RISK MATRIX
Figure 1. Effectively managing near-misses requires putting them into appropriate context.
THE BIG PICTUREToday, operations comprise a mix of automatic and manual or procedural steps. A plant may have hundreds of control loops, interlocks, permissives and safety systems. In addition, numerous procedures rely on human interaction — for instance, opening or closing a block valve, and starting, stopping or ramping a pump or compressor. In view of the large number of operations, it appears reasonable to anticipate that "something" will go wrong someday.Many factors can contribute to an incident, including:• inadequate training and poorly written procedures;• process complexity;• personnel issues;• age of the plant and equipment;• environmental problems such as severe weather; and• random events (i.e., ones we can't really predict).From a managerial perspective, the key is to identify near-miss incidents that under slightly different circumstances could become major events with adverse consequences. Unfortunately, the experience-based approach likely won't spot all such worrisome incidents. That's not surprising because staff may lack sufficient breadth and depth of experience to truly determine which incidents could become big. In addition, in some cases, there's a tendency to disregard existing data on equipment or procedural failures.
THE RM APPROACHAs an alternative, plants should consider using an RM technique. Such a semi-quantitative method can better identify near-misses that could have led to dire consequences under a slightly different set of circumstances. Basically, it screens incidents by likelihood and severity to pick out those with a potential to develop into "large impact" events.The approach involves estimating the likelihood and consequence of the near-miss and assigning these an index value to use for the risk matrix.
Figure 1 shows such a risk matrix while Table 1 provides guidance on how to assign appropriate indices.Plants with a long history of operations can use near-miss records to estimate likelihood and consequence of relatively similar incidents. New plants should consider data from similar incidents elsewhere in the chemical industry.In the absence of data, you can turn to plant operations and maintenance personnel for estimates. However, in the long-run, the goal should be to collect relevant data.The accident data must be arranged in some useful classes. For instance, you might consider the following categories: flammable or toxic gas leak, flammable or toxic liquid spill, dropped object, procedural error, design deficiency, and other. From the data on near-misses you've collected, estimate the frequency (e.g., times/year) for each category and use these estimates to come up with a likelihood value for the matrix, with 3 being highest likelihood.Then categorize consequence in three classes: relatively low level of impact (1 on the matrix), moderate level (2), and high level (3). Be as specific as possible — e.g.:
• Category 1: no injury, no lost time, <100-lb gas release, <3-gal spill, no fire, <$1,000 loss;
• Category 2: one injury, lost time; 100–500-lb gas release, 3–5-gal spill, no fire, $1,000–$2,000 loss; and
• Category 3: one lost-time injury, 500–1,000-lb gas release, 5–10-gal spill, no fire, >$2,000 loss.
Once you've placed an incident in the matrix, you're in a position to decide whether it demands corrective action based on your plant's risk acceptance criteria.
