Three Presses, Three Diagnoses - and a Failure No One Could Explain

By Ronen Dolev | 2026

Three Different Failures in the Same Plant

At an industrial printing plant in Europe, three digital presses using Liquid Electrophotography (LEP) had been producing consistently to customer specifications and established quality-control procedures. Then, over a few days, all three began showing unwanted print-quality disturbances.

On the first press, a faint tonal shadow appeared after long areas of heavy ink coverage. It was most apparent when the subsequent image contained light or midtone areas. The maintenance team suspected instability in image transfer and began investigating the printing system.

Two days later, a disturbance from the same visual family appeared on a second press, though its intensity and frequency differed. That press team believed the issue involved calibration or unstable tonal response.

Toward the end of the week, the third press produced a more visible disturbance, mainly on one particular substrate. Its operators concentrated on the material and the substrate-related working conditions.

Three presses. Three groups of experienced people. Three diagnoses that sounded entirely reasonable.

Throughout the week, the teams carried out inspections, calibrations, cleaning procedures, operating-point adjustments, and checks of printing components. Some interventions improved the result. None convincingly established why the disturbance started or why it returned.

The more changes the teams made, the harder it became to reconstruct the original conditions. Management began to wonder whether a common problem was affecting the plant. Yet three unrelated faults occurring in the same week remained a possibility.

After local troubleshooting failed to produce a defensible conclusion, an application expert from the press manufacturer arrived.

The First Question Was Not About a Press

Three investigation folders were waiting on the meeting-room table. Each contained print samples, measurement records, maintenance history, and recommendations from the team responsible for that press.

The expert looked through them and asked, "What makes us so certain these are three different failures?"

The maintenance engineer answered immediately. "The symptoms are not identical. Press one has a transfer problem. Press two has a calibration problem. Press three points to the substrate."

"Those are three possible explanations," the expert replied. "But have we examined what the cases share?"

They compared the dates on which the disturbances first appeared. Although the presses had been running different jobs on different materials, the complaints had emerged within a relatively short period.

The expert did not claim that one common cause must exist. Nor did he dismiss the local teams' observations. He proposed a different starting point: instead of investigating each press only as an isolated machine, they should first ask what had changed across the plant.

That was the first turning point. The unit of investigation expanded from individual presses to a production system with shared materials, procedures, and decisions.

When Experience Becomes a Constraint

When a disturbance appears on a press, it is natural to investigate the subsystem that seems most closely related to the visible defect.

A shadow following heavy image coverage suggests checking image formation and transfer. Tonal variation suggests calibration, color response, and operating conditions. A defect that is strongest on one substrate suggests material suitability and substrate settings.

None of those lines of inquiry was unprofessional. In fact, each was a defensible first hypothesis. The problem was that hypotheses were gradually being treated as established explanations.

A temporary improvement after calibration was taken as evidence that calibration had caused the problem. A change following a transfer adjustment reinforced the mechanical theory. The greater visibility of the symptom on one substrate appeared to confirm that the substrate was responsible.

Yet an intervention followed by improvement does not, by itself, demonstrate causation.

The expert asked the teams to distinguish three things: the observed symptom, the proposed mechanism, and the evidence supporting that mechanism. He also asked for a complete record of interventions made before his arrival.

On the first press, the list included pressure adjustments, another calibration, cleaning, and a modified working point. The second press had accumulated a different sequence of changes. The third had been investigated primarily through substrate comparisons.

By trying to improve the process, the teams had unintentionally made the original failure harder to reproduce and understand. The expert therefore proposed returning, wherever safe and practical, to documented, qualified baseline conditions.

An engineer objected. "You want to undo adjustments that made the output better."

"I want us to understand what made it better," the expert replied. "Without a repeatable failure under known conditions, we cannot reliably prove what resolves it."

Several hours were spent reconstructing the testing conditions rather than making further corrections. When the first press was returned to its controlled baseline, the tonal disturbance reappeared.

This time, that was progress. For the first time in days, the team had a reproducible symptom that could be challenged through a meaningful experiment.

Two Changes Nobody Had Thought to Connect

A process engineer began comparing the histories of all three presses: production jobs, substrates, maintenance, shift patterns, environmental records, and changes introduced shortly before the complaints.

Most obvious common explanations weakened. The presses had not been producing the same jobs. No common file family matched the timing. Substrates and maintenance histories differed.

Then two details emerged.

First, all three presses had recently received consumable material from the same new supplier production lot. The consumable remained approved, incoming checks had not identified a failure, and no formulation or product change was declared.

Second, the plant had revised a cleaning procedure during the same period. The cleaning material itself remained approved. The adjustment concerned the sequence and timing of the procedure, introduced to make cleaning more consistent between shifts.

Neither change looked inherently problematic. Purchasing had no reason to reject the new consumable lot. Maintenance had already validated the revised cleaning procedure. Both were considered acceptable, and the procedural change had been introduced as an improvement.

One engineer pointed to the consumable. "That must be it. All three presses received the lot before the trouble started."

"And the cleaning procedure?" the expert asked.

"That was changed too, but the revised procedure was tested and approved."

"So was the consumable," the expert replied.

For a moment nobody spoke. The two changes had each been assessed independently. Nobody had tested how they behaved together.

What If the Problem Is the Interaction?

Instead of asking the team to choose between the new consumable and the revised cleaning procedure, the expert proposed a third possibility: neither might be sufficient on its own to create a significant disturbance, while their combination could make the process more sensitive.

In engineering, this is the idea of an interaction effect: the response to one variable depends on the level or condition of another.

A consumable may perform acceptably under one cleaning method. A different cleaning method may work well with a consumable whose surface properties sit elsewhere within its approved manufacturing range. Combined, however, those otherwise permissible conditions may narrow the stable process window.

That does not mean either element is automatically defective. It means the system's response cannot always be predicted by examining each input in isolation.

The possibility challenged the plant's usual troubleshooting practice, which had largely involved changing one parameter at a time while leaving the current production state in place. In this case, changing a single factor might improve the symptom without revealing its dependence on the other factor.

The expert proposed a structured comparison of both variables.

Four Controlled Combinations Instead of Twenty More Adjustments

The first press was selected because it showed the clearest reproducible disturbance. Two factors were examined at two levels each: the earlier versus the new consumable lot, and the earlier versus the revised cleaning procedure.

Together, these produced four controlled combinations:

These were comparative, qualitative observations. The case did not provide numerical measurements and replicated statistical analysis sufficient to calculate an interaction coefficient or establish formal statistical significance. The design nevertheless allowed testing whether the combined conditions behaved differently from the original baseline.

The team began with the current production combination: new consumable and revised cleaning. The tonal disturbance appeared again.

They then introduced retained material from the earlier lot while keeping the revised cleaning procedure. The disturbance weakened, but it did not disappear completely.

Next, they restored the new lot while using the earlier cleaning procedure. The result improved again.

Finally, they combined the earlier lot with the earlier cleaning procedure. This produced the cleanest image. The comparisons were repeated to check consistency. When the team returned to the current production combination, the disturbance became stronger once more.

For the first time, the evidence showed that both changes mattered under the tested conditions. The issue was no longer plausibly described as a simple failure of one component.

Within Specification — but Without Enough Margin

The supplier data were reviewed again. The new consumable lot remained within the approved specification. Incoming inspection had not identified a nonconformity.

A closer comparison with retained material from the previous lot, however, revealed a small difference in a surface-response indicator. Both values were inside the accepted manufacturing range, but they occupied different parts of that range.

The revised cleaning procedure was not intrinsically invalid either. It passed its own qualification checks and caused no obvious problems in most production work.

The comparison showed that the new combination left the process more vulnerable to a sensitive image sequence. It did not establish the complete microscopic or chemical mechanism connecting surface response and cleaning history; that remained an engineering and supplier investigation.

The central lesson was more fundamental. Specifications for individual inputs do not necessarily describe every interaction within the assembled process.

If each component or procedure is permitted some normal variation, particular combinations may leave too little operating margin for demanding applications. Compliance is essential, but compliance alone does not guarantee the same robustness under every allowable combination of conditions.

Why Did the Symptom Follow Heavy Ink Coverage?

One question still needed attention. If the shared sensitivity involved consumable condition and cleaning practice, why was the disturbance most obvious after areas of high image coverage?

The expert asked the team to treat the artwork sequence as a diagnostic probe rather than automatically defining it as the root cause.

In LEP printing, image formation, development, and transfer rely on several electro-physical and material interfaces operating in a controlled process window. A test sequence in which high-coverage image areas are followed by light or midtone regions can make particular forms of residual tonal or transfer instability more visible.

That does not prove that heavy coverage chemically altered the consumable or created the interaction. The experiment showed that this image sequence was particularly effective at revealing the sensitivity associated with the tested production conditions.

The distinction matters. A customer job may expose a weakness without causing it. When engineers confuse the trigger that makes a defect visible with the mechanism that makes the process vulnerable, they can spend considerable effort changing artwork or color settings without addressing the underlying margin loss.

The Shared Explanation Was Still Not Enough

After the successful comparisons on press one, several team members believed the investigation was essentially over. The expert disagreed. Evidence of a common factor did not prove that all three presses had the same failure mechanism.

They took the sensitive test-image sequence to press two and performed only the comparisons needed to challenge the shared hypothesis there.

The same pattern appeared: the combination of the new consumable and revised cleaning increased the disturbance. But the result was stronger and somewhat different from the effect seen on press one.

If the common interaction were the only relevant factor, the remaining difference would be difficult to explain. Further inspection identified a local alignment sensitivity that amplified one portion of the visual disturbance.

The second press therefore had two interacting layers of concern: a sensitivity shared with the others, plus a local alignment condition that worsened its manifestation.

The press-two team had not been entirely wrong to suspect its own equipment. It had identified part of the picture, not the whole picture. Treating the common explanation as final would have concealed that additional local issue.

The Third Press Added Another Layer

Press three behaved differently again. The common consumable-and-cleaning combination could be shown to influence the result, but the disturbance was comparatively weak on some jobs.

On one substrate, however, it became visually prominent.

That had originally led the third team to identify the substrate as the cause. The new comparisons let the expert test a more precise explanation: the substrate's surface and optical characteristics could increase the visibility of a disturbance generated by the broader process conditions.

In other words, the substrate did not necessarily create the common sensitivity. It made that sensitivity easier to see.

The local team's observations were valid, but its causal conclusion went beyond what the evidence justified. A substrate that makes a defect obvious is not necessarily the substrate that produces the defect.

By the end of the investigation, the differences across the three presses finally fit together. Press one exposed the shared sensitivity most cleanly. Press two had that sensitivity plus an additional alignment condition. Press three showed the common effect more subtly until a particular substrate amplified it visually.

Three different diagnoses had each captured something real. None alone described the full system.

A System of Relationships, Not Three Isolated Failures

The team returned to the meeting room where the three case folders had been placed at the beginning of the week.

The picture was now much clearer. Two changes had entered the plant within a similar period: a new consumable lot and a revised cleaning procedure. Each satisfied the acceptance requirements applied to it individually. Together, under a sensitive image sequence, they produced a less stable response.

On top of that shared sensitivity, each press had its own conditions affecting how strongly the disturbance appeared and how it looked.

There was no defensible reason to label the entire new lot defective or to declare the revised cleaning procedure universally unacceptable. The results supported a problematic combination under the tested conditions, not a universal verdict on either element.

They also showed why the local variations could not simply be collapsed into a single, elegant root-cause sentence.

Keeping Production Running Without Pretending Every Question Was Answered

The plant could not wait for a complete materials-science explanation before making production decisions. Customers expected deliveries and management needed an actionable containment plan.

For the most sensitive work, the team temporarily restored the earlier cleaning procedure under controlled conditions. The team used retained stock from the earlier consumable lot for critical jobs, supported by test results and available inventory.

A supplier investigation was opened to determine whether ordinary lot-to-lot variation, although still within specification, could reduce margin when combined with the revised cleaning process. The cleaning change itself was also reviewed, not because it had been proved universally wrong, but because its original qualification had not revealed this interaction.

Press two required separate attention to the local alignment sensitivity. On press three, quality control needed to recognize that a substrate could amplify the visibility of an existing disturbance without being the sole underlying cause.

The process and service teams also developed a short, sensitive test-image sequence. It could be used after significant changes to consumable lots or cleaning practice, so that reduced process margin could be detected before a customer job exposed it.

The containment actions supported controlled production. The supplier and engineering work remained open.

Why Closing the Case Too Early Would Have Been a Mistake

Industrial quality systems often expect an investigation to produce one identified root cause, one corrective action, and one closure date. That structure supports accountability, but not every technical failure fits neatly into one line of a report.

In this case, the team had demonstrated a reproducible relationship between two changed conditions and the tonal disturbance. It had also identified additional local factors that altered how the effect appeared on two of the presses.

What had not yet been demonstrated was every physical or chemical step responsible for the shared response. Understanding that mechanism more completely would require further examination of the consumable, its interfaces, and the cleaning process.

There is a difference between gathering enough evidence for a controlled production decision and proving a complete failure mechanism at the material level.

Recognizing that difference is not a weakness. It is disciplined engineering. Prematurely labeling an acceptable lot 'defective' or a qualified procedure 'wrong' could lead to costly decisions that the available data did not justify.

What Was Different About the Expert's Approach?

The visiting expert did not arrive knowing that the consumable lot and cleaning procedure would prove important. He did not know about the extra alignment sensitivity on press two or the optical amplification on press three.

His contribution was the way he framed the investigation.

Instead of treating each press as a separate technical world, he searched for changes shared across the plant. Instead of choosing between two competing suspects, he asked whether their effects depended on one another. Instead of accepting improvement after one change, he proposed comparisons that separated the combinations.

And when a common influence emerged, he did not stop. He asked why the presses still behaved differently.

That was the outside-the-box thinking the original investigation had been missing. No secret parameter, no unusual part replacement, and no miraculous adjustment: simply the willingness to change the level of analysis, from one press to a whole production system, and to examine relationships that isolated checks could not reveal.

The Most Important Lesson: Compliant Does Not Necessarily Mean Robust

Industrial printing relies on specifications, tolerances, approved operating windows, and repeatable quality procedures. Those controls are essential.

But a component meeting its specification does not guarantee that every permitted combination of components and procedures will deliver the same result. A validated cleaning procedure can expose an unexpected sensitivity when another input changes within its normal tolerance.

A process that produces one excellent sample at a favorable operating point is not necessarily a process that can sustain production across ordinary, allowable variation.

A robust process needs adequate margin to absorb expected material and operational differences without moving the finished product outside its acceptance requirements.

In this case, the new combination reduced that margin. A sensitive image sequence exposed the disturbance, and local press or substrate conditions changed how visible it became.

That is why a small, well-designed experiment can teach a team more than dozens of interventions whose variables are never properly separated.

Beyond Solving the Fault: Changing the Investigation Method

At the final technical review, production, quality, maintenance, and engineering no longer presented three unrelated case reports. They presented one investigation explaining what the presses shared, where they differed, what had been demonstrated, and what still required study.

The immediate production actions were clear. Containment was in place. Supplier work continued. The sensitive test sequence was available for future change validation.

More importantly, the plant had learned to investigate differently.

When several presses develop related symptoms within a short period, compare plant-wide changes before allowing every press to build its own independent theory. When more than one relevant change is identified, test whether their effects depend on one another. After finding a shared influence, keep looking for the local conditions that alter the outcome on each machine.

Complex problems often require us to understand the interactions between components and processes, not merely the components themselves.

Conclusion: Sometimes There Is No Single Component to Blame

Three LEP presses began showing quality disturbances during the same week. One team suspected image transfer. Another focused on calibration. A third blamed the substrate.

Each team had observed something real. None had seen the complete picture.

The investigation showed that two individually acceptable changes could combine to make the process more sensitive. Local conditions on the other presses then shaped the disturbance's severity and appearance.

The operational recovery did not come from replacing one part or performing another calibration. It came from structured comparisons that exposed dependencies between variables, enabled better-controlled production conditions, and honestly defined the unresolved questions.

Industrial reality does not always conform to a simple chain of one cause and one effect. Sometimes several compliant inputs, approved procedures, and legitimate operating conditions interact to produce an unexpected result.

Thinking outside the box is not simply seeing something others missed. It is recognizing that the problem may lie not in any factor alone, but in the relationship between them.

Once we understand that, we stop asking only which component needs fixing. We begin asking how the whole system behaves—and why.

Author's note: This article is adapted from "Three Presses, the Same Sign," in The Personal Journey of an Industrial Printing Expert by Ronen Dolev. It presents a fictionalized professional case informed by industrial troubleshooting principles, not a documented account of an identifiable plant. The purpose is to illustrate controlled experimentation, interactions between variables, and the distinction between demonstrated evidence and mechanisms still under investigation.