Recognizing hazards in a plant

You cannot control a hazard you have not seen. In a plant, hazards hide in plain sight: a guard that has been off so long that nobody notices, a hose stretched across a walkway, a pallet stacked a little too high, a hissing steam trap, a smell that has always been there. HSE-103 covers hazard recognition and risk assessment in general. This lesson applies those ideas to plants, using a method built around energy that works well for industrial equipment and processes.

Hazards are energy

Almost every serious plant injury involves an uncontrolled release or transfer of energy into the body. Thinking in terms of energy helps you find hazards that a checklist might miss, because it makes you ask what energy is present, where it is stored, and what could release it. The energy wheel below groups plant hazards into ten forms of energy. Walking an area and asking about each form in turn is a quick and thorough way to see what is there.

High-energy hazards and direct controls

Research on serious injuries in industry has found that the amount of energy present is one of the best predictors of whether an incident could be fatal. Some studies use rough thresholds, such as a fall of more than about 1.2 m (4 ft), a suspended load over about 225 kg (500 lb), a vehicle moving faster than about 50 km/h (30 mph), or electrical energy above 50 V, to flag high-energy hazards. The exact numbers matter less than the habit: when you see high energy, ask whether there is a direct control that will work even if someone makes a mistake, such as a guard, an interlock, a barrier or isolation. Training, signs and PPE are useful, but they depend on people doing the right thing every time.

# Rough energy estimates for common plant hazards (illustrative values).
g = 9.81
hazards = {
    "200 kg die dropped 1.5 m":            200 * g * 1.5,
    "25 kg box falling 3 m from a rack":   25 * g * 3,
    "4,500 kg forklift at 10 km/h":        0.5 * 4500 * (10 / 3.6) ** 2,
    "80 kg worker falling 4 m":            80 * g * 4,
}
for name, joules in sorted(hazards.items(), key=lambda kv: -kv[1]):
    flag = "HIGH" if joules > 1500 else "lower"
    print(f"{name:<36} {joules / 1000:6.2f} kJ  {flag}")

4,500 kg forklift at 10 km/h          17.36 kJ  HIGH
80 kg worker falling 4 m               3.14 kJ  HIGH
200 kg die dropped 1.5 m               2.94 kJ  HIGH
25 kg box falling 3 m from a rack      0.74 kJ  lower

Walking the plant

Formal inspections are covered in lesson 12 and HSE-407, but hazard recognition happens every time someone walks the floor. A few habits make it more effective.

Normalization of deviance

One of the strongest forces working against hazard recognition is normalization of deviance. A small departure from a rule or design, such as running a machine with a guard removed for a quick adjustment, works without incident. It happens again, and again. Over time the deviation becomes the normal way of working, and nobody sees it as a hazard any more. The term comes from Diane Vaughan's study of the 1986 Challenger shuttle loss, but the same pattern appears in plants everywhere. Signs of it include phrases like 'we always do it this way', tools left hanging on machines for bypassing interlocks, and taped-over sensors.

Reporting hazards and near misses

Hazard recognition is useful only if what people see is reported and acted on. Plants make reporting easy with simple cards, phone apps, boards in each area and a standing agenda item at shift meetings. The most important factor, though, is what happens after a report. If workers see their reports fixed, or get an explanation when a fix will take time, they keep reporting. If reports disappear, reporting stops.

# Hazard report follow-up: how fast are reports closed? (illustrative data, days to close)
import statistics
reports = {"guarding": [2, 5, 1, 9, 3], "housekeeping": [1, 1, 2, 1], "traffic": [14, 30, 7],
           "electrical": [3, 4], "ergonomics": [21, 45, 12, 60]}
for cat, days in reports.items():
    over = sum(d > 14 for d in days)
    print(f"{cat:<13} n={len(days)}  median {statistics.median(days):>4} days  over 14 days: {over}")
allds = [d for v in reports.values() for d in v]
print(f"All reports: median {statistics.median(allds)} days, {sum(d > 14 for d in allds)} of {len(allds)} took over two weeks")

guarding      n=5  median    3 days  over 14 days: 0
housekeeping  n=4  median  1.0 days  over 14 days: 0
traffic       n=3  median   14 days  over 14 days: 1
electrical    n=2  median  3.5 days  over 14 days: 0
ergonomics    n=4  median 33.0 days  over 14 days: 3
All reports: median 4.5 days, 4 of 18 took over two weeks

From hazard to risk

Recognizing a hazard is the first step. The next is judging the risk, which depends on how likely harm is and how severe it would be. HSE-103 and HSE-223 cover risk matrices and other methods in depth. In plants, a simple matrix helps rank the hazards found during a walk so that the most serious get attention first.

# A 5 x 5 risk matrix applied to hazards found on a plant walk (illustrative ratings).
def band(s):
    return "HIGH" if s >= 15 else "MEDIUM" if s >= 8 else "LOW"
found = [  # (hazard, likelihood 1-5, severity 1-5)
    ("pedestrians in forklift aisle", 4, 5),
    ("taped light curtain on palletizer", 3, 5),
    ("hose across walkway", 4, 2),
    ("missing rack end protector", 2, 4),
    ("spray bottle with no label", 3, 2),
]
for h, l, s in sorted(found, key=lambda x: -(x[1] * x[2])):
    print(f"{h:<36} L{l} x S{s} = {l * s:>2}  {band(l * s)}")

pedestrians in forklift aisle        L4 x S5 = 20  HIGH
taped light curtain on palletizer    L3 x S5 = 15  HIGH
hose across walkway                  L4 x S2 =  8  MEDIUM
missing rack end protector           L2 x S4 =  8  MEDIUM
spray bottle with no label           L3 x S2 =  6  LOW

Hazard recognition for contractors and visitors

People new to a plant see some hazards more clearly than regulars, because nothing is normal to them yet, but they also miss hazards that are specific to the process, such as a line that carries ammonia or a hopper that can contain combustible dust. A good host shows contractors the plant-specific hazards in their work area before they start, and encourages them to report anything that worries them. Some plants give contractors a quick hazard card listing the energies present in each area.

Job hazard analysis in plants

For tasks that are done regularly, plants record hazards and controls in a job hazard analysis (JHA), sometimes called a job safety analysis. The task is broken into steps, the hazards of each step are listed using the energy wheel as a prompt, and the controls are recorded. JHAs are most useful when they are written with the people who do the job, kept short, posted where the work is done, and reviewed when the task, equipment or materials change. For non-routine tasks such as unusual repairs, many plants use a quick field-level hazard assessment at the start of the job instead, which HSE-230 covers.