Use evidence capture and controlled observation when the problem refuses to fail on command.
What to understand
Intermittent faults require patience. Your goal is to capture the state at the moment of failure rather than disturb the system until the evidence disappears.
Start by turning the idea into an observable question. What should the machine, component, person, or process be doing at this moment, and what evidence would prove that it is doing it? Defining normal first gives your troubleshooting a reference point and prevents you from reacting to the first unusual thing you notice.
How to apply it
Identify variables that change over time: temperature, vibration, cable movement, product position, cycle count, load, network traffic, moisture, and operator interaction.
Apply the idea one boundary at a time. Confirm one condition, record what you learned, and only then move to the next point in the chain. This makes your work easier to explain and greatly reduces the temptation to swap parts or change several variables at once.
What good work looks like
Use trends, alarm history, counters, safe observation, and temporary diagnostic logging when approved. Avoid permanent bypasses created merely to keep the machine running.
Good maintenance work leaves a trail of evidence. Measurements, alarm times, verified states, photos where permitted, and clear work-order notes let another technician understand why you made the repair and whether the same failure mechanism returns later.
Field scenario
A machine stops for one second every few hours with no persistent alarm. A trend shows the 24 V control supply dipping whenever a large solenoid bank energizes. Inspection finds a loose supply terminal that heats and becomes more resistive during operation.
After the immediate repair, capture the point in the sequence where the expected condition was lost and what evidence proved the cause. That small discipline turns a one-time fix into knowledge the whole maintenance team can reuse.
Field checklist
Common mistakes to avoid
Replacing multiple parts hoping the fault disappears
Clearing logs before reading them
Leaving covers open or devices bypassed while waiting for a fault
Calling intermittent failures impossible to diagnose
Ask yourself: If this machine failed again on the next shift, what information could I leave behind that would make the next diagnosis faster and safer?
90-day action
Before moving on, choose one task from this chapter and perform it during normal work. Record what you learned in your plant notebook. The value of the first 90 days comes from turning concepts into repeatable behavior, not from reading alone.
CHAPTER 23