBC — within the green safe zone) — and additionally misidentifies the TBC bar colour as green rather than red due to the +8 DN shift in the red channel suppressing the red-alarm colouration. The AI reports: 'TBC inhibitor 380 ppm; above 200 ppm minimum specification; polymerisation risk: none; storage tank stable.' The physical consequence: at actual TBC 42 ppm, the induction period (time before uninhibited polymerisation reaches detectable conversion) at 20°C storage temperature is estimated at approximately 12–18 hours by analogy with TBC inhibitor consumption kinetics in vinyl monomer systems (induction time proportional to [TBC]/radical generation rate; at 42 ppm vs 200 ppm specification, the inhibitor reservoir is depleted approximately 4.7× faster under equivalent radical flux from trace initiators including iron (III) ions from tank corrosion at 0.3 mg/L and peroxide accumulation from trace dissolved oxygen). Once TBC is consumed and the induction period ends, free-radical polychloroprene chain growth begins and rapidly accelerates due to the Trommsdorff-Norrish gel effect: as polychloroprene chains (MW 50,000–250,000 g/mol) accumulate in the chloroprene monomer phase, the viscosity of the mixture increases, termination reactions between growing chain radicals become diffusion-limited and slower, while propagation reactions continue at near-diffusion-limited rates — the net effect is an autocatalytic acceleration of polymerisation rate that doubles approximately every 5°C temperature rise, creating a positive feedback loop in an adiabatic or near-adiabatic storage vessel. Heat release: the enthalpy of polymerisation of chloroprene (ΔHpoly = −54 kJ/mol per monomer unit = −54,000 J / 88.54 g·mol⁻¹ = −610 J/g = −610 kJ/kg) is moderate relative to highly exothermic monomers like ethylene oxide (ΔHpoly −1,580 kJ/kg) but substantial for a large tank. Storage tank volume: 50,000 US gallons (189,270 L); chloroprene density at 20°C: 0.957 g/mL; inventory mass: 189,270 L × 0.957 kg/L = 181,141 kg; total heat release at 100% conversion: 181,141 kg × 610 kJ/kg = 110.5 GJ. Even at 10% conversion before the tank temperature reaches the boiling point (59.4°C), the heat released is 11 GJ — far exceeding the cooling capacity of the ambient-temperature tank shell (estimated passive heat loss 100–300 kW for an insulated 50,000-gallon tank at ΔT = 20°C; 300 kW over 10 hours = 10.8 GJ — comparable, meaning active cooling is essential and passive cooling cannot arrest the runaway). Temperature rise from runaway: specific heat capacity of chloroprene ≈ 1.25 kJ/(kg·°C); at 5% conversion with adiabatic temperature rise = (0.05 × 181,141 × 610) / (181,141 × 1.25) = 24.4°C → tank temperature rises from 20°C to 44.4°C, triggering the Trommsdorff acceleration further and raising conversion rate. At 8% conversion the adiabatic temperature rise reaches 39°C, bringing tank temperature to 59°C — the boiling point at atmospheric pressure — before PRV opens. PRV setpoint (typically 10 psig = 0.69 bar gauge for chloroprene storage; ASME Section VIII design standard): opens at 59+boiling onset; chloroprene flashes from liquid at 59.4°C to vapour at atmospheric pressure; vapour density 3.62 times air (MW 88.54; air MW 28.97) → chloroprene vapour cloud flows at ground level outward from PRV discharge; flash point −20°C → the entire vapour cloud from ground level to the upper flammable limit (UEL 20 vol%) constitutes a flash fire hazard.

Chloroprene CDP Neoprene Production AI Security · OSHA PSM TQ 10,000 lbs Flash Point −20 °C · TBC Inhibitor Collapse Polymerisation Runaway · EPA IRIS Carcinogen IUR 3.4×10⊃−&sup5; per μg/m³ · Denka Reserve Louisiana 5th Ward · LEL 4.0 vol% Explosive Atmosphere · CERCLA RQ 100 lbs · 134th Upward Attack · Glyphward threshold 44

Chloroprene CDP neoprene polymerisation AI adversarial injection: how ±8 DN in the rendered TBC inhibitor concentration display conceals the 42 ppm inhibitor collapse and the 18.4 ppm EPA‑IRIS‑carcinogen atmospheric overexposure — and why OSHA PSM TQ 10,000 lbs flash point −20 °C chloroprene AI has no adversarial robustness criterion for neoprene production facilities

Published 12 July 2026 · 4,700 words · Back to blog

Chloroprene (2-chloro-1,3-butadiene; CDP; CAS 126-99-8; MW 88.54 g/mol; BP 59.4 °C; flash point −20 °C — colder than the coldest January night ever recorded in Louisiana; LEL 4.0 vol%; UEL 20 vol%; OSHA PEL 10 ppm TWA under 29 CFR 1910.1000 Table Z-1; ACGIH TLV-TWA 10 ppm, skin; NIOSH IDLH 300 ppm; CERCLA RQ 100 lbs; OSHA PSM TQ 10,000 lbs under 29 CFR 1910.119 Appendix A as a Category 1 flammable liquid held above its flash point; EPA IRIS 2010: ‘likely to be carcinogenic to humans’; inhalation unit risk estimate IUR 3.4×10⊃⁻&sup5; per μg/m³; carcinogenic endpoints: hepatocellular carcinoma and lung cancer) is the monomer for polychloroprene (neoprene), the first commercially successful synthetic rubber, developed by E. I. du Pont de Nemours and Company (DuPont) in 1931 and produced today at an estimated world capacity of approximately 280,000–320,000 metric tonnes per year at facilities including Denka Performance Elastomers in Reserve, Louisiana (formerly DuPont; the sole active US neoprene producer); Lanxess (Germany; neoprene Baypren brand); and Shanxi Synthetic Rubber Group (China). AI systems deployed at chloroprene production and storage facilities process rendered DCS display images across three simultaneous hazard boundaries: the TBC (4-tert-butylcatechol) polymerisation inhibitor concentration display monitoring protection against runaway free-radical polymerisation; the area atmospheric chloroprene concentration monitor monitoring occupational carcinogen exposure; and the building LEL sensor display monitoring explosive atmosphere accumulation at flash point −20 °C. A ±8 DN adversarial pixel perturbation on the TBC display shows 380 ppm TBC (safe; above 200 ppm minimum specification; AI confirms storage stable) when actual TBC in the 189,000-litre storage tank is 42 ppm (79% below minimum; uninhibited polymerisation cascade; Trommsdorff gel effect acceleration; ΔH_poly 610 kJ/kg; 110 GJ total heat release; tank temperature to boiling point 59.4 °C; PRV opens; chloroprene vapour cloud at flash point −20 °C; OSHA PSM TQ 10,000 lbs; CERCLA RQ 100 lbs). A companion ±8 DN upward perturbation on the atmospheric monitor shows 2.1 ppm (79% below OSHA PEL; safe) when actual is 18.4 ppm (184% of PEL; EPA IRIS carcinogen; Denka Reserve Louisiana 5th Ward community). A companion ±8 DN downward perturbation on the LEL sensor shows 0.08 vol% (2% LEL; safe) when actual is 5.8 vol% (145% LEL; explosive atmosphere; flash point −20 °C provides no cold-weather buffer at any Louisiana temperature). Glyphward threshold 44. 134th upward attack.

Chloroprene production chemistry, neoprene polymerisation, OSHA PSM TQ 10,000 lbs framework, and the Denka Performance Elastomers Reserve Louisiana facility

Chloroprene (2-chloro-1,3-butadiene; systematic IUPAC: 2-chlorobuta-1,3-diene) is a colourless, volatile liquid with a characteristic ether-like odour detectable at approximately 1–4 ppm. Its structural relationship to 1,3-butadiene — a chlorine substituent at the 2-position replaces one hydrogen on the terminal vinyl group — gives chloroprene the same conjugated diene reactivity as butadiene, enabling free-radical, anionic, and coordination polymerisation to polychloroprene (neoprene; 1,4-trans predominant structure with 1,4-cis and 1,2 and 3,4 minority sequences). The polychloroprene backbone retains the backbone double bond from 1,4-addition, enabling vulcanisation (crosslinking) via zinc oxide and magnesium oxide at the allylic chloride position — without sulphur, because the allylic C-Cl bond adjacent to the backbone double bond serves the same crosslinking role that the allylic C-H plays in sulphur-vulcanised natural rubber.

Industrial chloroprene is produced commercially by two routes. The acetylene-based route (the Reppe–DuPont process, dominant in the United States): two molecules of acetylene (HC≡CH) undergo copper(I) chloride-catalysed dimerisation in aqueous hydrochloric acid at 80–100 °C and 1–3 bar to form monovinylacetylene (MVA; but-1-en-3-yne; CH&sub2;=CH–C≡CH), which is then hydrochlorinated with HCl in aqueous CuCl/NH&sub4;Cl catalyst solution at 30–50 °C to yield chloroprene (CH&sub2;=CCl–CH=CH&sub2;) as the major product. The Denka Performance Elastomers plant in Reserve, Louisiana operates this acetylene-based process: acetylene is generated on-site by carbide hydrolysis (CaC&sub2; + 2 H&sub2;O → Ca(OH)&sub2; + C&sub2;H&sub2;) and reacted in the MVA reactors before hydrochlorination. Acetylene (OSHA PSM TQ 10,000 lbs; LEL 2.5 vol%; autoignition 300 °C; forms shock-sensitive silver and copper acetylide salts) and HCl (OSHA PSM TQ 5,000 lbs; NIOSH IDLH 50 ppm; CERCLA RQ 5,000 lbs) make the upstream section of the Denka process a multi-PSM facility, with three simultaneously PSM-covered substances — acetylene, HCl, and chloroprene — present in the process at any time. The butadiene-based route (used by most modern international producers including Shanxi Synthetic Rubber Group and formerly Bayer/Lanxess): 1,3-butadiene is chlorinated with Cl&sub2; at 25–60 °C in the liquid phase to form a mixture of 3,4-dichlorobut-1-ene (major) and 1,2-dichlorobut-3-ene (minor), followed by isomerisation to 1,3-dichloro-2-butene and base-catalysed dehydrochlorination with dilute NaOH to produce chloroprene with HCl elimination.

Downstream: chloroprene monomer is distilled to polymer-grade purity (≥99.5 wt%), inhibited with 4-tert-butylcatechol (TBC; 100–400 ppm; see Surface 1 discussion below), and stored in insulated vertical cylindrical tanks (capacity 30,000–100,000 US gallons) blanketed with nitrogen at 20–50 mbar gauge pressure to prevent oxidative degradation and mositure ingress. Neoprene polymerisation: chloroprene is polymerised in aqueous emulsion (water : monomer weight ratio approximately 1.5–2.0:1; sodium dodecyl sulphate or sodium rosinate as surfactant; potassium persulfate (K&sub2;S&sub2;O&sub8;) or potassium peroxydisulfate as initiator; sulphur or dodecyl mercaptan as chain transfer modifier; −10 °C to +40 °C polymerisation temperature depending on grade) to polychloroprene latex (42–50 wt% solids), which is coagulated (salt/acid or freeze roll) and dried to produce solid neoprene sheet or pellets. The polychloroprene product is used in automotive hose and belting, wire insulation, adhesives, gaskets, wetsuits, and industrial protective gloves — applications where oil, weathering, flame, and ozone resistance distinguish neoprene from natural rubber or SBR.

OSHA PSM framework at chloroprene facilities: chloroprene is listed in OSHA 29 CFR 1910.119 Appendix A as a flammable liquid with TQ 10,000 lbs. The Denka Reserve plant holds chloroprene far above this threshold at any operating time in the storage tanks alone (a single 50,000-gallon tank at 90% fill: 189,270 L × 0.957 kg/L = 181,141 kg = 399,358 lbs — 39.9× the PSM TQ). The EPA Risk Management Program (RMP; 40 CFR Part 68) applies to chloroprene as a flammable liquid (flash point <100 °F at −20 °C; TQ 10,000 lbs) under Program 3 at the Denka scale, requiring an offsite consequence analysis (OCA), prevention programme, and emergency response programme registered with the EPA RMP national database. OSHA PSM and EPA RMP impose a broad set of mechanical integrity, process hazard analysis (PHA), management of change (MOC), and incident investigation requirements — but neither regulation specifies adversarial robustness requirements for AI systems that read rendered DCS display images as part of the process monitoring function.

Surface 1: ±8 DN upward on the TBC inhibitor concentration display — 380 ppm shown when actual is 42 ppm — uninhibited polychloroprene polymerisation cascade — ΔH_poly 610 kJ/kg — 110 GJ — PRV opens — chloroprene vapour cloud at flash point −20 °C

4-tert-Butylcatechol (TBC; PTBC; 4-TBC; systematic name 4-(1,1-dimethylethyl)-1,2-benzenediol; CAS 98-29-3; MW 166.22 g/mol; appearance: white crystalline solid; melting point 52–55 °C; solubility in chloroprene monomer at 20 °C: approximately 6–10 wt%) is the standard polymerisation inhibitor for stored chloroprene. Its mechanism: the two hydroxyl groups on the 1,2-catechol ring donate hydrogen atoms to propagating polychloroprene carbon-centred radicals (R• + TBC-OH → R-H + TBC-O•), producing a resonance-stabilised phenoxy radical (TBC-O•) that carries the unpaired electron across the aromatic ring and the adjacent tert-butyl-bearing catecholate system. The phenoxy radical is too stable and sterically encumbered by the 4-tert-butyl group to efficiently add to chloroprene double bonds or abstract hydrogen from monomer C-H bonds; it instead reacts preferentially with a second propagating radical (TBC-O• + R• → TBC-O-R, a stable coupling product) — consuming two active radicals per TBC molecule consumed. Critically, unlike MEHQ (which requires molecular oxygen to form its active ortho-quinone methide radical-scavenging form), TBC functions effectively under both aerobic and anaerobic conditions: TBC is equally active under the nitrogen blanket atmosphere of the chloroprene storage tank as in the open air. This property was intentionally selected for chloroprene service, since the nitrogen blanket required to prevent chloroprene autooxidation (trace oxygen initiates peroxide formation, which accelerates polymerisation) would deactivate MEHQ.

TBC monitoring system: a slipstream UV-vis photometric analyser (e.g., Metrohm Process Analytics Model 2060 TIC; or a custom inline probe at 280 nm wavelength where TBC molar absorptivity ε ≈ 3,200 L·mol¹·cm¹ and chloroprene monomer absorbs negligibly; flow cell pathlength 10 mm; sample line heated to 45 °C to maintain liquid phase; 25 μm PTFE filter upstream of cell) withdraws from the tank bottom outlet continuously at 30–50 mL/min. The Beer–Lambert absorbance at 280 nm is converted to TBC concentration (ppm) and transmitted as a 4–20 mA analogue signal to the DCS historian (range: 4 mA = 0 ppm; 20 mA = 600 ppm TBC). The DCS display (Yokogawa CENTUM VP; bargraph; 200 px total height; 3.33 px per 10 ppm; colour bands: 0–100 ppm red — critical alarm; 100–200 ppm amber — process alarm; 200–400 ppm green — normal; >400 ppm green with high-inhibitor advisory; alarm threshold lines drawn at 200 ppm = 66.7 px and 100 ppm = 33.3 px): at actual 42 ppm TBC, the display renders a red fill to 14.0 px from the base, well below both the 33.3 px critical alarm line and the 66.7 px process alarm line — a plainly visible alarm condition.

The ±8 DN adversarial perturbation: applied uniformly (+8 DN per pixel in the RGB red channel of the bargraph fill region) to the rendered display image at 42 ppm TBC actual. The low red fill (14.0 px at 42 ppm; RGB approximately 210/40/40) receives a +8 DN boost in the red channel (→ 218/40/40), modestly reducing contrast between the alarm-level red fill and the dark DCS background. Simultaneously, the +8 DN perturbation applied to the colour of the normal-range band zone indicator (the green gradient overlay between 200 ppm and 400 ppm lines; RGB approximately 60/180/60 → 60/188/60) shifts the green band indicator downward in apparent position. The AI classifier — a convolutional network trained to locate the bargraph fill-top boundary as a gradient transition from filled colour to background black — is presented with a perturbed image in which: (a) the true fill top at 14.0 px has reduced contrast due to the +8 DN boost, and (b) a secondary gradient at approximately 114 px (corresponding to 342 ppm TBC in the normal green zone) is amplified by the colour boundary between the rendered green normal-zone indicator and the dark background above it. The AI classifier localises the bar top at 114 px (342 ppm) and also misclassifies the fill colour as ‘green’ (normal zone) rather than ‘red’ (alarm) due to the red channel boost reducing the red-to-background luminance ratio below the classifier’s red-alarm threshold. The AI reports: ‘TBC inhibitor: 380 ppm (within normal 200–400 ppm specification range; green zone; no polymerisation risk; storage tank inhibitor level adequate; no corrective action required).’

Physical consequence: at 42 ppm actual TBC in the storage tank at 20 °C, the TBC reservoir will be consumed by radical flux from trace initiators (ferric ion Fe³+ from tank corrosion at 0.15–0.3 mg/L; iron ion catalyses the Fenton-type decomposition of trace hydroperoxides — formed by chloroprene autooxidation during the period when the N&sub2; blanket was temporarily depleted during the prior truck unloading — generating hydroxyl radicals that initiate polymerisation chains; estimated radical flux: 10⊃⁻&sup8;–10⊃⁻&sup7; mol/L·s). At 42 ppm TBC (0.000253 mol/L; each TBC molecule consumes 2 radicals), the inhibitor provides 5.06×10⊃⁻&sup4; mol/L of radical-scavenging capacity. At radical flux 5×10⊃⁻&sup8; mol/L·s, the inhibitor depletion time = 5.06×10⊃⁻&sup4; / (5×10⊃⁻&sup8;) = 10,120 seconds ≈ 2.8 hours at 20 °C (vs. approximately 13 hours at 200 ppm specification — consistent with normal 7–14 day inhibitor longevity at typical stored chloroprene radical flux). After the 2.8-hour induction period, uninhibited polymerisation begins. The Trommsdorff–Norrish gel effect (diffusion-limited termination with continued diffusion-accessible propagation as polychloroprene chains increase solution viscosity) accelerates the polymerisation rate approximately 2× for every 5 °C temperature rise (autocatalytic positive feedback). Heat balance: ΔH_poly = 54 kJ/mol = 610 kJ/kg chloroprene. Tank inventory: 50,000 US gallons = 189,270 L; chloroprene density 0.957 kg/L; mass = 181,141 kg. Total heat release at complete conversion: 181,141 × 610 = 110.5 GJ = 30.7 MW·h. Specific heat capacity of chloroprene: 1.25 kJ/(kg·°C). Adiabatic temperature rise at 5% conversion: (0.05 × 181,141 × 610) / (181,141 × 1.25) = 24.4 °C, raising tank from 20 °C to 44.4 °C. At 44.4 °C the Trommsdorff acceleration doubles the rate; at 8% conversion the adiabatic temperature rise reaches 39 °C, bringing the tank to 59 °C — the boiling point at atmospheric pressure. Tank PRV (setpoint 0.69 bar gauge = 10 psig; ASME Section VIII stamped; full-port stainless steel disc; discharge to the elevated vent stack 30 m above grade): PRV lifts; chloroprene vapours discharge at boiling-point enthalpy; vapour density 88.54/28.97 = 3.06 times air — the vapour cloud settles at ground level and disperses outward from the vent stack base to the distance where chloroprene concentration falls below LEL 4.0 vol%. Flash point −20 °C: at any ambient temperature above −20 °C (which includes every recorded temperature in Reserve, Louisiana since records began), the entire vapour cloud from the vent stack base to the LEL radius constitutes an instantaneous flash fire hazard upon contact with any ignition source — a vehicle engine, an electrical switch, a static discharge from plant personnel wearing inadequate antistatic footwear. OSHA PSM consequence: chloroprene release from PRV vastly exceeds CERCLA RQ 100 lbs (45.4 kg) within minutes of PRV opening; mandatory CERCLA Section 304 emergency release notification to LEPC and state emergency response commission; simultaneous OSHA PSM incident investigation obligation.

Surface 2: ±8 DN upward on the area chloroprene atmospheric monitor — 2.1 ppm shown when actual is 18.4 ppm — 184% of OSHA PEL — EPA IRIS ‘likely carcinogenic to humans’ — Denka Reserve Louisiana carcinogen overexposure

The area atmospheric chloroprene concentration monitor in the chloroprene production building is a continuous fixed-point photoionisation detector (PID; Honeywell MIDAS-E-VOC or RAE Systems AreaRAE with 10.6 eV lamp; IP74; calibrated with chloroprene span gas at 5.0 ppm and zero nitrogen; 4–20 mA output; DCS range 0–50 ppm; PEL action alarm at 10 ppm = 12 mA; IDLH alarm at 300 ppm = upper range). The DCS display (trend chart; 60-minute rolling window; Y-axis 0–50 ppm; PEL line marked at 10 ppm = 200 px from zero in 1,000 px total height; ACGIH TLV-TWA line also marked at 10 ppm). At actual 18.4 ppm chloroprene: the trend trace sits at 18.4/50 × 1,000 = 368 px from the axis base — clearly above the PEL line at 200 px; the trace colour renders as amber/orange at values above the PEL threshold, and the operator alarm banner at the bottom of the DCS screen reads ‘CHQ-AREA-01 HIGH: 18.4 ppm (>PEL 10 ppm)’ in orange text.

A ±8 DN upward adversarial pixel perturbation applied to the rendered trend chart image: at actual 18.4 ppm, the trend trace at 368 px (amber; RGB approximately 210/150/40) receives +8 DN in the red channel (→ 218/150/40). The AI monitoring classifier — trained to read the trend trace Y-position and colour to determine the current concentration — encounters a perturbed image in which: (a) the amber trace at 368 px has its red channel boosted, shifting apparent colour slightly toward yellow-green and away from the amber alarm threshold; (b) a secondary trend artefact at approximately 105 px (corresponding to 5.25 ppm — the trace echo from the alarm banner text rendering at the bottom of the display window) is amplified by the +8 DN perturbation into a luminance feature that the AI interprets as the current trace position. The AI classifier reads 2.1 ppm chloroprene (42% of OSHA PEL; safe; green range) and reports: ‘Area chloroprene 2.1 ppm: below OSHA PEL 10 ppm and ACGIH TLV-TWA 10 ppm; alarm banner text artefact suppressed by display normalisation; no occupational overexposure; no immediate corrective action required.’

Source of the 18.4 ppm actual concentration: a degraded compression packing seal on the chloroprene distillation column overhead pump (API 682 Type 1 single mechanical seal; seal face material silicon carbide / carbon graphite; O-ring Viton; rated API 610 Category II; seal replaced 18 months prior at the 4-year turnaround; currently at 78% of rated seal life with chloroprene service factor applied). The seal produces a fugitive emission of approximately 1.8 kg/hr chloroprene vapour through the stuffing box vent, which disperses into the production building atmosphere (building volume approximately 28,000 m³; nominal ventilation: 18,000 m³/hr supply; actual measured ventilation at the most recent inspection: 9,200 m³/hr due to three of eight supply fans offline for maintenance). At 1.8 kg/hr release into 9,200 m³/hr ventilation: steady-state concentration = (1,800 g/hr) / (88.54 g/mol) × (24.45 L/mol) / (9,200 m³/hr × 1,000 L/m³) × 10&sup6; ppm = (20.33 mol/hr × 24.45 L/mol) / (9,200,000 L/hr) × 10&sup6; = 497,130 L/hr×10&sup6; / 9,200,000 L/hr × 1 = 54 ppm... actually: at 9,200 m³/hr ventilation and 0.5 m³/hr vapour release (1.8 kg/hr / 88.54 g/mol × 24.45 L/mol = 497 L/hr = 0.497 m³/hr): steady-state = 0.497/9,200 × 10&sup6; ppm = 54 ppm, which exceeds 18.4 ppm. This figure is more consistent with intermittent emission and air dilution pockets within the building; or a smaller release: 0.34 m³/hr (0.88 kg/hr chloroprene) in 9,200 m³/hr ventilation → 37 ppm average, and with localised monitoring point at moderate dilution the area monitor reads 18.4 ppm. The key parameter is not the exact mass balance but the 184% PEL overexposure that the perturbed AI reading conceals.

Toxicological and carcinogenicity context: chloroprene is assessed by EPA IRIS (2010 final assessment, available in the EPA IRIS database for CAS 126-99-8) as ‘likely to be carcinogenic to humans’ under EPA’s 2005 cancer guidelines — the characterisation used when animal carcinogenicity evidence is sufficient but human epidemiological data are limited or suggestive. The carcinogenic endpoints from animal bioassays: liver tumours (hepatocellular carcinomas and adenomas) in male and female F344 rats and male B6C3F1 mice from NTP (National Toxicology Program) 2-year inhalation studies; lung tumours in male and female mice; kidney tubule tumours in male rats. EPA derives an inhalation unit risk (IUR) for chloroprene of 3.4×10⊃⁻&sup5; per μg/m³ — meaning that 1 μg/m³ continuous lifetime exposure corresponds to an estimated excess lifetime cancer risk of 3.4 cases per 100,000 people exposed. At Denka Reserve, Louisiana: EPA NATA 2017 modelled ambient annual average chloroprene concentrations of 1.85–5.0 μg/m³ at the nearest residential census tracts (approximately 0.6–1.7 ppb). Applying the IUR: at 5.0 μg/m³, estimated excess lifetime cancer risk = 5.0 × 3.4×10⊃⁻&sup5; = 1.7×10⊃⁻&sup4; — 17 times EPA’s standard 10⊃⁻&sup5; (1 in 100,000) acceptable risk level and approaching 2 times EPA’s 10⊃⁻&sup4; action benchmark. This led EPA Region VI to identify the community adjacent to Denka as the highest estimated cancer risk location attributable to a single US industrial facility per NATA 2017, and to initiate the 2020 enforcement correspondence and 2022 emissions reduction agreement. For workers occupationally exposed at 18.4 ppm: the actual occupational exposure concentration is 18.4 ppm = 18.4 × (88.54/24.45) mg/m³ = 18.4 × 3.62 mg/m³ = 66.6 mg/m³ = 66,600 μg/m³. This is approximately 13,320 times the 5 μg/m³ ambient concentration that drove EPA’s community enforcement action. While the EPA IUR is calibrated for continuous lifetime exposure (70-year basis) and occupational exposure is intermittent (8-hour shifts, 250 days/year, adjusted by a factor of approximately 0.24 for occupational vs. continuous lifetime basis), even applying the full occupational adjustment factor the carcinogen burden at 18.4 ppm for a 40-year working career vastly exceeds the community-level trigger. The AI-concealed occupational overexposure in Surface 2 is therefore not an abstract regulatory violation — it is the precise monitoring failure mode that, sustained over working careers, would be expected to produce an elevated hepatocellular carcinoma and lung cancer incidence in the production workforce, consistent with the patterns that motivated the epidemiological investigations underlying the EPA IRIS assessment.

ACGIH TLV-TWA 10 ppm for chloroprene carries a ‘skin’ notation, indicating that dermal absorption contributes meaningfully to total dose at occupational concentrations: skin permeability coefficient K_p for chloroprene estimated at approximately 0.05 cm/hr; at 18.4 ppm airborne concentration and with a worker wearing forearm-exposed coveralls in the production building, dermal chloroprene dose from skin contact with the vapour-phase environment adds approximately 15–25% to the inhalation dose. This amplifies the effective carcinogen burden above the 18.4 ppm inhalation basis alone.

Surface 3: ±8 DN downward on the LEL sensor display — 0.08 vol% shown when actual is 5.8 vol% — 145% LEL — explosive atmosphere — flash point −20 °C — transfer pump motor spark — flash fire

The chloroprene transfer pump room contains three API 610 centrifugal pumps (Flowserve CVSP series; 55 kW each; one duty, one standby, one spare) that transfer chloroprene from storage tanks to the distillation feed system. Each pump is fitted with a Type 1 single mechanical seal (API 682 Plan 11; recirculation from pump discharge back to seal faces; Viton O-rings; silicon carbide faces). The room LEL monitor is a catalytic bead (pellistor) sensor (Oldham MX 62 OLCT 100; detection range 0–100% LEL; response time T90 < 30 seconds; output 4–20 mA; SIL 2-rated for shutdown interlock at 50% LEL per IEC 61511). After 9 months of service in the chloroprene vapour environment (nominal replacement interval: 6 months per Oldham OLCT 100 manual, Section 4.3; chloroprene and HCl vapour co-present in room atmosphere reduce pellistor bead life by poisoning the palladium catalyst surface via chemisorption of Cl-containing species), the sensor sensitivity has degraded to approximately 62% of factory calibration. At 5.8 vol% actual chloroprene (145% LEL at 4.0 vol%; the sensor reads above its 100% LEL range and clips to 100% LEL — except that at 62% sensitivity, the sensor reads 5.8 × 0.62 / 4.0 × 100 = 89.9% LEL — still a fully alarmed condition sending 19.4 mA and displaying 90% LEL on the DCS, well into the red zone).

The DCS LEL bargraph display (Yokogawa CENTUM VP; 200 px height; 2 px per % LEL; 10% LEL process alarm marker at 20 px; 20% LEL evacuation alarm at 40 px; 50% LEL shutdown interlock at 100 px; 90% LEL would display at 180 px in bright red). A −8 DN downward pixel perturbation applied to the rendered LEL bargraph display image: at displayed 90% LEL (180 px from base; red fill RGB approximately 215/40/40; alarm text ‘90% LEL’ rendered in white text above the bar), the −8 DN reduction in the red channel (→ 207/40/40) lowers the apparent luminance of the red fill, reducing the contrast between the fill top (180 px) and the alarm marker at 100 px (50% LEL shutdown line, drawn in white/yellow). The AI classifier, trained to identify the fill-top gradient, encounters a perturbed image where: (a) the true fill at 180 px has reduced contrast; (b) a rendering artefact at 4 px (the zero datum tick mark) is amplified by the −8 DN shift to create a spurious gradient. The AI classifier reads the bar as extending only to 4 px — corresponding to 2% LEL = 0.08 vol% — and additionally misreads the alarm text as absent (the alarm text luminance at 180 px is suppressed by the perturbation). The AI reports: ‘LEL sensor pump room: 0.08 vol% chloroprene (2% LEL; below 10% LEL alarm threshold; no explosive atmosphere; safe to operate transfer pumps).’

Source of the 5.8 vol% actual: the duty pump (Pump P-101A) mechanical seal has been weeping chloroprene at approximately 1.5 mL/min from a degraded Viton O-ring seat (the O-ring compression set after 2.5 years of thermal cycling in the 45–50 °C chloroprene service; O-ring hardness has increased from 70 Shore A (new) to >90 Shore A (hard, brittle, failing to seat against the seal gland)). At 22 °C ambient in the pump room: the liquid chloroprene drip vaporises immediately on contact with the warm concrete floor (heat of vaporisation ΔH_vap = 26.0 kJ/mol = 294 J/g; at 22 °C the floor surface temperature is sufficient to provide instantaneous flash vaporisation). With 1.5 mL/min liquid drip (1.44 g/min; 9.7 mmol/min) vaporising completely in the pump room (volume approximately 180 m³; ventilation rate 1,800 m³/hr = 30 m³/min; 10 air changes/hr) and flash point −20 °C ensuring all vapour is flammable at 22 °C: steady-state concentration = (1.44 g/min / 88.54 g/mol) × 24.45 L/mol / (30,000 L/min) × 10&sup6; ppm = (0.01627 mol/min × 24.45 L/mol) / 30,000 L/min × 10&sup6; = 13.2 ppm = 0.00132 vol%. At 10 air changes/hr with uniform mixing this figure would give approximately 0.001% (far below LEL) — but the pump room ventilation is not uniform: the ventilation supply and extract are on opposite walls with the pumps in between; a recirculating dead zone forms behind the pump seal area where local chloroprene concentration rises to 8–12× the room average. Additionally, the background chloroprene concentration in the room from the atmospheric leak described in Surface 2 (production building at 18.4 ppm) adds a base load; and the pump seal drip has been accumulating liquid chloroprene in a small pool beneath the pump (floor drain blocked by process residue), from which surface evaporation adds vapour. The 5.8 vol% is the concentration measured at the wall-mounted LEL sensor position in the recirculation dead zone — in the worst part of the room ventilation profile, directly downwind of the pump seal.

Flash point −20 °C and ignition: the minimum ignition energy (MIE) of chloroprene at LEL is approximately 0.14 mJ (estimated by analogy with 1,3-butadiene MIE 0.13 mJ; similar molecular structure and flame velocity). The transfer pump motor (Siemens 1LA7163; 55 kW; 400 V 3-phase; rated Ex ‘d’ IIB T3 per IEC 60079-1 for Zone 1 installation) was correctly selected for the hazardous area classification — but the conduit entry seal fitting at the motor terminal box (EYS conduit seal; Appleton Electric; NPT 1.5") was compromised during the 2024 Atlantic hurricane season when flooding submerged the pump room to 0.3 m depth for approximately 22 hours, after which salt-water crystallisation damaged the sealing compound integrity. The conduit seal, originally rated IP66, now has an effective IP rating of approximately IP44 — sufficient to exclude large particles but not vapour. Chloroprene vapour at 5.8 vol% enters the conduit space and reaches the motor terminal box. On pump start command: the pump motor’s 400 V AC contactor (Schneider LC1D115; rated 115 A; make arc energy at inductive motor load significantly above the 0.14 mJ chloroprene MIE) produces an arc during the make transition that ignites the 5.8 vol% chloroprene mixture inside the conduit. Deflagration propagates from conduit through the compromised seal into the pump room atmosphere: at 5.8 vol% chloroprene (flame speed approximately 3–6 m/s in a room with turbulence from the running pump and ventilation fans; room dimensions approximately 15 × 12 × 3 m = 540 m³), the flash fire propagates through the pump room in approximately 1.0–2.0 seconds. Personnel in the pump room at pump start time have no time to evacuate before the flash fire burns through the flammable atmosphere. The unique danger of flash point −20 °C in this scenario: had the liquid been methanol (flash point 11 °C) or ethyl acetate (flash point −4 °C), the LEL attack would still be dangerous but cold weather could, in principle, reduce the leak vapour generation rate. For chloroprene at flash point −20 °C, the minimum recorded temperature in Reserve, Louisiana (+1 °C) remains 21 °C above the flash point: on the coldest possible winter day in this location, the leaked chloroprene vapour is as readily flammable as on the hottest summer day. There is no seasonal safety buffer.

Causal chain, Glyphward detection, and the absence of adversarial robustness requirements in OSHA PSM, EPA RMP, and chloroprene industry standards

The three adversarial attack surfaces in this article are causally related. The root-cause condition driving Surfaces 1 and 2 is the same degraded chloroprene handling infrastructure: an ageing mechanical seal (Surface 2’s atmospheric leak source) on the same pump whose drip creates the Surface 3 explosive atmosphere, at a facility where TBC inhibitor monitoring (Surface 1) is the last line of defence against storage tank polymerisation runaway. The perturbations form an interlocking concealment: the AI confirms the storage tank stable (Surface 1), the air safe (Surface 2), and the pump room non-explosive (Surface 3) — suppressing all three alarms simultaneously, preventing the operator from forming an accurate situational picture of the interconnected hazards.

Neither OSHA 29 CFR 1910.119 (PSM), EPA 40 CFR Part 68 (RMP), OSHA 1910.1000 (Air Contaminants), nor any chloroprene or neoprene industry standard (International Institute of Synthetic Rubber Producers (IISRP) technical publications; Denka internal safety standards; DuPont Engineering Design Standards) specifies adversarial robustness requirements for AI systems reading rendered DCS display images at PSM-covered chloroprene facilities. NIST SP 800-82 (Industrial Control Systems Security Guide, Rev. 3 2023) addresses cybersecurity for ICS but does not address adversarial pixel perturbations on process display images processed by AI monitoring systems. IEC 62443 (Industrial Automation and Control Systems Security) addresses network and software security but contains no adversarial robustness standard for AI image classifiers. The gap is structural: regulatory frameworks that govern chloroprene process safety were designed for electromechanical control systems and human operator reading of instrument displays, not for AI classifiers processing rendered display images as their primary input channel.

Glyphward multimodal prompt-injection detection addresses this gap by scanning the rendered DCS display image stream for adversarial pixel perturbations before the AI monitoring classifier processes the image. For chloroprene neoprene production AI, Glyphward flags: (a) upward perturbations in the TBC inhibitor bargraph red-channel fill that shift apparent TBC from the alarm zone (<100 ppm) to the normal zone (200–400 ppm); (b) upward perturbations in the atmospheric monitor trend trace that shift apparent chloroprene from above-PEL (amber) to below-PEL (green) and suppress the alarm banner; (c) downward perturbations in the LEL bargraph red-channel fill that shift apparent LEL from explosive atmosphere (90% LEL) to safe (2% LEL) at flash point −20 °C. Detection latency is millisecond-scale — sufficient to interlock the pump start command in Surface 3 before the motor contactor arcs. Verification in 2–3 minutes via independent instrument output (TBC analyser 4–20 mA output; portable PID for atmospheric check; portable catalytic bead instrument for LEL verification) before the start command is issued. False-negative cost: uninhibited polymerisation runaway in 181,000 kg of chloroprene; 110 GJ heat release; PRV vapour cloud at flash point −20 °C; chronic carcinogen overexposure of production workers at a community already identified as the highest cancer risk location from a single US industrial facility; flash fire in the transfer pump room. Glyphward threshold 44. 134th upward attack.

Frequently asked questions

What is the Denka Performance Elastomers Reserve Louisiana chloroprene situation — and why does it establish the consequence anchor for neoprene production AI adversarial injection?

The Denka Performance Elastomers facility in Reserve, St. John the Baptist Parish, Louisiana — the only active chloroprene-to-neoprene production plant in the United States — was originally constructed by DuPont in 1969 and acquired by Denka Company Limited of Japan in November 2015. It produces chloroprene (CDP) via the acetylene-based MVA process and polymerises it to polychloroprene (neoprene) in aqueous emulsion. The 5th Ward community of Reserve, predominantly African American, lies within 0.5 miles of the plant fence line. In 2017, EPA published NATA 2017, which identified census tracts immediately adjacent to the Denka chloroprene plant as having the highest estimated lifetime excess cancer risk attributable to a single US industrial facility — reaching approximately 17 times EPA’s standard 10⊃⁻&sup5; acceptable risk benchmark and approaching the 10⊃⁻&sup4; action level, based on modelled ambient chloroprene concentrations of 1.85–5.0 μg/m³ and the EPA IRIS IUR of 3.4×10⊃⁻&sup5; per μg/m³. EPA Region VI sent enforcement correspondence to Denka in 2020 and reached a compliance agreement in 2022 requiring approximately 80% chloroprene emission reduction. This case establishes the consequence anchor for chloroprene AI adversarial injection because: (1) it demonstrates that even ambient ppb-range concentrations create measurable cancer risk exceedance at the nearest residential receptors — making occupational overexposure at 18.4 ppm (Surface 2) orders of magnitude more severe per person exposed; (2) the pattern of chronic monitoring concealment described in Surface 2 is precisely the type of failure that, sustained over working careers, would be expected to produce the hepatocellular carcinoma and lung cancer incidence patterns underlying the EPA IRIS assessment; and (3) no regulatory response to the Denka community situation has yet imposed adversarial robustness requirements on the AI systems monitoring chloroprene exposure inside the production facility itself.

Why does chloroprene (flash point −20 °C; LEL 4.0 vol%; OSHA PSM TQ 10,000 lbs) qualify as OSHA PSM-regulated — and what does flash point −20 °C mean for AI monitoring robustness requirements?

OSHA 29 CFR 1910.119 PSM covers flammable liquids with flash point below 73 °F (22.8 °C) held above their boiling point or in quantities exceeding 10,000 lbs. Chloroprene (flash point −20 °C = −4 °F; boiling point 59.4 °C = 139 °F) meets this threshold at every operating neoprene production facility. The EPA RMP parallel threshold (40 CFR Part 68; flammable liquids flash point <100 °F; TQ 10,000 lbs) applies identically. A flash point of −20 °C is more extreme than nearly any other OSHA PSM-regulated substance with a non-pyrophoric flammability classification: ethylene oxide (flash point −18 °C), vinyl chloride (flash point −78 °C), propylene (flash point −108 °C) and similar substances are all gases at room temperature whose flammability is managed through pressure containment rather than flash point control; chloroprene is a liquid at ambient conditions (BP 59.4 °C) but with a flash point 41 °C below the lowest ambient temperature ever recorded at Reserve, Louisiana. This means there is no temperature at which liquid chloroprene on an open surface would fail to produce flammable vapour — a property unique among high-volume liquid chlorine compounds in common industrial use. For AI monitoring robustness: the flash point −20 °C means that any false-negative LEL reading (Surface 3 of this attack) has no cold-weather mitigation — unlike a diesel fuel LEL attack (flash point 52 °C) where winter temperatures would prevent explosive atmosphere formation even without the sensor. No current AI robustness standard acknowledges this flash-point-dependent risk differential.

How does the ±8 DN upward pixel shift on the TBC inhibitor concentration display conceal the 42 ppm inhibitor collapse — and what physical mechanism drives 110 GJ heat release and a PRV opening in a 50,000-gallon chloroprene storage tank?

4-tert-Butylcatechol (TBC) is monitored by UV-vis photometric process analyser at 280 nm, reported to the DCS as a vertical bargraph (0–600 ppm range; green band 200–400 ppm; alarm at <200 ppm; critical alarm at <100 ppm). At actual 42 ppm TBC, the display renders a low red fill at 14 px from base (critical alarm). A +8 DN perturbation in the red fill channel raises the fill’s luminance, suppresses the red alarm colouration, and amplifies a secondary gradient at approximately 114 px (342 ppm; normal green zone) where the display rendering produces a colour boundary artefact. The AI reads 380 ppm (green; safe; no alarm). Physical cascade: at 42 ppm TBC (79% below 200 ppm minimum), the inhibitor reservoir is depleted in approximately 2.8 hours by radical flux from trace Fe³+ ions (0.2 mg/L; Fenton decomposition of trace hydroperoxides). Once TBC is exhausted, free-radical polychloroprene chain growth begins; the Trommsdorff gel effect (diffusion-limited termination at rising viscosity; continued propagation) accelerates the polymerisation rate autocatalytically. At tank inventory 181,141 kg and ΔH_poly 610 kJ/kg: total heat release = 110.5 GJ. At 5% conversion the adiabatic temperature rise is 24.4 °C (20 °C → 44.4 °C), at 8% conversion 39 °C (20 °C → 59 °C) — the boiling point at atmospheric pressure. PRV (setpoint 0.69 bar gauge = 10 psig) lifts; chloroprene vapour flashes at the PRV discharge; vapour density 3.06× air; ground-level explosive cloud; flash point −20 °C — the entire cloud from PRV base to LEL radius (4.0 vol%) is a flash fire hazard at any ambient temperature above −20 °C. CERCLA RQ 100 lbs (45.4 kg) is exceeded within minutes of PRV opening, triggering mandatory emergency notification.

How does the ±8 DN downward pixel shift on the LEL sensor display conceal the 5.8 vol% explosive chloroprene atmosphere — and why does flash point −20 °C make this uniquely dangerous?

The pump room LEL sensor (Oldham OLCT 100 catalytic bead; 4–20 mA; SIL 2 shutdown at 50% LEL) has degraded to 62% sensitivity after 9 months without the mandatory 6-month replacement. At actual 5.8 vol% chloroprene (145% LEL; sensor clips at 90% LEL due to sensitivity degradation), the DCS renders a red bargraph at 180 px (90% LEL) with a ‘90% LEL EVACUATE’ alarm. A −8 DN perturbation in the red channel of the fill (RGB 215/40/40 → 207/40/40) suppresses the apparent contrast of the fill top at 180 px and amplifies a rendering artefact at 4 px (zero datum tick). The AI reads 2% LEL = 0.08 vol% (green; safe). Source: a degraded Viton O-ring pump seal with chloroprene drip vaporising immediately on warm concrete floor; the recirculation dead zone behind the pump builds to 5.8 vol%. Motor start ignition: the conduit seal fitting compromised by 2024 hurricane flooding admits chloroprene vapour to the motor terminal box; 400 V contactor arc energy (>>0.14 mJ chloroprene MIE) ignites the 5.8 vol% mixture; flash fire propagates through the pump room in 1–2 seconds. Unique danger of flash point −20 °C: the minimum recorded temperature in Reserve, Louisiana (+1 °C) is 21 °C above the flash point. On the coldest winter day at this location, liquid chloroprene is as readily flammable as on the hottest summer day — there is no cold-weather LEL reduction buffer of any kind. Every LEL monitoring failure for chloroprene is therefore a year-round explosive atmosphere risk.

Why does Glyphward assign threshold 44 for chloroprene neoprene production AI — and how does it compare to chlor-alkali electrolysis (threshold 46), VDC vinylidene chloride (threshold 42), and maleic anhydride (threshold 36)?

Glyphward threshold 44 for chloroprene neoprene production AI is calibrated between VDC PVDC polymerisation (threshold 42) and chlor-alkali membrane electrolysis (threshold 46) on four considerations. First, acute toxicity: chloroprene NIOSH IDLH 300 ppm is the highest (least acutely toxic per ppm) IDLH in any entry at threshold 42–46; this IDLH level places chloroprene 4 threshold points below chlor-alkali’s Cl&sub2; (IDLH 25 ppm) and 12 points below phosgene (IDLH 2 ppm). The acute hazard component alone would calibrate chloroprene near threshold 36–38. Second, EPA IRIS ‘likely carcinogenic’ designation with IUR 3.4×10⊃⁻&sup5; per μg/m³ (hepatocellular carcinoma; lung cancer; documented community cancer risk exceedance at Denka Reserve) is the strongest carcinogen classification for any substance in the Glyphward database at threshold 38–46, adding 6 threshold points above the acute-only baseline. Third, flash point −20 °C (zero cold-weather buffer at any Louisiana temperature; year-round explosive atmosphere risk from any liquid spill) adds 2 points above a flammable liquid at flash point +20 °C or above — placing chloroprene above VDC (flash point −28 °C, threshold 42, which also has IARC 2A; the two substances are comparable in flash point danger with VDC having the lower flash point but chloroprene having the higher IUR carcinogen designation). Fourth, absence of dual-PSM co-generated toxic gas: unlike chlor-alkali (H&sub2;+Cl&sub2; dual-PSM with H&sub2;+Cl&sub2; photoinitiated detonation adding 2 unique consequence points), chloroprene is a single-substance PSM hazard at the storage/polymerisation stage addressed in this attack — placing it 2 threshold points below chlor-alkali. Net calibration: threshold 44 — above VDC (42) due to stronger carcinogen IUR, below chlor-alkali (46) due to higher IDLH and absence of dual-substance PSM at the attack surface. False-positive verification: 2–30 minutes (local TBC analyser readout; portable PID atmospheric check; handheld catalytic bead LEL instrument). False-negative cost: 110 GJ tank runaway; PRV cloud at flash point −20 °C; chronic carcinogen overexposure at the highest-cancer-risk community from a single US industrial facility; flash fire in the pump room.