Intermittent Fault Diagnosis and Test Planning

The hardest jobs in a shop are the ones that will not fail while you are watching. The customer says the car stalled twice this month. The code is history. Nothing is wrong today. A beginner clears the code and hopes. A weak technician replaces a part and hopes. A strong technician writes a plan, captures the fault or forces it to appear, and only then sells a repair.

A good plan matters on any hard job, not just intermittent ones. It keeps you from wandering. It lets you tell the advisor how long you need and what it will cost. And it means another technician could pick up the job and carry on.

What You Will Be Able to Do

• Get the facts from a customer that actually help you recreate a fault.
• Write a test plan with expected results, and choose tests that cut the problem in half.
• Recreate heat, cold, moisture, vibration and load faults, or record them when you cannot.
• Test terminals and connections the right way.
• Decide when you have proved a cause, and when to stop.

Start With the Customer

Vague complaints make vague diagnoses. Ask questions that give you conditions you can repeat.

Ride with the customer if you can. Ten minutes in the passenger seat can save hours.

Build the Test Plan

A test plan is a short written list that you make before you start testing. It does not need to be fancy. It does need five things.

1. The complaint in one sentence, with its conditions.
2. What must be true for this symptom to happen. For a stall, the engine has lost fuel, spark, air or computer control.
3. Possible causes, ranked. Put the likely and the cheap to test at the top.
4. The test for each one, with the result you expect before you run it.
5. A time limit, and what you will do when you reach it.

Writing the expected result first is what makes this work. If you do not say in advance what good looks like, then you will talk yourself into whatever you see.

Choose Tests That Split the Problem

The best test is one that rules out half of the possible causes, whichever way the result goes. Think of a fuel pump circuit with a fuse, relay, wiring, connector, pump and ground. Do not test them one at a time from the fuse. Test in the middle.

Worked example: A pump will not run. Six items could be at fault. Measure voltage at the pump connector while the pump is commanded on. If voltage is present, the fuse, relay and power wiring are fine, and you have cut the list in half with one test. If it is missing, the pump and its ground are not the first problem. A test in the middle takes two or three steps to find the fault, while going from one end can take six.

Rank Tests by Value

A test's value depends on how likely it is to find the fault, how much it tells you, and how long it takes. A five-minute test that clears a whole system beats an hour on a rare cause.

Worked Example: A Written Plan

Complaint: Engine stalls when hot, at idle, about once a week. It restarts after 5 to 10 minutes. No current codes. A history code P0335, crank position sensor circuit.

Time limit: 90 minutes before calling the advisor with results and a next step. The plan says up front how long the job will take, so nobody is surprised, and the advisor can get approval.

Recreating the Fault

Most intermittent faults need a trigger. You can usually supply it, safely.

Keep the wiggle test gentle, and work in sections, so you know exactly where it failed. Watch a live signal while you do it.

Warning: Do not spray water near an open fuse box, a hot exhaust or high-voltage parts. Keep heat guns away from fuel lines, plastic and wiring insulation. Heat only as much as the part would see in service, or you may create a new fault.

Capturing the Fault

If you cannot make it fail, record it until it does.

• Scan tool recording. Many tools can record live data and save it when a code sets or when you press a button. Choose only the values that matter, so it records quickly.
• Freeze frame. Read it for the conditions. Even a history code often has one.
• Scope recording. A long recording with a trigger, or a long stored capture, can catch a dropout that scan data will miss. Use several channels to see which signal failed first.
• Vehicle data recorder. A small logger can stay on the car while the customer drives. Check your shop's policy, and mount it so it cannot interfere with the pedals or the driver.

The order of events often solves the case. If the crank signal dropped before the engine stalled, the signal caused it. If the engine stalled first and then the signal stopped, the signal was only reporting the stall.

Testing Connections the Right Way

Most intermittent faults are electrical connections. A meter on a resting circuit often says that they are fine, because one strand of wire can pass a continuity test and still fail under load.

Voltage Drop Under Load

Measure across a connection, wire or ground while the circuit is working. A healthy connection drops very little. A common guide is no more than about 0.1 volt for each connection, with a little more allowed for a full cable or circuit, but limits vary. Use the spec if there is one.

Worked example: A blower motor cuts out on rough roads. With the blower on high, you read 0.04 volt across the power connector and 0.9 volt across the ground. That ground is your fault. Clean or repair it, and retest. A resistance reading with the power off would likely have passed.

Pin Drag Testing

A terminal can look fine and still have lost its grip. To test it, push the correct size of test pin into each female terminal, and feel for a light, even drag. Compare with its neighbors. A pin that slides in with almost no resistance has lost its tension and can drop out over a bump. Use the matching terminal test tool from a repair kit. Never force a meter probe into a terminal, because that spreads it, and you create the fault you are looking for.

Also check that:

• Each terminal is fully locked in its connector. Pull gently on the wire.
• There is no green corrosion, water or burned plastic.
• The wire is not broken inside its insulation close to the terminal.

Common Hiding Places

• Grounds at the engine, the body and behind the kick panels
• Harness flex points, such as doors, the trunk hinge, engine mounts and steering columns
• Add-on devices like remote starters, dash cameras, trackers and OBD plug-ins
• Battery cables with corrosion under the insulation
• Water leaks dripping into a connector or a module
• Rodent damage in hidden parts of the harness

Is It Proved?

A cause is proved when you can make the fault come and go. Moving a wire makes the signal drop out, and letting go makes it return. Heating a sensor kills the signal, and cooling it brings the signal back. If you only have a theory, say so. Say what you found, what you did, and what it means.

Selling a part in the hope it fixes an intermittent fault costs the customer money and the shop its reputation. An honest "we could not reproduce it, here is what we tested" is a better answer.

Verify every repair under the conditions that caused the fault, not just once around the block. Write down what was tested, what was found and what was fixed, because a good record makes the next visit shorter.

Common Mistakes

Check Your Understanding

1. A customer says the car stalled twice last month. What do you ask? When it started, whether it was hot or cold, what the car was doing, what lamps came on, what changed, and how it restarted.
2. A misfire shows up only on wet mornings. How could you recreate it in the bay? Use a light mist of water on the ignition parts and connectors while you watch a live signal. Keep water away from an open fuse box and a hot exhaust.
3. What five things does a test plan need? The complaint with its conditions, what must be true for the symptom to happen, ranked causes, each test with its expected result, and a time limit.
4. You found a corroded connector, but you cannot link it to the stall. What do you do? Fix it, tell the customer it may not be the cause, and arrange follow-up.
5. After your time limit, you have found nothing. A coworker says to replace the fuel pump, as it could be that. What do you do? Report what you tested, and offer a recorder or more time. Do not sell a guess.

Key Takeaways

• Get conditions from the customer, not just symptoms.
• Write a plan with expected results and a time limit before you test.
• Test in the middle to cut the problem in half.
• Supply the trigger, or record the fault and read the order of events.
• Test connections with voltage drop under load and a pin drag test.
• A cause is proved when you can make the fault come and go.