# Unplanned Cracking Unit Downtime

*/Problems/Unplanned_Cracking_Unit_Downtime*

## Problem Overview

Refinery operations managers experience unplanned shutdowns of fluid catalytic or steam cracking units when internal coking, thermal stress, or catalyst degradation exceed operational thresholds. A single day of unexpected downtime halts the conversion of heavy crude into high-value distillates, costing millions of dollars in lost yield and emergency maintenance. Because these units sit at the critical bottleneck of the refining process flow, any sudden outage immediately cascades through downstream chemical synthesis and blending operations.

The internal conditions of a cracking unit are notoriously opaque and volatile, driven by constantly changing feedstock compositions and extreme operating temperatures. Current reliability strategies rely on static schedules or lagging indicators like external temperature drops and pressure differentials, which only trigger alerts after a localized failure or heavy fouling is already underway. Engineers lack dynamic, predictive visibility into the specific chemical fouling rates and metallurgical stress points as they evolve in real time.

Traditional physics-based simulation tools cannot process the high-frequency sensor data required to map localized coke deposition or catalyst deactivation on the fly. As refineries process increasingly heavy and variable crude blends, the margin for operational error shrinks. Operators are forced to choose between running units conservatively, which sacrifices yield, or running near the limit and risking catastrophic mechanical failure.

## Problem Severity Frequency

_Illustrative — target and order-of-magnitude estimate figures, not an achieved track record (this Thing is concept-stage)._

**Severity**: 5
**Frequency**: event-driven
**Budget Reality**:
- **Price Ceiling**: ~$150k–500k/yr per refinery — anchors to existing enterprise asset performance management (APM) software budgets, not the $10M+ cost-of-pain
- **Who Controls Spend**: VP of Operations or Plant Manager signs, Reliability Engineering Manager recommends
- **Existing Budget Line**: true
- **Switching Cost From Status Quo**: high: requires complex data integration with plant historians (e.g. OSISoft PI), tuning machine learning models to specific unit metallurgy, and overcoming process engineers' deep skepticism of black-box AI algorithms touching safety-critical systems
**Regulatory Risk**: high
**Time Cost Per Event**: ~3–10 days
**Money Cost Per Event**: ~$2M–10M lost yield and emergency repairs
**Annual Cost Per Affected Entity**: ~$10M–25M all-in (including margin sacrificed running conservatively)

## Problem Why Now

Refineries currently process increasingly heavy, sour, and highly variable opportunity crudes to defend shrinking margins. This shift pushes fluid catalytic and steam cracking units past their historical baselines, accelerating internal coking and metallurgical stress at unpredictable rates. With average downtime costs exceeding $1 million per day per industry estimates circa 2023, operators can no longer rely on static maintenance schedules built for stable feedstock.

Until recently, predicting localized fouling required complex computational fluid dynamics models that took days to process, rendering them useless for active operational decisions. Today, physics-informed neural networks combine thermodynamic rules with high-frequency acoustic and temperature sensor streams at the edge. This specific technological threshold allows continuous, sub-second mapping of catalyst deactivation and coke deposition before lagging indicators trigger a shutdown.

## Problem Current Solutions

**Status Quo**: Reliability engineers monitor lagging indicators like pressure drops in plant historians and rely on static physics-based simulations to schedule preventative maintenance. Operators typically run cracking units below maximum capacity to create a safety buffer against unexpected fouling and thermal stress.
**Workarounds**:
- running units below maximum capacity
- exporting historian data to spreadsheets
- relying on static decoking schedules
- setting reactive pressure drop alarms
**Named Tools In Use**:
- [OSIsoft PI System](/Products/OSIsoft_PI_System)
- [Aspen HYSYS](/Products/Aspen_HYSYS)
- [GE Digital APM](/Products/GE_Digital_APM)
- [Honeywell Forge](/Products/Honeywell_Forge)
- [Aveva Predictive Analytics](/Products/Aveva_Predictive_Analytics)
**Why Insufficient**: Traditional physics-based simulators cannot process high-frequency sensor data on the fly, while existing asset management platforms rely on lagging indicators that only trigger after heavy fouling is already underway. These systems cannot dynamically predict localized coke deposition or catalyst deactivation in real time as complex feedstock compositions change.

## Problem Market Profile

**Incumbents**:
- [OSIsoft PI System](/Problems/Unplanned_Cracking_Unit_Downtime/Competitors/OSIsoft_PI_System)
- [Aspen HYSYS](/Problems/Unplanned_Cracking_Unit_Downtime/Competitors/Aspen_HYSYS)
- [GE Digital APM](/Problems/Unplanned_Cracking_Unit_Downtime/Competitors/GE_Digital_APM)
- [Honeywell Forge](/Problems/Unplanned_Cracking_Unit_Downtime/Competitors/Honeywell_Forge)
- [Aveva Predictive Analytics](/Problems/Unplanned_Cracking_Unit_Downtime/Competitors/Aveva_Predictive_Analytics)
**Substitutes**:
- Running units below maximum capacity
- Exporting historian data to spreadsheets
- Relying on static decoking schedules
- Setting reactive pressure drop alarms
**Position Axes**:
- Data Latency (Historical/Batch vs. High-Frequency Live Data)
- Insight Generation (First-Principles Physics vs. Data-Driven ML)
**Market Dynamics**: The market is consolidating as heavy industrial software providers attempt to bolt predictive machine learning modules onto legacy data historian architectures. Simultaneously, independent industrial AI vendors are attempting to bypass traditional asset management platforms by reading directly from edge IoT sensors.
**Competition Concentration**: Incumbents like Aspen HYSYS cluster in the batch-latency, first-principles physics quadrant, providing high-fidelity models that lack real-time operational utility. Platforms like OSIsoft PI and GE Digital APM dominate the live data space but pair it primarily with historical rules and reactive alarms. The quadrant combining high-frequency live data ingestion with dynamic, data-driven ML forecasting remains sparsely populated, as legacy platforms struggle to model localized catalyst degradation at runtime.

## Mint Vocabulary Bag

**Action Verbs**:
- calibrate
- monitor
- purge
- throttle
- bypass
- lubricate
- inspect
**Gerund Stems**:
- monitor
- calibrat
- purg
- throttl
- lubricat
- inspect
**Abstract Nouns**:
- fatigue
- erosion
- thermal
- pressure
- corrosion
- amplitude
- viscosity
**Concrete Nouns**:
- reactor
- catalyst
- manifold
- impeller
- furnace
- valve
- sensor
**Metaphor Nouns**:
- anchor
- pulse
- keel
- tether
- steady
- gauge
- core
**Structure Nouns**:
- vessel
- plenum
- stack
- column
- array
- chamber

## Problem Candidate Solutions

- [Detonationridge](/Problems/Unplanned_Cracking_Unit_Downtime/Startups/Detonationridge) — Software
- [Plenumgate](/Problems/Unplanned_Cracking_Unit_Downtime/Startups/Plenumgate) — Agent
- [Monitorharbor](/Problems/Unplanned_Cracking_Unit_Downtime/Startups/Monitorharbor) — Service-as-Software
- [Impellerhaven](/Problems/Unplanned_Cracking_Unit_Downtime/Startups/Impellerhaven) — Software
- [Lucen](/Problems/Unplanned_Cracking_Unit_Downtime/Startups/Lucen) — Agent
- [Agack](/Problems/Unplanned_Cracking_Unit_Downtime/Startups/Agack) — Service-as-Software

## Problem Solution Space2x2

```mermaid
quadrantChart
    x-axis Component Level --> System Level
    y-axis Vibration Data --> Thermodynamic Data
    Detonationridge: [0.2, 0.3]
    Plenumgate: [0.8, 0.8]
    Monitorharbor: [0.7, 0.2]
    Impellerhaven: [0.3, 0.4]
    Lucen: [0.9, 0.7]
    Agack: [0.3, 0.8]
```

## Problem Affected Roles

- Refinery Operations Manager — Plant Leadership
- Reliability Engineer — Equipment Maintenance
- Unit Process Engineer — FCC Optimization
- Production Planner — Yield Scheduling
- Turnaround Manager — Outage Planning
- Process Control Engineer — Automation
- Blending Supervisor — Downstream Operations

## Problem Affected Companies

- Petroleum Refineries — Downstream O&G
- Petrochemical Manufacturers — Chemical Production
- Integrated Energy Producers — Oil & Gas
- Heavy Oil Upgraders — Midstream Processing
- Olefin Production Facilities — Petrochemicals
- Independent Oil Refiners — Independent Refining
- Downstream Polymer Manufacturers — Derivatives

## Problem Affected Processes

- Fluid Catalytic Cracking — Core Operations
- Turnaround Planning — Maintenance Strategy
- Feedstock Optimization — Input Blending
- Emergency Maintenance Execution — Repair Operations
- Catalyst Lifecycle Management — Material Management
- Production Scheduling — Yield Management
- Asset Integrity Management — Reliability

## Problem Matching Opportunities

- Acoustic Anomaly Detection for Refineries — Audio Diagnostics
- Predictive Catalyst Tracking for Petrochemicals — Time Series Analytics
- Automated Thermal Inspection for Downstream — Computer Vision
- Algorithmic Vibration Analysis for Hydrocrackers — Sensor Processing
- Dynamic Feedstock Profiling for Refineries — Process Optimization

## Problem Token Hero

**Genre**: problem-hero
**Rendered**: Refinery operations managers experience unplanned shutdowns of fluid catalytic or steam cracking units when internal coking, thermal stress, or catalyst degradation exceed operational thresholds.
**Mechanism**: overview-derived-v1
**Template Id**: problem-overview-derived
**Vocab Fingerprint**: 3842a6dc9153b6ab

## Neighborhood

### Who exposes this

- [Chemical refineries](/Customers/Chemical_refineries) — exposes problem · Customers

### What it's used for

- [OSIsoft PI](/Products/OSIsoft_PI) — used for · Products
- [Aveva Predictive Analytics](/Products/Aveva_Predictive_Analytics) — used for · Products
- [GE Digital APM](/Products/GE_Digital_APM) — used for · Products
- [Honeywell Forge](/Products/Honeywell_Forge) — used for · Products
- [Aspen HYSYS](/Products/Aspen_HYSYS) — used for · Products

### Competitors

- [GE Digital APM](/Competitors/GE_Digital_APM) — competes with · Competitors
- [Honeywell Forge](/Competitors/Honeywell_Forge) — competes with · Competitors
- [OSIsoft PI System](/Competitors/OSIsoft_PI_System) — competes with · Competitors
- [Aspen HYSYS](/Competitors/Aspen_HYSYS) — competes with · Competitors
- [Aveva Predictive Analytics](/Competitors/Aveva_Predictive_Analytics) — competes with · Competitors

### Entails child problem

- [Metallurgical Health Tracking](/Problems/Metallurgical_Health_Tracking) — entails child problem · Problems
- [Thermal Stress Mitigation](/Problems/Thermal_Stress_Mitigation) — entails child problem · Problems
- [Catalyst Degradation Monitoring](/Problems/Catalyst_Degradation_Monitoring) — entails child problem · Problems
- [Decoking Schedule Optimization](/Problems/Decoking_Schedule_Optimization) — entails child problem · Problems
- [Feedstock Variability Management](/Problems/Feedstock_Variability_Management) — entails child problem · Problems
- [Localized Coke Deposition](/Problems/Localized_Coke_Deposition) — entails child problem · Problems

### Solves problem

- [Detonationridge](/Startups/Detonationridge) — candidate solution for · Startups
- [Impellerhaven](/Startups/Impellerhaven) — candidate solution for · Startups
- [Lucen](/Startups/Lucen) — candidate solution for · Startups
- [Monitorharbor](/Startups/Monitorharbor) — candidate solution for · Startups
- [Plenumgate](/Startups/Plenumgate) — candidate solution for · Startups
- [Agack](/Startups/Agack) — candidate solution for · Startups

### Similar Problems

- [Prevent Unplanned Unit Outages](/Problems/Prevent_Unplanned_Unit_Outages) — similar · Problems
- [Catalyst Degradation Prediction](/Problems/Catalyst_Degradation_Prediction) — similar · Problems
- [Unplanned Unit Downtime](/Problems/Unplanned_Unit_Downtime) — similar · Problems
- [Facility Turnaround Cost Overruns](/Problems/Facility_Turnaround_Cost_Overruns) — similar · Problems
- [Distillation Yield Sub-Optimization](/Problems/Distillation_Yield_Sub-Optimization) — similar · Problems
- [Unplanned Furnace Downtime](/Industries/Steel_Mills/Problems/Unplanned_Furnace_Downtime) — similar · Problems
- [Equipment Downtime Costs](/Problems/Equipment_Downtime_Costs) — similar · Problems
- [Reduce Unplanned Reactor Downtime](/Problems/Reduce_Unplanned_Reactor_Downtime) — similar · Problems
- [Unplanned Furnace Downtime](/Industries/Iron_and_Steel_Mills_and_Ferroalloy_Manufacturing/Problems/Unplanned_Furnace_Downtime) — similar · Problems
- [Unplanned Process Downtime](/Problems/Unplanned_Process_Downtime) — similar · Problems
- [Maintenance Turnaround Cost Overruns](/Problems/Maintenance_Turnaround_Cost_Overruns) — similar · Problems
- [Track Flare Emissions Compliance](/Problems/Track_Flare_Emissions_Compliance) — similar · Problems
- [Unplanned Control Loop Failures](/Problems/Unplanned_Control_Loop_Failures) — similar · Problems
- [Minimize Unplanned Machine Downtime](/Problems/Minimize_Unplanned_Machine_Downtime) — similar · Problems
- [Corrosive Asset Degradation](/Problems/Corrosive_Asset_Degradation) — similar · Problems
- [Unplanned Equipment Downtime](/Problems/Unplanned_Equipment_Downtime) — similar · Problems
- [Crusher Plant Unplanned Downtime](/Problems/Crusher_Plant_Unplanned_Downtime) — similar · Problems
- [Prevent Chemical Batch Spoilage](/Problems/Prevent_Chemical_Batch_Spoilage) — similar · Problems
- [Minimize Production Line Downtime](/Problems/Minimize_Production_Line_Downtime) — similar · Problems