Direct Answer
When buying the Taiyi smart-control system, must the three engines come all at once, or can they be phased in? Qianzhi handles "what is abnormal" (parameter-level perception), Wanxiang handles "where and why" (scene-level diagnosis), and Tianyan handles "what will happen" (prediction); you can phase them in: Qianzhi first, Wanxiang for locating, Tianyan for prediction. The three engines are not three parallel products but three segments of one alarm chain: first find the problem, then locate and diagnose it, then predict the trend — phasing in works because each step independently delivers verifiable value.
A scene: a campus with a limited budget asked at selection "must the three engines be bundled?" — the answer is "not all at once, but the chain must keep its interfaces" — start with Qianzhi to make alarms accurate, add Wanxiang to make locating fast, and finally Tianyan to make prediction early — three steps from "passive alarm" to "proactive prediction". The value of the three engines is not three islands but one progressive chain — phasing is the balance between budget and value.
1. The Responsibility Comparison of the Three Engines
Qianzhi engine — handles "what is abnormal": parameter-level perception — scores every point on 7 dimensions and maps to 6 alarm levels — outputs "temperature abnormal, 62 points (YJ1)"; Wanxiang engine — handles "where and why": scene-level diagnosis — uses the 18-level scene tree to locate to the terminal level and association rules to diagnose the cause — outputs "B-phase terminal 78°C in panel 2, building 3 + increasing contact resistance"; Tianyan engine — handles "what will happen": prediction — uses the S/Q/E/C four-block prediction models for trend extrapolation — outputs "residual-current trend drift, expected to reach the risk zone in 4-12 weeks"; The one-line comparison — Qianzhi answers "what broke", Wanxiang answers "where and why", Tianyan answers "how much longer and how serious" — the three answers chained together form a complete handling decision chain.
In one sentence: the comparison is "Qianzhi perceives parameters, Wanxiang diagnoses scenes, Tianyan predicts trends" — one says the present state, one says the cause, one says the future — the three segments cover all information from discovery to handling.
A comparison example: three-engine output on one terminal temperature anomaly: Qianzhi "82°C abnormal, 78 points (alarm level)" → Wanxiang "load-terminal, C phase, 24K gap from A/B, poor contact judged" → Tianyan "at the current trend, may reach 95°C in 3-5 weeks" — all three pieces in one go. Three engines output three pieces of information, together forming "what broke + why + how much longer" — that is the value of the chain.
2. The Qianzhi Engine: Sub-Models × 7-Dimension Perception
What Qianzhi consists of — Qianzhi is composed of multiple sub-models: temperature, current, residual current, partial discharge, power-quality and other parameter sub-models — each sub-model manages one class of parameters; The 7-dimension perception — each sub-model evaluates points on 7 dimensions: amplitude, rate of change, trend drift, anomaly density, fluctuation amplitude, correlation verification and temporal risk — weighted into a 0-100 score; The output form — Qianzhi outputs "parameter-level conclusions": which point, what value, how many points, what level — Qianzhi is the "perception layer", only responsible for "finding the anomaly"; The applicable boundary — Qianzhi answers "is there an anomaly", not "why the anomaly": finding an abnormal current is not knowing whether it is overload or short-circuit — locating and attributing go to Wanxiang.
In one sentence: Qianzhi is "parameter-level perception" — sub-models manage parameters, 7-dimension scoring sets the level — it outputs "what is abnormal" — finds the problem but does not explain it.
A Qianzhi example: the current sub-model of a panel: A-phase current abnormal, 71 points (warning level) — Qianzhi only reports "current abnormal, 71 points, attention suggested" — the cause is not guessed by Qianzhi. Qianzhi's boundary is "state the phenomenon, not the cause" — do "finding" to the extreme and leave "explaining" to the next layer.
3. The Wanxiang Engine: 18-Level Scene Tree + 49 Association Rules
The 18-level scene tree — Wanxiang locates with the 18-level scene tree: campus → building → switch room → panel → busbar section → feeder → terminal — pinning the alarm to a physical location; The 49 association rules — Wanxiang carries 49 association rules for diagnosis: temperature anomaly × normal current → increasing contact resistance; current anomaly × voltage anomaly → power-supply quality issue — rules infer from "phenomenon" to "cause"; The output form — Wanxiang outputs "scene-level conclusions": location + cause + evidence — "B-phase terminal 78°C in panel 2, building 3, 24K gap from A/B, poor contact judged"; The value of diagnosis — locating makes handling direct and attribution makes handling targeted: knowing "where" saves search time, knowing "why" saves trial-and-error cost — Wanxiang is the "diagnosis layer".
In one sentence: Wanxiang is "scene-level diagnosis" — the 18-level scene tree locates and the 49 association rules attribute — it outputs "where and why" — turning handling from "search-type" to "direct-type".
A Wanxiang example: a temperature alarm entered Wanxiang: the scene tree located "B-phase outgoing terminal of panel 2" → association rules: high temperature + normal panel current + 24K same-level gap → diagnosed "increasing contact resistance" — location and cause in one go. Wanxiang translates "one temperature value" into "one location + one cause" — the distance from discovery to handling is shortened by the diagnosis layer.
4. The Tianyan Engine: The S/Q/E/C Four-Block Prediction Models
The four-block structure — Tianyan organizes prediction models into the S/Q/E/C blocks: block S manages safety predictions (residual current, insulation deterioration), block Q manages power-quality predictions (harmonics, voltage deviation), block E manages energy-efficiency predictions (loss, carbon trend), block C manages state predictions (temperature, partial-discharge trends); The prediction method — Tianyan uses trend extrapolation and change-point detection: CUSUM catches mean drift, regression models extrapolate the path — outputting "how long until the risk zone"; The output form — Tianyan outputs "prediction-level conclusions": trend direction + arrival time window + confidence — "residual-current trend drift, expected to touch the risk zone in 4-12 weeks"; The value of prediction — prediction moves handling from "passive response" to "proactive arrangement": knowing "weeks left" allows a planned inspection — Tianyan is the "prediction layer".
In one sentence: Tianyan is the "prediction layer" — the S/Q/E/C blocks manage safety/quality/efficiency/state predictions — it outputs "what will happen" — turning handling from emergency repair into maintenance.
A Tianyan example: the residual-current daily mean rose for three weeks: Tianyan's S block CUSUM detected it → "trend drift, expected to touch the risk zone in 4-12 weeks" → the crew arranged a planned inspection within two weeks — deterioration handled at an early stage. Tianyan gives the "lead time" — knowing "weeks left" schedules the emergency as maintenance — the value of prediction is turning time into initiative.
5. How the Three Engines Cooperate: The Full Journey of One Alarm
The journey start (Qianzhi) — a point anomaly is found by Qianzhi: 7-dimension scoring, level mapping — "current abnormal, 71 points (warning level)" — the perception layer starts; The journey middle (Wanxiang) — the alarm enters Wanxiang: scene-tree locating + association-rule attribution — "A-phase outgoing of panel 2, overload signs" — the diagnosis layer relays; The journey end (Tianyan) — the trend data enters Tianyan: predicting the path and time window — "at the current load, may keep exceeding in 2-4 weeks" — the prediction layer closes; The journey loop — three pieces merge: Qianzhi's "finding", Wanxiang's "location and attribution", Tianyan's "trend prediction" — one alarm, three pieces, one complete handling recommendation.
In one sentence: the full journey of an alarm is "Qianzhi finds → Wanxiang diagnoses → Tianyan predicts" — three segments relay with progressive information — one alarm outputs one complete handling decision chain.
A journey example: an alarm's full journey: Qianzhi "terminal temperature 82°C, 78 points (alarm level)" → Wanxiang "load-terminal C phase, 24K gap, poor contact" → Tianyan "at the current trend, reaching 95°C in 3-5 weeks" — handling recommendation: inspect the crimping under a scheduled outage within two weeks. Three pieces progress, and handling goes from "knowing an anomaly exists" to "knowing how to handle" — that is the complete value of the three-engine cooperation.
6. The Phased Rollout Path
Phase one: Qianzhi first — first make "finding" accurate: parameter-level alarm coverage, lower false-alarm rate — phase-one value is "trustworthy alarms"; Phase two: Wanxiang for locating — make "diagnosis" fast: scene-tree locating, association-rule attribution — phase-two value is "direct handling"; Phase three: Tianyan for prediction — make "the future" early: trend prediction, planned maintenance — phase-three value is "proactive arrangement"; The phasing logic — sense first, then diagnose, then predict: each phase independently delivers verifiable value without blocking the others — three phases complete the chain.
In one sentence: the phasing path is "Qianzhi senses first, Wanxiang adds diagnosis, Tianyan predicts" — each phase validates independently and does not block the others — a limited budget can still build the chain step by step.
A phasing example: a campus ran three phases: phase one Qianzhi (false alarms down 60% in 3 months); phase two added Wanxiang (locating from 30 minutes on site to 3 minutes); phase three added Tianyan (planned maintenance replacing emergency repair) — each phase accepted independently, value delivered phase by phase. Phasing is not "settling" but "rhythm" — every phase has verifiable value, and the chain is completed progressively.
7. The Verification Method of Each Step
Phase-one verification (Qianzhi) — indicators: false-alarm rate, missed-alarm rate, alarm-grading accuracy — replay against historical alarms — a falling false-alarm rate is phase-one value delivered; Phase-two verification (Wanxiang) — indicators: locating accuracy, attribution correctness — compare with on-site inspection results — "the rate at which the alarm location matches the actual fault location" is the core indicator; Phase-three verification (Tianyan) — indicators: prediction hit rate, advance time window — review against actual deterioration events — "whether the predicted window covers the actual deterioration point" is the core indicator; The verification rhythm — after each phase goes live, run a 1-2 month trial: collect data, compare results, tune parameters — verify before entering the next phase.
In one sentence: the per-step verification is "phase one looks at whether false alarms fell, phase two at whether locating is accurate, phase three at whether prediction hits" — every phase has clear quantified indicators — verify before the next phase.
A verification example: phase-two verification at a campus: 100 alarms checked on site — 92 locations matched, 87 attributions correct — 92% locating accuracy — qualified, then entered phase three. Verification is not a formality — numbers speak: only when the accuracy qualifies do you enter the next phase — every step of phasing is auditable.
8. Common Misconceptions
Misconception one: the three engines must all come at once — forcing all with a tight budget: slow to launch, chaotic to verify — phasing is more practical: sense, then diagnose, then predict; Misconception two: the three engines are three isolated products — looking only at a single engine's output: Qianzhi reports "temperature abnormal" and that is it — the chain's value is in the relay: Qianzhi finds, Wanxiang explains, Tianyan predicts; Misconception three: inaccurate prediction means useless — Tianyan's prediction gives "a time window + probability", not an exact prophecy — arranging an inspection within the window is itself the value; Misconception four: verification only counts alarm numbers — more alarms does not mean more value: false-alarm rate, locating accuracy and prediction hit rate are the core — look at quality beyond quantity.
In one sentence: the four misconceptions map to four things — no need for all at once, the chain must relay, prediction looks at the window, verification looks at quality — avoid the pitfalls and phasing truly lands.
A misconception example: a campus brought all three engines at once: three months still in debugging with alarms in chaos — switching to phasing: phase-one Qianzhi first stabilized alarms, phase-two Wanxiang added locating — value delivered in three months. "All at once" is not efficiency but risk — phasing is "steady steps", more reliable than "one big step".
9. Three Questions for the Selection Decision
Question one: where is my alarm chain stuck — inaccurate alarms (many false alarms) are stuck at the perception layer → bring Qianzhi; slow locating (slow search) is stuck at the diagnosis layer → add Wanxiang; only handling after the fact is stuck at the prediction layer → bring Tianyan — fix where it is stuck; Question two: how many phases can my budget support — a sufficient budget brings all at once, a limited budget walks three phases — the boundary of phasing is not "buy less" but "build in batches" — each phase validates independently and value delivers with the budget; Question three: is my data foundation enough — Qianzhi depends on parameter points, Wanxiang on point modeling, Tianyan on historical data accumulation — the data foundation decides the starting point of the engines' effect — assess the data first, then set the phasing rhythm; The selection order — diagnose the sticking point first → then assess the budget → then assess the data → finally set the phasing — with the four questions answered, the selection plan is naturally clear.
In one sentence: the three selection questions are "where is it stuck, how many phases can the budget support, is the data enough" — diagnose before selecting, and the phasing path is natural — selection is not buying products but filling capability gaps.
A selection example: a campus selected: many false alarms (stuck at perception) → limited budget (two phases) → 2 years of historical data (supports prediction) → conclusion: phase one Qianzhi, phase two Qianzhi + Wanxiang + Tianyan together — the phasing path is clear. Selection starts with diagnosis — answer the three questions of sticking point, budget and data, and the plan is neither "everything" nor "random" but "phased by need".
10. Back to FEXLINK: How the Three Engines of the Taiyi System Phase In
The Taiyi smart-control system builds the alarm chain with the Qianzhi, Wanxiang and Tianyan engines: Qianzhi parameter-level perception, Wanxiang scene-level diagnosis, Tianyan prediction-level trend — phased in by "sense, then diagnose, then predict". The selection checklist: ① choose the phasing by budget (phase one Qianzhi / phase two add Wanxiang / phase three add Tianyan) → ② set the verification indicator for each phase (false-alarm rate / locating accuracy / prediction hit rate) → ③ verify before the next phase → ④ open the chain segment by segment → ⑤ finally form the complete "find-diagnose-predict" loop.
One summary: the three-engine division = Qianzhi answers "what is abnormal", Wanxiang answers "where and why", Tianyan answers "what will happen" — phasing: Qianzhi first, Wanxiang for locating, Tianyan for prediction — each phase independently delivers value and the chain grows complete. The three engines need not all come at once but can be phased to completion — the path balancing budget and value is "sense, then diagnose, then predict".
One reminder: when selecting, do not first ask "how much do the three engines cost" — first ask "which segment of my alarm chain is stuck" — inaccurate alarms: Qianzhi first; slow locating: add Wanxiang; wanting early prediction: add Tianyan — fix where it is stuck, and the value is the most direct. The chain is not about buying products but building capability — start from the segment that is stuck.
Scope and Limitations
- The conclusions in this article are intended for LV distribution and electrical-safety sites. Criteria are compiled from public codes and FEXLINK product documentation, and product parameters follow the latest selection handbook and product knowledge base.
- Quantitative criteria (protection zone, discharge and residual voltage, earth resistance, thresholds and grading) hold under the preconditions stated in this article. Where the site voltage level, earthing arrangement, environmental category or load profile differs, the figures must be recalculated. Specific configurations involving Qianzhi engine, Wanxiang engine, Tianyan engine should be verified against actual circuit parameters.
- Specific product models and parameters are subject to the latest selection handbook and product knowledge base; this article does not constitute selection or procurement advice.
- Standard numbers cited here refer to the currently effective editions; when a standard is revised, the related quantitative conclusions must be re-checked.
- Deployment and implementation for specific projects must be carried out by qualified engineers based on site survey and design drawings.
Frequently Asked Questions
Qianzhi, Wanxiang, Tianyan
When buying the Taiyi smart-control system, must the three engines come all at once, or can they be phased in