Engineering Knowledge Cards.
Confidence HighUpdated 2026-10-18

Time parameters of activity-on-node networks: the recursion of time lag LAG(i,j), free float FF = min LAG and total float TF = min(LAG + TF of successors); the virtual end node's duration value and critical path identification (JGJ/T 121-2015 clauses 5.3.5~5.3.9, 5.5.1, 5.5.2)

Per JGJ/T 121-2015 clauses 5.3.7~5.3.9: for adjacent activities, LAG(i,j) = ES of successor - EF of the activity (with a virtual end node, LAG(i,n) = Tp - EFi); free float FF = min of the LAGs to all successors; total float must be computed activity by activity backwards from the end node - TF of the end activity = Tp - Tc, others TF = min(LAG + successor TF). With no required duration, Tp = Tc, so the end node's TF = 0; with a required duration Tr < Tc it is negative. 'The activities with the least total float are critical'; the critical path further requires zero lag between adjacent critical activities (all LAG = 0) along the path.

Applicable Codes

  • JGJ/T 121-2015 工程网络计划技术规程 第5.3.5条
  • JGJ/T 121-2015 第5.3.6条
  • JGJ/T 121-2015 第5.3.7条
  • JGJ/T 121-2015 第5.3.8条
  • JGJ/T 121-2015 第5.3.9条
  • JGJ/T 121-2015 第5.5.1条
  • JGJ/T 121-2015 第5.5.2条
  • JGJ/T 121-2015 术语 2.1.35

Positioning

The "time-lag method" (LAG method) for activity-on-node networks (nodes are activities, arrows are logic): the recursions and order for the time lag LAG, free float FF and total float TF, the end (virtual) node's duration value, and the criteria for critical activities and the critical path. Basis: JGJ/T 121-2015 clauses 5.3.5~5.3.9, 5.5.1, 5.5.2 and term 2.1.35.

Core conclusions

1. Time lag LAG (clause 5.3.7) - two cases:

  • with a virtual end node: LAG(i,n) = Tp - EF of activity i (planned duration minus its earliest finish);
  • otherwise: LAG(i,j) = ES of j - EF of i.

2. Earliest times (clauses 5.3.3, 5.3.4) The start node's activity has ES = 0 unless prescribed; other activities ES = max of the predecessors' EF; EF = ES + duration; computed forwards along the arrows.

3. Computed and planned durations (clauses 5.3.5, 5.3.6)

  • Computed duration Tc = EF of the end activity.
  • Planned duration Tp: with a required duration Tr, Tp <= Tr; without, Tp = Tc.

4. Free float (clause 5.3.9)

  • End activity n: FF = Tp - EF of n;
  • others: FF = min of the LAGs to all successors.

5. Total float (clause 5.3.8) Direction fixed by the code: "the total float TF of activity i shall be computed item by item from the end node of the network, against the arrow direction."

  • end activity: TF = Tp - Tc;
  • others: TF = min of (LAG(i,j) + TF of j) over all successors j - successors first.

Hence: without a required duration Tp = Tc, so the end node's TF = 0 (and its FF = 0); with Tr < Tc, the end node's TF = Tr - Tc is negative and the network minimum total float is negative - critical activities are then those with the least total float, not those with TF = 0.

6. The two procedures offered by the code First: earliest starts and finishes - lags - free and total floats from the lags - latest starts and finishes from the total floats. Second: earliest starts and finishes - latest finishes and starts - total and free floats. Both give the same result; the first pivots on LAG and is the more direct for AON networks (no prior backward LF/LS pass).

7. Critical activities and path (clauses 5.5.1, 5.5.2; term 2.1.35)

  • Clause 5.5.1: "The activities with the least total float shall be identified as critical." With Tp = Tc the least total float is 0 - TF = 0 means critical.
  • Clause 5.5.2: "The path identified as critical is one composed wholly of critical activities from start to end with zero lag between them, or the path of longest total duration"; mark it bold, double or coloured. Term 2.1.35 for AON: "a path composed of critical activities with zero time lag between them, or the path of longest total duration".
  • So an AON critical path needs both: all activities on it critical, and every adjacent LAG = 0 (unbroken). A path of all-critical activities with some LAG non-zero is not critical - the key difference from AOA, where "composed of critical activities" suffices.

8. Worked example (LAG method) Activity A has lags 0 and 2 to two successors whose total floats are 3 and 2:

  • FF of A = min(0, 2) = 0 (one zero LAG makes the free float 0 - A's early finish pins a successor's early start);
  • TF of A = min(0+3, 2+2) = 3;
  • check: FF <= TF (0 <= 3) holds; A's TF is not 0, so A is neither critical nor on the critical path.

Applicable boundaries

  • For AON networks without time lags (STS/FTF/STF/FTS); AON networks with overlaps follow the relevant chapters of the code.
  • The LAG method presupposes ES/EF already computed forwards with the max rule; otherwise LAG, FF and TF have no basis.
  • Total float must be computed backwards; computing forwards as you go is forbidden - all successors' TF must be known first (the virtual end node first).
  • Tp fixes the sign of the end node's TF: with Tp = Tc, TF = 0 identifies critical activities; with Tr < Tc the least (negative) total float does - "TF = 0" no longer works.
  • Where the computed duration exceeds the required duration, adjust the network (compress critical activities or change logic) before fixing Tp - float arithmetic must not paper over an unmet duration.
  • Always FF <= TF; a violation means the recursion or a LAG is wrong.

Further reading

  • JGJ/T 121-2015 clauses 5.3.5~5.3.9, 5.5.1, 5.5.2 and term 2.1.35; its commentary on the two AON procedures; AOA six-parameter computation in clause set 4.3; time-scaled direct reading in 4.4 and 4.5.

Keywords