"Dr. Cole," she said, "the chairman's authorization is in the system. He's cleared to observe and provide nonbinding diagnostic suggestions."
"Nonbinding," Cole repeated.
"That's what the document says."
Cole folded his arms.
"Fine. He observes. He does not touch the console, he does not interrupt active testing, and he does not give instructions to my team."
Jamal nodded.
"That's reasonable."
The lack of resistance seemed to irritate Cole more than an argument would have.
"Good. Then we'll proceed."
Jamal moved to the designated observation area. A technician brought up the previous test logs on a large display.
The data appeared as a series of colored lines, each representing a different system variable over time.
For the first few minutes, everything looked normal.
The engine warmed gradually. Rotational speed increased according to the programmed schedule. The AI controller made small adjustments to maintain the desired operating range.
At minute twelve, the temperature curve began to rise more quickly.
At minute thirteen, the vibration readings changed.
At fourteen minutes and thirty-seven seconds, the warning appeared.
HARMONIC DISRUPTION DETECTED.
The system entered protective shutdown.
The room fell silent.
Jamal watched the screen.
"Can you display the raw sensor values?" he asked.
Cole turned toward him.
"Which values?"
"Temperature, vibration, and the feedback channel associated with the replacement sensor. Raw values, before the controller filters or normalizes them."
"We already analyzed the data."
"I'm sure you did. I'd like to see the time series."
Cole hesitated before giving the technician a curt nod.
The display changed.
Jamal moved closer, keeping his hands behind his back.
He studied the curves for several minutes.
Then he pointed toward a narrow section of the temperature trace.
"Can you zoom in here?"
The technician complied.
The line appeared almost smooth, but tiny irregularities became visible near the beginning of the failure sequence.
"What's that?" Priya asked.
Jamal leaned closer.
"I don't know yet. It could be measurement noise. But the pattern begins before the system reports a harmonic disruption."
Cole shook his head.
"That channel has been filtered."
"Then I would like to compare it with the unfiltered record."
"The unfiltered data is noisy."
"That is why it should be examined alongside the filtered data, not instead of it."
Priya turned toward Cole.
"He's right about the comparison."
Cole's jaw tightened.
The technician opened the unfiltered record.
The small irregularities became more pronounced. They appeared in a repeating pattern, increasing in amplitude as the test progressed.
Jamal did not smile.
He did not announce a discovery.
He simply asked for the sensor's identification number.
Priya retrieved the component record.
The original sensor had been manufactured by a supplier called Meridian Instruments. The replacement came from a different production line, though the catalog description listed it as an approved equivalent.
The paperwork showed that the replacement had passed a standard electrical continuity test.
It also showed that the installation team had recorded a successful low-load verification.
But there was no evidence of a complete dynamic calibration under the prototype's maximum operating conditions.
"Why wasn't the full calibration performed?" Jamal asked.
Cole answered immediately.
"Because the sensor was certified as a compatible replacement."
"By whom?"
"The procurement engineering group."
"Did they test its response curve under full load?"
"I don't have that information in front of me."
"Could we retrieve it?"
Cole looked at him sharply.
"You're asking a lot of questions for someone who hasn't been here ten minutes."
"I'm trying to understand the record."
"Then understand this: the engine has been examined by specialists with decades of combined experience. We have already checked the sensor."
Jamal nodded.
"Then we should be able to rule it out with the calibration data."
The room became very still.
Priya lowered her eyes toward the console, but Jamal noticed the corner of her mouth twitch.
Cole turned back to the screen.
"Run the standard test again."
The technician initiated the sequence.
The engine started with a low mechanical hum.
The room filled with the familiar sounds of fans, pumps, and monitoring equipment. Numbers shifted across the displays.
At twelve minutes, the temperature curve began to climb.
At thirteen, the vibration trace changed.
At fourteen minutes and thirty-seven seconds, the warning appeared.
The engine shut down.
Jamal looked at the clock.
"Exactly the same interval?"
"Within a fraction of a second," the technician replied.
Jamal turned toward the diagnostic screen.
"Can we overlay the last three tests?"
The technician did so.
The three failure traces aligned almost perfectly.
Priya stepped closer.
"That's unusually consistent."
"Which means we should examine what changes predictably before the failure," Jamal said. "Not just what the system reports when it stops."
Cole rubbed his forehead.
"We've done that."
"Then perhaps the answer is hidden in the data you haven't compared."
The chief engineer stared at him.
For a moment, Jamal wondered whether he had gone too far.
Then Priya spoke.
"Let's compare the sensor response with the controller's expected response curve."
Cole looked at her.
"That will take an hour."
"Then it will take an hour."
The technician loaded the reference data.
The new graph showed the expected sensor response as a thin line and the actual response as a second line. At first, the lines were nearly indistinguishable.
As the engine approached operating temperature, they began to separate.
The difference was small.
But it was consistent.
At minute twelve, the discrepancy became measurable.
At minute thirteen, it widened.
The control system responded by making increasingly aggressive adjustments. Each correction introduced another small change into the system.
The engine was not failing because the sensor had stopped working.
It was failing because the sensor was reporting a value that appeared plausible but did not match the response the controller had been designed to expect.
The system was correcting a problem that did not exist in the way it believed it did.
Jamal stared at the display.
"There," he said quietly.
Priya moved closer.
"The sensor's dynamic response is lagging behind the reference curve."
"Yes. The controller sees a delayed change and compensates. The compensation increases the oscillation instead of reducing it."
Cole looked at the data, then at the replacement component's record.
His face had lost some of its color.