Skip to main content

Risk Assessment Techniques

Risk Assessment Techniques

Risk assessments are a valuable tool to systematically evaluate the potential for harm, damage, and effects on business operations from systems or components. As outlined below, risk assessment methods are qualitative, semi-quantitative or quantitative and differ in approach, data requirements, and depth of the analysis.  Appreciating the differences and limitations of a risk assessment tool and applying the method correctly is key to ensuring an accurate characterization of potential consequences and providing appropriate safeguards.

The table below illustrates these assessment types and the information required at various project phases.  This table also should be revisited through the project phases as more information is available and to continually evaluate system safety. It is common to conduct multiple risk assessments on the same component or system throughout the project life-cycle.

Table 1: Information Needed for Various Assessment Types
Phase Information or Conceptual System Design Detailed Event Scenarios
Assessment Type Qualitative Qualitative, semi-quantitative, or quantitative Quantitative
Information Needed Purpose of the system, major components, and high-level functional description P&IDs, plot plans, electrical design, controls details, component specifications, administrative controls, and operation and maintenance procedures
  • Leak or failure-rate data*
  • Experimental or operational data*
  • Validated consequence models* (e.g., dispersion, flame, and overpressure)
  • Harm criteria*

*Can be a significant source of error.

Approaches:

There are two fundamental approaches when performing qualitative and semi-quantitative analysis.

1. Top Down: Identify high level hazards/consequences, identify risks, evaluate detailed systems design

2. Bottom Up: Evaluation of detailed systems design/function, identify risks, assess hazards/consequence

The illustration below shows how risks can be characterized for qualitative and semi-quantitative risk analysis.  For more detail also see risk ranking.

Example Ranking Matrix

When experimental data such as failure rate, ignition probability and validated dispersion models are available, a more detailed quantitative assessment can help determine how best to lower the probability of occurrence and how to reduce consequences to help determine:

  • Specific starting events that will lead to system failure
  • Severity of consequences from system failure (operational, property damage, injuries/fatalities, etc.)
  • Frequency of consequences occurrences

 

For example, severity of consequences and the likelihood of occurrences resulting from a given activity in the work scope can be quantitatively categorized and expressed as the number of people injured or the dollar amount of damage done during an operational period. The challenge is the accuracy of quantitative models are highly dependent on the accuracy of the underlying data, such as leak frequency as a function of leak size and ignition probability.  Damage levels can also be highly variable.  For quantitative assessments the results should be repeatable, and identical given the same input values and a sensitivity analysis and/or potential error should be specified in recognition of model limitations.

Depending on the method used and especially for a qualitative and semi-quantitative risk assessment process, the following should be included as applicable:

  • Input from SMEs, workers and management
  • Prioritization or ranking of the consequence and likelihood for each failure mode
  • Suggested controls and layers of protection (engineering and/or administrative) for each hazard identified

 

References:

  • Hazard Analysis and Risk Assessment | H2tools | Hydrogen Tools
  • Center for Chemical Process Safety (CCPS). Guidelines for Chemical Process Quantitative Risk Analysis. American Institute of Chemical Engineers.
  • NFPA 2, Hydrogen Technologies Code, Chapter 5 Performance Based Option, 2023 Edition, National Fire Protection Association.
  • Sandia Report - B. Ehrhart, E. Hecht, and M. Devin. 2025 update “Hydrogen Plus Other Alternative Fuels Risk Assessment Models Version 6.0 Technical Reference Manual (HyRAM+ – Energy).” Sandia National Laboratories, SAND2025-04942.

 

Qualitative Assessments

Qualitative analysis is exploratory in nature involving a process of examining non-numerical data to evaluate risks using subjective judgment and experience. It frequently involves brainstorming realistic and conservative scenarios to identify potential risk and consequences depending on the objectives. Risks are often categorized as “low”, “medium” or “high” with the primary goal of establishing high likelihood and high-risk events to improve understanding of risk and drive towards appropriate mitigation strategies.

When used: New system concept review; identify primary risks to the system; drive major design decisions; evaluate proposed design changes at the concept level; form basis for further research on high-risk topics.

Challenges: Quality of results depends on the knowledge/skill of participants. Potential to miss consequences based on the availability of information. May not effectively evaluate a detailed design.

Methods:

  • Bow-Tie: This analysis is a visual risk assessment methodology used to map and communicate the pathways leading to a hazardous event and its potential consequences. The diagram gets its name from its distinctive shape, which resembles a bow tie, with a central "knot" representing a critical event. The analysis centers on a top event—the point at which control over a hazard is lost. To the left of the top event are threats (or initiating mechanisms) that could lead to the event. To the right are consequences that may follow if the event occurs. Connecting these elements are barriers (or controls), which are preventive measures on the left (to stop threats from causing the top event) and mitigative measures on the right (to reduce the impact of consequences). This method combines elements of fault tree analysis (causal logic on the left) and event tree analysis (consequence logic on the right), but in a simplified, qualitative format.
  • Fault Tree Analysis: Method for determining which initiating events lead to a system fault. When used semi-quantitatively, each initiating event is assigned a probability, allowing for a calculation of the system fault probability.
  • HAZID: Short for Hazard Identification, is a structured, qualitative risk assessment technique used to systematically identify potential hazards in a process, system, or facility—particularly during early design or operational phases. It involves a multidisciplinary team conducting a workshop to brainstorm and evaluate hazards based on guide-words such as, fire, explosion, chemical release and scenarios, focusing on both internal process risks and external threats such as natural disasters, security breaches, or neighboring facility incidents. The goal is to detect hazards early to enable cost-effective design changes, support regulatory compliance, and establish a foundation for more detailed analyses like HAZOP. HAZID is typically performed at the conceptual or FEED (Front-End Engineering Design) stage, before detailed engineering, and is used to identify major accident events, their causes, consequences, and existing safeguards.
  • What-If: This analysis is a strategic and analytical technique used to explore the potential outcomes of a model, system, or decision by changing the values of its input variables (assumptions). It answers the fundamental question: "What would happen if a specific condition or assumption changed?"

 

Semi-Quantitative Assessments

Semi-quantitative analysis is a systematic process that couples qualitative with quantitative analysis, predicts likelihood and severity of risks using alpha numeric values based on data and experience.

When used: Throughout the process, first step refinement in qualitative assessments; detailed design review; assessment of potential consequences, evaluation of proposed design changes, determination for when full quantitative analysis is needed.

Challenges: Blending the best of both worlds with data and good safety judgment.

Methods:

  • Failure Modes and Effects Analysis (FMEA): Systematic method for examining the effects of component failures on system performance and surroundings. FMEA determines which initiating events (component failures, operating conditions, external events, etc.) will lead to significant adverse consequences.
  • Hazard and Operability Analysis (HAZOP):  Method that focuses on the systematic identification of causes, consequences, and safeguards from defined deviation key words. Effective at identifying potential problems and issues within complex systems and reducing risk through recommendations.
  • Layer of Protection Analysis (LOPA): is a risk assessment technique used to evaluate the adequacy of protective measures against high-consequence hazard scenarios. It builds upon qualitative hazard identification methods like HAZOP by adding a structured, rule-based approach to estimate risk reduction. LOPA assesses the frequency of initiating events, the severity of potential consequences, and the effectiveness of Independent Protection Layers (IPLs)—such as safety instrumented systems, alarms, relief valves, or operator intervention—that prevent or mitigate an incident.

 

Quantitative Assessments

Quantitative analysis is a method that uses mathematical and statistical techniques and is reliant on data to analyze or model data to postulate potential risk and consequences. The analysis may focus on one or more elements such as number of fatalities per year, probability of outcomes, or single scenario consequences.

When used: Fully designed system to answer specific questions such as, What is the probability of X happening? What is the expected consequence? How does design impact probability of event? Is the likelihood/consequence acceptable?

Challenges: The quality or difficulty obtaining useful data, adjustments to frequencies, easy to be misleading. Unknown error or unspecified confidence levels in results.

Methods:        

  • Computational Fluid Dynamics (CFD): Is a model process using numerical methods and algorithms to simulate and analyze fluid flow, heat transfer, chemical reactions, and related phenomena by solving the governing equations of fluid mechanics on a computer. It enables a prediction of how fluids—such as gases, liquids, or multi-phase mixtures—behave in real-world scenarios without the need for physical experiments.
  • Consequence Modeling: Includes jet fire model, blast over-pressure model, dispersion model, etc. 
  • Monte Carlo Simulation: Assesses the probability of different outcomes by running numerous simulations with random inputs based on defined probability distributions.

 

Using Computer Modeling Software

When performing a quantitative risk assessment (QRA), it is often the case that specific data is not available leading to a wide range of hazard characterizations and assumptions impacting the results.  Standards requiring QRA and guidance documents may be ambiguous or vague about the type of risk assessment required.  When conducting a QRA using a computer model/software and developing assumptions, there may be the inclination to adjust model input values to obtain desired results to accommodate project schedule, budget and/or build location constraints. While all these factors are a consideration of any project, they should not be used to influence or compromise the integrity of a QRA. The best practices outlined below are recommended to promote safety and trust in the hydrogen industry and to encourage a pragmatic approach to avoid simply “checking the box” to satisfy code compliance or a desired predetermined outcome.  Ultimately, the challenge is to ensure a practical level of safety accounting for qualitative unknowns. Do not understate or overstate the hazard--understand the hazard.

1.     Selected scenarios should be credible with reasonably conservative assumptions applied. Credible means scenarios that could be expected to occur in the setting or operation, such as assuming a H2 leak will ignite.  An example of a non-credible event may be a storage container being hit by a meteor.  Reasonably conservative means selecting input values such as the “mean” from applicable test data if available, if the data establishes high, mean and low values for a particular release event or component failure. In cases where such data is not available or uncertainty is high, best practice is to use conservative input values.

2.     Scenarios should include the spectrum of high risk-low probability to low risk-high probability.  For example, part of this bounding range would include scenarios that consider 100% ignition probability of entire container contents and rapid release operational events to understand potential worst-case consequences if all safety features fail (high risk-low probability) and scenarios that consider ignition of small leaks with limited fuel source (low risk-high probability).  Scenarios should consider releases from operational, security and credible natural hazards. The key is to have a comprehensive appreciation of the potential consequences.

3.     All assumptions should be clearly documented to ensure limitations or applicability are transparent.

4.     All model input values should be clearly documented so the scenarios can be repeated/duplicated by other users and analysts, etc.

5.     Risk mitigation and reduction features used to reduce potential consequences should be clearly documented with a technical explanation of how the feature would apply and accomplish the intended reduction.  These are discussed in the next best practice section Risk mitigation.