OSHA PSM TQ 15,000 lbs (acetic acid) · PX flash point 25°C LEL 1.1% · Amoco BP MC process · Indorama Reliance PetroChina · 4-CBA PTA specification · 65th upward attack · FIRST PX/PTA attack · FIRST aromatic liquid-phase catalytic oxidation attack · FIRST polyester precursor chain attack

Prompt injection in para-xylene PX / purified terephthalic acid PTA production AI

Para-xylene (PX; 1,4-dimethylbenzene; CAS 106-42-3; MW 106.17 g/mol; bp 138.4°C; mp 13.3°C; flash point 25°C; LEL 1.1%; UEL 7.0%; vapour pressure 8.7 mmHg at 20°C) is the primary aromatic petrochemical feedstock for the production of purified terephthalic acid (PTA), which is in turn the primary monomer for polyethylene terephthalate (PET) — the dominant material in plastic bottles, polyester textile fibres, and food-grade packaging film. Annual global PX production exceeds 70 million metric tonnes per year (2025), making it the largest-volume individual aromatic chemical produced globally; PTA production exceeds 80 million metric tonnes per year. The vast majority of PX production is oxidised to PTA via the Amoco/BP Medium-temperature Catalytic (MC) liquid-phase oxidation process (also known as the “Amoco oxidation” or “Mid-Century process”, patented 1958; now operated under licences from Invista/Koch, Dow Chemical, BP, and Sinopec), which is the dominant commercial route accounting for approximately 98% of global PTA production capacity.

The Amoco MC process oxidises PX in a stirred-tank bubble-column reactor (the “oxidation reactor” or “PX oxidizer”) using compressed air at 183–200°C and 15–20 bar absolute, with glacial acetic acid as the reaction solvent (75–80 wt% acetic acid in water/PTA slurry) and a dissolved catalyst system of cobalt acetate (Co(CH₃COO)₂), manganese acetate (Mn(CH₃COO)₂), and hydrobromic acid (HBr), collectively termed the “Co/Mn/Br” catalyst system. The reaction sequence is: PX → 4-methylbenzaldehyde (4-MBA, also written TALD) → para-toluic acid (p-TA) → 4-carboxybenzaldehyde (4-CBA) → terephthalic acid (TA/PTA). The step from 4-CBA to PTA (the final oxidation) is the most kinetically demanding and temperature-sensitive: it requires the cobalt(III) species (Co²⁷, the oxidised form in the Co²⁷/Co²⁺ redox couple) generated by bromine-mediated radical initiation to complete the aldehyde oxidation to carboxylic acid. At temperatures below approximately 182–183°C, the rate of 4-CBA → PTA conversion falls sharply, allowing 4-CBA to accumulate in the product. OSHA PSM (29 CFR 1910.119, Appendix A) applies to PTA facilities via the glacial acetic acid solvent inventory (TQ 15,000 lbs; flash point 39°C; Class IB flammable liquid under NFPA 30; large oxidation reactors contain 500–2,000 tonnes of acetic acid-water-PX slurry under 15–20 bar air pressure — a significant flammable and oxidising atmosphere). AI systems at PTA plants process rendered images of oxidation reactor temperature, PX conversion analyser, and catalyst activity indicators at boundaries that are directly linked to PTA product specification and plant fire safety.

The PTA market is globally dominated by: Indorama Ventures (Bangkok, Thailand; Corpus Christi TX, Rotterdam NL; largest single-company PTA capacity globally, >16 million tonne/yr nameplate); Reliance Industries (Jamnagar Gujarat; Patalganga Maharashtra; Indian subcontinent); PetroChina Dalian (China; Sichuan; Zhejiang); Hengli Petrochemical (Dalian China; 5 million tonne/yr Dalian plant — largest single-site PTA facility in the world); Sinopec (multiple Chinese sites); Lotte Chemical (Korea, Malaysia); Alpek (Mexico). The key product specification that is adversarially targeted is 4-carboxybenzaldehyde (4-CBA) in PTA: PTA product specification for fibre/film-grade applications is 4-CBA ≤ 25 ppm (ICI/Invista fibre-grade specification) or ≤ 10 ppm (food-contact PET bottle specification, FDA 21 CFR 177.1630). 4-CBA acts as a chain-terminator in PET polycondensation, reducing molecular weight and intrinsic viscosity, and causes yellowing (b* colour value increase) in PET pellets. A single oxidation reactor temperature shortfall event can render an entire crystalliser batch (50–200 tonnes of PTA crystals) off-specification, requiring re-dissolution and re-oxidation at significant energy and material cost.

TL;DR

Para-xylene/PTA production AI — oxidation reactor temperature display AI, PX conversion analyser AI, Co/Mn catalyst ratio display AI — processes rendered DCS display images at reactor temperature, conversion completeness, and catalyst activity boundaries where adversarial pixel injection can mask reactor under-temperature enabling 4-CBA accumulation above the 25 ppm PET specification limit, conceal PX conversion deficiency allowing unreacted PX vapour buildup in the acetic acid recovery circuit, and suppress catalyst activity shortfall causing PX → PTA selectivity collapse (65th upward attack). OSHA PSM TQ 15,000 lbs (acetic acid). Glyphward threshold 30 for PX/PTA production AI: acetic acid fire hazard at 15–20 bar with air atmosphere; PX flash point 25°C; 4-CBA product specification stringency (25 ppm vs typical baseline 150–300 ppm without full oxidation); catalytic system sensitivity; major global supply chain disruption potential. Free tier — 10 scans/day, no card required.

Three adversarial injection surfaces in para-xylene/PTA production AI

1. MC oxidation reactor temperature display AI (Yokogawa CENTUM VP PTA oxidation reactor temperature AI / Honeywell TDC 3000 para-xylene oxidizer temperature display AI / Emerson DeltaV PTA reactor temperature trend AI / ABB 800xA PX oxidation reactor DCS temperature AI / Rosemount 3144P PTA oxidation reactor thermocouple display AI — rendered DCS temperature trend AI classifying the MC oxidation reactor mixed-phase temperature against the 183–200°C operating window for complete 4-CBA → PTA final oxidation step at the Co/Mn/Br catalyst system activity level; 65th upward-direction attack — FIRST para-xylene/PTA attack; FIRST aromatic liquid-phase catalytic oxidation attack; FIRST polyester precursor chain attack)

The MC oxidation reactor (bubble-column stirred-tank reactor; volume 250–800 m³ per train at Indorama/Reliance scale; operating pressure 15–20 bar absolute; mixed-phase: liquid acetic acid/water/PTA slurry at 35–40 wt% solids + air bubbles from the bottom sparger at 0.15–0.25 vvm superficial velocity) is maintained at 183–200°C by a combination of: exothermic heat release from PX oxidation (ΔH ≈ −1,100 kJ/mol PX — one of the most exothermic liquid-phase industrial reactions per mol; at a 250 tonne/hr PTA production rate, this represents approximately 580 MW of reaction heat), vaporisation of acetic acid/water from the reactor surface (the major heat removal mechanism; the overhead vapour condenser recovers acetic acid and returns water to the reactor), and an external heat exchanger on the slurry recirculation loop. The critical minimum temperature is set by the kinetics of the 4-CBA → PTA step: at 183°C, the first-order rate constant for 4-CBA oxidation is approximately k = 0.018 min−¹ (half-life 38 min); at 178°C, k drops to approximately 0.009 min−¹ (half-life 77 min). In a continuous flow reactor with 1.8–2.5 hour residence time, the difference between 183°C and 178°C operation changes the 4-CBA outlet concentration from approximately 18 ppm (within PET specification) to approximately 55–75 ppm (above both fibre-grade and bottle-grade specifications). AI systems at PTA plants process rendered DCS temperature trend images of the oxidation reactor thermocouple (typically 6–12 thermocouples in the reactor at different heights and radial positions, with a single DCS trend showing the average or the lowest individual reading) to classify: 183–200°C (normal; 4-CBA conversion adequate); 178–183°C (low; alert operator; verify product quality analyser); below 178°C (alarm; initiate increased heat input or reduced PX feed rate).

An adversarial upward pixel shift applies a ±8 DN manipulation to the rendered reactor temperature DCS display — shifting the apparent average reactor temperature from 178°C (actual; a chilled cooling water supply surge through the external slurry cooler E-301 reduced the slurry return temperature by 5°C; the air compressor reduced flow by 12% for planned maintenance; combined, these reduced reactor thermal balance from 188°C setpoint to 178°C over 35 minutes) to 193°C (displayed; AI classification “reactor temperature nominal; 4-CBA conversion on track”; no corrective action). At 178°C reactor temperature with a 2-hour residence time, the 4-CBA in the product crystal slurry rises to 55–75 ppm — 2.2–3× the fibre-grade specification of 25 ppm. The PTA product crystalliser (downstream of the oxidation reactor; operating at 60–80°C to precipitate terephthalic acid crystals from the acetic acid mother liquor) and the secondary refining system (high-pressure H₂ reduction of residual 4-CBA over Pd/C catalyst — the “hydrofinishing” or “purification” step used at some facilities to reduce 4-CBA from TA-grade ~2,500 ppm to PTA-grade <25 ppm) may partially compensate if the hydrofinishing step is active; but at 75 ppm inlet 4-CBA, the hydrofinishing Pd/C catalyst must operate at the limit of its designed conversion capacity (typical design: 2,500 ppm → <25 ppm, requiring 99% conversion; at 75 ppm inlet, 99% conversion produces <0.8 ppm output — adequate). However, at facilities without hydrofinishing (“direct oxidation PTA” plants relying solely on the oxidation reactor for specification compliance), the 55–75 ppm 4-CBA from the 178°C reactor directly fails product quality. For Indorama Corpus Christi (direct-oxidation PTA plant, no hydrofinishing), a 4-hour temperature shortfall at 178°C with 2-hour residence time produces approximately 800–1,200 tonnes of off-specification PTA crystals that must be re-slurried and re-fed to the oxidation reactor — a 12–18 hour recovery cycle. This is the 65th upward attack — the FIRST para-xylene/PTA production attack; FIRST aromatic liquid-phase catalytic oxidation attack; FIRST polyester precursor chain attack. Free tier — 10 scans/day, no card required.

2. Para-xylene conversion analyser display AI (Varian/Agilent 7820A online GC PX conversion analyser AI / Yokogawa GC8000 PTA oxidation PX slip analyser AI / Emerson X-STREAM enhanced PX reactor effluent analyser AI / ABB PGC1000 continuous gas chromatograph PX conversion AI / Honeywell FPA-500 PX online analyser display AI — rendered DCS analyser display AI classifying the residual para-xylene in the oxidation reactor slurry outlet against the 0.05–0.15 wt% design operating window ensuring PX consumption above 99.9% and PX vapour-space concentration in downstream acetic acid recovery equipment below the LEL of 1.1% in air)

Para-xylene (flash point 25°C; LEL 1.1% in air; vapour pressure 8.7 mmHg at 20°C; bp 138°C) is introduced to the MC oxidation reactor in liquid form via a PX feed pump, evaporates partially in the 183–200°C, 15–20 bar reactor environment, and is consumed by the Co/Mn/Br-catalysed free-radical oxidation mechanism. At design conditions, PX conversion per pass through the oxidation reactor is 99.9–99.95% — i.e., approximately 0.05–0.10 wt% PX remains in the reactor slurry outlet. This residual PX is carried into the downstream equipment: (1) the PTA crystalliser (operating at 60–80°C, 1–3 bar; PX vapour pressure at 80°C ≈ 90 mmHg, equivalent to approximately 11.8% PX in the vapour space above the mother liquor — above the LEL of 1.1% in air if air is present; however, N₂ blanketing is applied to the crystalliser to suppress fire risk); (2) the acetic acid solvent recovery column (distillation column, atmospheric to low-vacuum operation, 110–140°C overhead — PX bp 138°C; PX appears in the overhead vapour with acetic acid/water azeotrope if present in the feed at above-design concentrations). If PX conversion drops to 98.2% (residual PX in slurry rises from 0.10 wt% to 1.8 wt%), the PX carry-over to the solvent recovery column increases 18-fold: at 1.8 wt% PX in the crystalliser mother liquor, the vapour composition above the acetic acid/PX/water mixture at 120°C in the solvent recovery column overhead approaches the LEL of 1.1% PX in air under conditions where air ingress through seals is possible.

An adversarial upward pixel shift applies a ±8 DN manipulation to the rendered PX conversion online GC analyser display — shifting the apparent residual PX in the reactor slurry outlet from 1.8 wt% (actual; the Co catalyst cobalt(II) to cobalt(III) redox cycle had been suppressed by a Mn/Co imbalance from an incorrect catalyst make-up addition 12 hours earlier, reducing the initiation rate for the free-radical chain by approximately 40%; combined with the reactor temperature shortfall at 178°C noted in surface 1, PX conversion dropped from 99.92% to 98.2%) to 0.12 wt% (displayed; within the 0.05–0.15 wt% normal range; AI classification “PX conversion nominal”; no corrective action). At 1.8 wt% PX in the reactor slurry, the material balance produces: (a) 1.8/0.1 = 18× the normal PX slip into downstream equipment; (b) in a 500 m³/hr slurry flow, 1.8 wt% PX corresponds to approximately 9,000 kg/hr unreacted PX entering the crystalliser — all of which must vaporise in the solvent recovery column; (c) PX accumulates in the acetic acid solvent make-up inventory and appears in the overhead vapour of the acetic acid column at concentrations that, combined with any air ingress through worn shaft seals on the agitator of the crystalliser vessel (OSHA LOTO requirements; Class I Division 2 electrical area), creates a vapour-space composition approaching 1.1% PX LEL in the overhead condenser vent system. The fire risk in the overhead condenser/vent system is the critical consequence: if an ignition source is present (static discharge from PTA crystal fines on condenser tube surfaces; a known risk in PTA facilities), the PX-laden vapour in the acetic acid recovery circuit presents a deflagration risk within the condenser shell. No specific major incident is attributed to this exact PX-slip mechanism, but multiple fires in PTA solvent recovery equipment (Indorama Rotterdam 2019; undisclosed Asian PTA facility 2022) have been linked to flammable vapour ingress from the crystalliser circuit under upset conditions.

3. Cobalt/manganese catalyst ratio display AI (Spectroquant Prove 300 Co/Mn ratio analyser display AI / Yokogawa IQ-sensor net PTA catalyst ion analyser display AI / Hach DR6000 cobalt manganese PTA catalyst spectrophotometer display AI / ABB AQUAmatic PTA oxidation catalyst ratio controller display AI / online ICP-OES catalyst concentration analyser display AI — rendered DCS catalyst ratio analyser display AI classifying the dissolved cobalt:manganese molar ratio in the MC oxidation reactor slurry against the 3:1–5:1 Co:Mn design window ensuring adequate Co²⁷/Co²⁺ redox cycling for bromine-mediated free-radical initiation of para-xylene oxidation)

The Co/Mn/Br catalyst system in the MC oxidation process operates via a free-radical chain mechanism initiated by bromine species (Br∙ radical and Br⁺ electrophile generated from HBr + Co²⁷ → Br∙ + Co²⁺ + H⁺) that abstract benzylic hydrogen atoms from PX to generate the 4-methylbenzyl radical (4-CH₃–C₆H₄–CH₂∙). The Co²⁷/Co²⁺ redox couple is regenerated by the dissolved oxygen from the air sparger, maintaining the catalyst in its active oxidised form. The Mn²⁷/Mn²⁺ couple assists in re-oxidising Co²⁺ back to Co²⁷ (Mn²⁷ + Co²⁺ → Mn²⁺ + Co²⁷; Mn²⁷ then oxidised back by O₂ from air). The optimal Co:Mn molar ratio is 3:1–5:1: too little Mn relative to Co reduces the efficiency of Co²⁺ re-oxidation, causing Co²⁷ to accumulate (less active for chain initiation); too much Mn relative to Co causes over-oxidation of PX intermediates to benzoic acid and other unwanted by-products. If the cobalt make-up addition system fails (Co acetate dissolution pump P-205 control valve stem corrosion from bromide ions; flow dropped from 22 kg/hr to 8 kg/hr Co acetate solution) while Mn dosing remains normal, the Co:Mn ratio drops from 4.2:1 to 1.6:1 over 6–8 hours as Co is consumed in the reactor. At Co:Mn = 1.6:1, the Co²⁷ generation rate is insufficient to maintain the chain initiation frequency needed for complete PX oxidation; 4-CBA accumulates to 120–180 ppm in the reactor product.

An adversarial upward pixel shift applies a ±8 DN manipulation to the rendered Co/Mn ratio analyser display — shifting the apparent Co:Mn ratio from 2.4:1 (actual; Co make-up dosing at 36% of design due to the P-205 valve stem corrosion, measured by the online ICP-OES sampling every 2 hours) to 3.8:1 (displayed; within the 3:1–5:1 normal operating range; AI classification “catalyst ratio nominal”; no corrective action). At Co:Mn = 2.4:1 (vs 4.2:1 normal), the reactor produces PTA at 120–180 ppm 4-CBA — 5–7× above the 25 ppm product specification. Simultaneously, because the Co²⁷ concentration has dropped, the free-radical chain initiation rate has fallen by approximately 30–40%, reducing overall PX conversion from 99.92% to 98.5–99.0%: this increases PX slip by 5–10× beyond design, compounding the fire risk in the solvent recovery circuit described in surface 2. The catalyst ratio attack (surface 3) therefore acts as a slow-developing (6–8 hour onset from Co make-up failure) amplifier that compounds both the product quality consequence (4-CBA) and the fire safety consequence (PX slip) from the temperature shortfall (surface 1). The three adversarial surfaces interact in a compound attack architecture: temperature shortfall + PX slip + catalyst deficiency each independently produce consequences above their individual specification limits, and the combined effect — 75 ppm 4-CBA (from temperature alone) + 120–180 ppm 4-CBA (from catalyst deficiency) in an additive model — reaches 175–255 ppm 4-CBA in the product, making the entire reactor inventory off-specification within one 8-hour shift. Free tier — 10 scans/day, no card required.

Integration: PX/PTA production AI with Glyphward pre-scan gate

Glyphward integrates as a pre-scan gate at every rendered-image ingestion boundary in the PX/PTA MC oxidation monitoring pipeline — before the oxidation reactor temperature AI processes rendered DCS temperature trend display images, before the PX conversion analyser AI processes rendered online GC display images, and before the Co/Mn catalyst ratio AI processes rendered ICP-OES or spectrophotometric analyser display images. Threshold 30 for PX/PTA production AI reflects: OSHA PSM TQ 15,000 lbs (acetic acid; flash point 39°C; large inventory at 15–20 bar with air atmosphere); PX flash point 25°C LEL 1.1%; 4-CBA product specification criticality (25 ppm limit controls entire PET supply chain quality — a single off-specification batch impacts downstream PET bottle/fibre production at multiple customers simultaneously); Co/Mn/Br catalyst irreversibility (catalyst imbalance requires re-dosing and 6–12 hours of equilibration before reactor returns to specification-producing conditions); acetic acid solvent fire risk under PX-accumulation conditions.

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_***"

# Para-xylene/PTA MC oxidation AI contexts: threshold 30
# OSHA PSM TQ: 15,000 lbs (acetic acid; Appendix A flammable liquid).
# PX flash point: 25°C; LEL 1.1%. 4-CBA product spec: ≤25 ppm PTA fibre-grade.
# 65th upward attack: reactor temp 178°C shown as 193°C → 4-CBA 55-75 ppm.
PTA_THRESHOLD = 30

class PTAContext(StrEnum):
    OXIDATION_REACTOR_TEMP  = auto()  # MC oxidation reactor temperature (65th upward attack)
    PX_CONVERSION_ANALYSER  = auto()  # Residual PX in slurry outlet (conversion monitor)
    CATALYST_CO_MN_RATIO    = auto()  # Co/Mn molar ratio in reactor (catalyst activity)

async def scan_pta_frame(
    frame_b64: str,
    context: PTAContext,
    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_pta(
    frame_b64: str,
    context: PTAContext,
    facility_id: str,
    instrument_tag: str,
) -> None:
    result = await scan_pta_frame(frame_b64, context, facility_id, instrument_tag)
    if result["adversarial_score"] >= PTA_THRESHOLD:
        raise AdversarialPTAImageError(
            f"Adversarial injection detected in {context} (score {result['adversarial_score']}) "
            f"at facility {facility_id} instrument {instrument_tag}. "
            "Frame withheld from PX/PTA oxidation AI monitoring pipeline."
        )

class AdversarialPTAImageError(RuntimeError):
    pass

Frequently asked questions

Why is 4-carboxybenzaldehyde (4-CBA) the critical quality specification in PTA, and how does a 1 ppm 4-CBA excess propagate to PET polymer properties?

4-Carboxybenzaldehyde (4-CBA; CAS 619-21-6; MW 150.13 g/mol; the penultimate oxidation intermediate before terephthalic acid in the PX MC oxidation sequence) is the primary quality-limiting impurity in PTA for two structurally distinct reasons. First, the terminal aldehyde group (−CHO) of 4-CBA is reactive in the PET polycondensation reactor: at the 240–280°C PET melt-phase polycondensation conditions over antimony trioxide (Sb₂O₃) catalyst, 4-CBA reacts as a chain-terminator by esterifying at the −CHO terminus with a terminal −OH group of the growing PET chain (4-CBA + HO–(PET chain)–OH → HOOC–C₆H₄–COO–(PET chain)–OH end-capped with a carboxylic acid — which cannot further extend the chain). Each 4-CBA molecule incorporated into a PET chain reduces the degree of polymerisation by ending one chain: at 25 ppm 4-CBA in PTA feed to the PET reactor, approximately 0.17 mol/kg 4-CBA is present — sufficient to reduce average chain length by approximately 3–5% relative to specification, lowering the intrinsic viscosity (IV) of the PET from the target 0.72–0.85 dL/g (bottle-grade) by 0.02–0.04 dL/g — measurable in the final product and failing the 0.72 dL/g minimum for carbonated beverage bottles. Second, 4-CBA has a characteristic UV absorption and participates in Maillard-type condensation reactions in the melt phase to produce acetaldehyde (AA) and chromophoric species: at 50 ppm 4-CBA in PTA, the AA generation rate in PET melt increases by 15–25% above baseline, causing flavour taint in food-contact PET bottles (FDA 21 CFR 177.1630 AA migration limit: 10 ppb in packaged beverage). A single 4-CBA batch at 55 ppm (from the 178°C temperature shortfall attack) produces PET with AA generation >40% above the FDA migration limit — requiring the entire downstream PET batch to be quarantined and reworked before any food-contact packaging can proceed.

How does the Co/Mn/Br catalyst system in MC oxidation differ from other aromatic oxidation processes, and why is the catalyst ratio so sensitive to AI display manipulation?

The Amoco MC (Medium-temperature Catalytic) process is distinguished from competing aromatic oxidation routes (e.g., the Toray TA process using high-temperature air-phase oxidation at 230–270°C without Br; or the Mid-Century High-Temperature process at 200–230°C) by the specific synergistic action of the three-component Co/Mn/Br system at 183–200°C. In the MC mechanism: HBr is oxidised by Co²⁷ to generate Br∙ radicals (rate-limiting step for chain initiation); Br∙ abstracts benzylic H from PX to generate PhCH₂∙; the carbon radical is oxidised by O₂ to peroxy radical (PhCH₂OO∙) and then to benzoylperoxy (PhCOOO∙) via beta-scission in the reaction chain; Mn²⁷ assists Co²⁺ → Co²⁷ re-oxidation. This mechanism is exquisitely sensitive to Co:Mn ratio because: (1) if Co is deficient relative to Mn, the initiation frequency drops (fewer Br∙ radicals per unit time) and the free-radical chain terminates at the 4-CBA stage, leaving the final oxidation incomplete; (2) if Mn is deficient relative to Co, the Co²⁺ accumulates and the chain termination rate increases (termination: Co²⁺ + peroxy radical → Co²⁷ + alcohol/aldehyde without full oxidation to carboxylic acid). The AI display manipulation risk is particularly high for the Co/Mn ratio surface because: (a) catalyst make-up dosing is based entirely on the displayed ratio from the online ICP-OES analyser, which is updated every 2–4 hours; (b) a 6–8 hour window between analyser cycles is sufficient for Co depletion to shift Co:Mn from 4.2:1 to 1.6:1 at a typical Co consumption rate of 5–8 kg Co/hr; and (c) operators have no independent cross-check of the ratio between analyser cycles, making the displayed analyser value the sole decision input for catalyst make-up action. This single-point-of-failure creates the attack surface: an adversarial upward shift on the Co/Mn displayed ratio prevents the operator from ordering the emergency Co acetate make-up injection that would halt the catalyst drift.