4.2 Data Completeness, Quality Assessment, and Handling Data Gaps

Data Quality vs. Data Quantity in Risk-Based Inspection

A fundamental principle established in API RP 580 Section 8 is that high data quality is significantly more critical than sheer data quantity. An RBI algorithm supplied with large volumes of unverified, inaccurate, or outdated data will produce misleading risk rankings—a scenario often summarized in engineering as "garbage in, garbage out."

Data uncertainty directly amplifies calculated risk. Quantitative RBI models (such as API RP 581) are mathematically structured to penalize data uncertainty. When key parameters—such as material specification, operating temperature, process stream chemistry, or inspection history—are unknown, the risk calculation engine must substitute conservative default assumptions. These conservative assumptions artificially elevate calculated Damage Factors ($D_{\text{total}}$) and Consequence Areas ($C_{\text{area}}$), driving equipment into higher risk categories on the plant risk matrix.

---

Data Quality and Source Confidence Hierarchy

To ensure consistency across multi-disciplinary assessment teams, API RP 580 recommends establishing a formal Data Confidence Hierarchy. Input data is assigned confidence levels based on the verification rigor of its source records:

┌────────────────────────────────────────────────────────────────────────┐
│                     DATA SOURCE CONFIDENCE HIERARCHY                   │
├─────────────────┬──────────────────────────────────────────────────────┤
│ HIGH CONFIDENCE │ • Form U-1 Manufacturer's Data Reports (MDR) & MTRs  │
│                 │ • 100% Automated UT (AUT) thickness grid scans       │
│                 │ • Certified lab chemical analysis of stream samples  │
│                 │ • Continuous process historian continuous averages    │
├─────────────────┼──────────────────────────────────────────────────────┤
│ MEDIUM CONFID.  │ • Standard P&IDs, PFDs, and isometric drawings       │
│                 │ • Manual spot Ultrasonic Thickness (UT) measurements │
│                 │ • Design heat and material balances (HMB)            │
├─────────────────┼──────────────────────────────────────────────────────┤
│ LOW CONFIDENCE  │ • Unverified sketch drawings or assumed specifications│
│                 │ • Missing inspection records (Category E assumption) │
│                 │ • Operator verbal estimates of fluid temperature     │
└─────────────────┴──────────────────────────────────────────────────────┘

1. High Confidence (Low Uncertainty): Data verified by authoritative engineering documents, including ASME Form U-1 Manufacturer's Data Reports (MDRs), certified Material Test Reports (MTRs), continuous process historian (DCS) data logged over 12+ months, and 100% Automated Ultrasonic Testing (AUT) grid mapping.
2. Medium Confidence (Moderate Uncertainty): Data derived from standard operating P&IDs, design heat and material balances (HMB), periodic manual spot UT thickness readings, or routine unit operating logs.
3. Low Confidence (High Uncertainty): Data estimated from unverified site sketches, verbal operator estimates, missing inspection histories, or generic industry fluid property handbooks.

---

Identifying Common Data Gaps in Industrial Facilities

During initial RBI implementation across older refining or chemical manufacturing facilities, assessment teams frequently encounter significant data gaps. Common data gaps and their technical impacts include:

• Missing Manufacturer's Data Reports (MDRs) or MTRs: Leaves material yield strength, PWHT status, and nominal thickness unconfirmed. Impact: The model must assume non-PWHT status (elevating environmental cracking susceptibility) and lower SMYS baseline values.
• Unmonitored Small-Bore Piping (NPS 2 and Smaller): Small-bore bypass lines, drains, and vents are historically omitted from routine inspection programs. Impact: Unknown wall loss rates force conservative POF estimates, despite low fluid inventory.
• Unknown Insulation Installation Date and Condition: Lacking records on insulation age, jacketing integrity, or moisture barrier condition. Impact: Forces the RBI model to assume severe Corrosion Under Insulation (CUI) driving high damage factors ($D_{\text{cui}}$).
• Unanalyzed Trace Process Contaminants: Lacking laboratory stream analyses for chlorides, cyanides, or ammonium bisulfide wt%. Impact: Missing chemistry data prevents accurate identification of API RP 571 active damage mechanisms.

---

Managing Data Gaps: Conservative Assumptions vs. Data Gathering

API RP 580 Section 8 presents a clear methodology for resolving data gaps. The assessment team faces two primary pathways:

                             ┌───────────────────────────┐
                             │   DATA GAP IDENTIFIED     │
                             └─────────────┬─────────────┘
                                           │
                   ┌───────────────────────┴───────────────────────┐
                   ▼                                               ▼
┌──────────────────────────────────────┐       ┌──────────────────────────────────────┐
│ PATHWAY 1: CONSERVATIVE ASSUMPTIONS  │       │ PATHWAY 2: TARGETED DATA GATHERING   │
├──────────────────────────────────────┤       ├──────────────────────────────────────┤
│ • Assign worst-case corrosion rate   │       │ • Perform field AUT grid mapping     │
│ • Assume Category E (Ineffective) NDE│       │ • Perform alloy PMI (API RP 578)     │
│ • Assume non-PWHT carbon steel       │       │ • Sample process fluid for chlorides │
│ • High calculated POF / High Risk    │       │ • Low data uncertainty / True Risk   │
└──────────────────────────────────────┘       └──────────────────────────────────────┘

Economic Cost-Benefit Principle

While applying conservative default assumptions allows an RBI study to proceed without delay, over-reliance on conservative defaults creates an artificial "high-risk" burden. Facilities risk spending hundreds of thousands of dollars on premature equipment replacement or unnecessary turnaround shutdowns to address phantom risks created entirely by data uncertainty.

API RP 580 advocates performing a Cost-Benefit Evaluation: compare the cost of gathering accurate field data (e.g., executing a $3,500 targeted NDE scan or a $1,500 lab fluid sample) against the cost penalty of accepting conservative default risk assumptions (e.g., a $75,000 capital bundle replacement scheduled based on high uncertainty).

---

Principles and Execution of Sensitivity Analysis

Governed by API RP 580 Section 8, Sensitivity Analysis is a powerful analytical technique used to evaluate how variations in uncertain input parameters influence output POF, COF, and risk matrix placement.

Sensitivity Analysis Steps:

1. Identify High-Uncertainty Inputs: Pinpoint parameters with low data confidence (e.g., estimated corrosion rate, operating temperature $\pm 30^\circ\text{F}$, chloride concentration $10 \text{ ppm}$ vs $250 \text{ ppm}$, or isolable inventory mass).
2. Define Parameter Range: Establish realistic lower-bound, expected baseline, and upper-bound conservative values for the uncertain variable.
3. Execute Iterative Risk Calculations: Run the RBI model across the parameter range while holding all other inputs constant.
4. Evaluate Risk Matrix Stability: Determine whether parameter variation shifts the asset across risk matrix threshold boundaries (e.g., moving from Medium-High Risk to High Unacceptable Risk).
5. Prioritize Targeted Data Collection: If parameter sensitivity is high (causing a shift into the unacceptable risk zone), prioritize immediate field data collection to measure that specific parameter accurately.

---

Data Quality Management Summary

---

Technical Worked Example: Sensitivity Analysis for an Overhead Piping Circuit

Scenario

An Atmospheric Crude Distillation Column Overhead Vapor Line (16" Carbon Steel, SA-106 Gr B) operates at $240^\circ\text{F}$.

• Problem: Historical inspection records are missing. The corrosion specialist suspects Hydrochloric Acid (HCl) corrosion due to potential desalter wash water carryover, but aqueous chloride concentration is unmeasured.
• Consequence Area (COF): Flammable release area $\text{COF} = 15,000 \text{ ft}^2$.
• Acceptable Risk Threshold: Target risk limit = $600 \text{ ft}^2/\text{year}$.

Step 1: Baseline Assessment under Conservative Default Assumption

Lacking chloride data, the model assumes maximum severe wet HCl corrosion rate ($CR_{\text{default}} = 25.0 \text{ mpy}$) and Category E (Ineffective) inspection confidence.

• Calculated Damage Factor: $D_{\text{thin}} = 80.0$.
• Baseline POF: $\text{POF}_{\text{base}} = 8.0 \times 10^{-2} \text{ failures/year}$.
• Baseline Risk: $\text{Risk}_{\text{base}} = (8.0 \times 10^{-2}) \times 15,000 \text{ ft}^2 = 1,200 \text{ ft}^2/\text{year}$.
• Result: Exceeds acceptable threshold ($1,200 > 600 \text{ ft}^2/\text{yr}$). Asset is categorized as High Unacceptable Risk, triggering a recommendation for immediate $110,000 piping replacement.

Step 2: Sensitivity Analysis Evaluation

The engineer runs a sensitivity trial varying corrosion rate ($CR$) across three scenarios:

• Scenario A (Low Corrosion Rate - $4.0 \text{ mpy}$): Assumes effective overhead boot water wash and low chlorides. $\text{POF}_A = 1.28 \times 10^{-2} \implies \text{Risk}_A = 0.0128 \times 15,000 = 192 \text{ ft}^2/\text{year} \quad (\text{Acceptable})$
• Scenario B (Expected Corrosion Rate - $8.0 \text{ mpy}$): Assumes moderate chloride carryover. $\text{POF}_B = 2.56 \times 10^{-2} \implies \text{Risk}_B = 0.0256 \times 15,000 = 384 \text{ ft}^2/\text{year} \quad (\text{Acceptable})$
• Scenario C (Severe Corrosion Rate - $25.0 \text{ mpy}$): Default conservative baseline. $\text{POF}_C = 8.0 \times 10^{-2} \implies \text{Risk}_C = 0.0800 \times 15,000 = 1,200 \text{ ft}^2/\text{year} \quad (\text{Unacceptable})$

Step 3: Engineering Decision & Economic Justification

Sensitivity analysis proves that if actual corrosion rate is $\le 8.0 \text{ mpy}$, equipment risk is well within acceptable boundaries ($384 < 600 \text{ ft}^2/\text{yr}$).

• Decision: Rather than executing an unnecessary $110,000 piping replacement, the plant invests $4,500 to perform Automated UT (AUT) profile grid mapping and $1,200 for overhead boot water chloride sampling.
• Field Result: AUT measures actual wall thickness confirming historical corrosion rate is $5.2 \text{ mpy}$ (Category A Inspection Effectiveness achieved).
• Updated Model: Calculated Risk drops to $249.6 \text{ ft}^2/\text{year}$, saving the facility $110,000 in capital expenditure while ensuring absolute mechanical integrity compliance.