For many chemical production companies, equipment failure analysis is nothing unfamiliar. It usually sits in an Excel spreadsheet: during a unit overhaul or annual review, engineers pull it out, conduct an inspection and re-assessment, print it, sign it, and file it—and the next time it is opened is most likely a year later.
But completing an FMEA does not mean risks are under control. Identifying failure modes and assigning scores is only the first half of the journey; the second half is ensuring that every relatively high risk has corresponding measures, that those measures are executed by someone, that there are verification records, and that the risk is truly reduced—in other words, closing the loop. The difficulty lies in the fact that these actions take place in many locations and are handled by many people, and a static spreadsheet can hardly string them together.
01
Four Common Management Bottlenecks
-
Analysis relies on passing files around; the more versions there are, the less clear it becomes.
An analysis for a piece of equipment is produced in one version and sent to the process team; the process team revises it and sends it back; the equipment team then adds another round—so whose copy is authoritative in the end? Merging relies on visual comparison, conflict resolution relies on the memory of those involved, and there is no confidence that the finally archived copy is the latest and most complete. Historical versions generated by annual re-assessments are also difficult to retain, manage, and consult.
-
Scoring criteria are inconsistent, distorting risk levels.
Once the template is copied onto each person’s computer, it becomes their own version: scoring standards are hard to unify, different people interpret them differently, assessment results are easily influenced by personal factors, and cross-checking is inconvenient. After rating scale definitions are changed, items under assessment do not change accordingly, making old and new data easy to confuse.
-
Measures are filled into the table, but execution falls outside it.
The moment a measure is written down, management has only just begun: whether it has been scheduled into the maintenance plan, whether it has gone through management-of-change approval, where the verification records are, and how it was ultimately executed—all of this happens outside the spreadsheet, and status is backfilled manually. At the end of the year, when someone asks “How many items are still open?”, often no one can answer immediately.
-
Analysis has been done for years, but experience has not been retained.
Year after year, what accumulates is hundreds of files, not a searchable dataset. A large number of the plant’s equipment repair, maintenance, and defect-elimination records cannot be effectively utilized. The veteran engineer’s knowledge in his head—“where this type of equipment is most prone to failure”—may be inaccurate or forgotten, and cannot be passed on to newcomers, so it cannot contribute value in assessments.
02
How Gongzhidao FMEA Analysis Software Works
-
Risk inventory, visible at a glance on one screen
All analyses live in a single repository, everyone accesses the same interface, and different versions for the same equipment are clearly displayed. Existing Excel analysis results can be imported directly, with no need to start over. The home page summarizes the number of independently analyzed equipment items, the number of higher-risk failure modes, the number of open improvement measures, and the FMEA analysis completion rate—the risk baseline is there as soon as you open it, without having to pull a pivot table.
FMEA Results Summary Page: Four Persistent Stats Cards
-
Scoring Criteria: One Yardstick for the Entire Plant
The three configuration pages for Severity, Occurrence, and Detection, plus a risk matrix, serve as the sole maintenance entry point for plant-wide scales; a single change takes effect across the entire plant. The built-in scales are tailored to the realities of chemical enterprises: for each Severity level, personnel injury, shutdown scope, and direct loss are specified; Occurrence provides two sets of scales—frequency range f (occurrences per unit per year) and failure rate λ (failures per 10⁶ hours)—so that frequency is used when operational data is available, and failure rate when reliability data is available; Detection ranges from “SIS/ESD automatically detects and acts” to “no detectable indication whatsoever.” Each level also supports customized configuration according to the plant’s specific conditions. The risk acceptability level is automatically determined by the matrix based on S/O/D, with clear standards.
Occurrence Configuration Page: Dual Scales for Frequency Range and Failure Rate
Risk/Criticality Matrix: Automatic Level Determination
-
Structure Broken Down to the Component Level; Upper-Level Scores Automatically Inherited
A piece of equipment is expanded in a three-level indented structure: L1 equipment level, L2 component level, and L3 part level. Component-level failures are linked to the equipment-level failures they affect, and severity is automatically inherited with its source indicated—no duplicate scoring across levels, so during reviews there is no need to argue over what score the same consequence should receive.
-
Improvement Measures, Directly Flowing into Daily Operations
Naturally integrated with Gongzhidao’s existing business modules, each measure automatically flows to the corresponding module by type: initiate a change (MOC), add to inspection plan, add to maintenance plan, add to preventive maintenance, or add to work orders. Each measure is linked to the implementation evidence and verification records of the business measure; click to open it and you see the actual business document in the system. The full-process closed-loop tracking of improvement measures is automatically connected by the system.
In the analysis task, the “Improvement Measures/Implementation Status” column: measures are automatically linked to inspection, maintenance, change, work orders, and preventive maintenance, and display implementation status.
-
Historical Experience as the Starting Point for New Analyses
When creating an analysis for a new piece of equipment, there is no need to start from a blank sheet. AI-assisted assessment provides references from four sources at once: public industry datasets, plus the plant’s own historical data—failures that have occurred on the same equipment, records of same-model equipment at other locations, and equipment groups of the same type. Each Occurrence rating is tagged with its source and the number of records, and whether the sources are consistent or inconsistent is clearly flagged by the system. With a click, you can view document links and trace back to the original records. All AI-generated results are subject to human confirmation—AI drafts, and engineers finalize.
AI-Assisted Assessment: Four-Source Data and Consistency Annotation
03
Once It’s in Use, the Changes Show Up in These Areas
Risk baseline is always visible:
Higher-risk items, open measures, and analysis completion rate stay on the home page—no need for a year-end rush to compile them.
Scores are comparable across people and equipment:
One set of scales for the entire plant, so horizontal comparisons and year-over-year comparisons hold up.
Closed-loop measures are traceable with evidence:
Measures land on real business documents, and status, execution, and verification can all be traced back. The system integrates with the dual prevention mechanism and the equipment full-lifecycle management system, so ready-made records are available when inspections come.
Experience accumulates in the system:
Historical analyses are data, not files. Analyses for new equipment start from AI-assisted assessment, giving a starting point far higher than a blank sheet.
