Initial commit

A题/数值分析检验/Prompt/6.md

@@ -0,0 +1,33 @@
+TASK: Run convergence/robustness checks for baseline scenario.
+
+INPUT DATA:
+- Same configuration as Prompt 5, but run two numerics:
+  A) dt = 1.0 s
+  B) dt = 0.5 s
+- Use identical params and scenario.
+
+METHODOLOGY:
+For each z0:
+1) Simulate with dt and dt/2 until termination.
+2) Compare z(t) by resampling dt/2 solution at dt grid (take every 2nd sample).
+3) Compute:
+   z_diff_inf = max_k |z_dt[k] - z_dt2[2k]|
+   tte_rel_err = |TTE_dt - TTE_dt2| / TTE_dt2
+4) Event-location robustness:
+   For each run, report the last two bracketing samples for the triggering event and the interpolated t*.
+
+DELIVERABLES:
+A) “STEP_HALVING_TABLE_v1” CSV:
+   z0, z_diff_inf, tte_rel_err, pass_bool
+B) “EVENT_BRACKET_REPORT_v1” JSON:
+   for each z0: {reason, (t_k-1, g_k-1), (t_k, g_k), t_star}
+C) Single line verdict:
+   "ROBUSTNESS_PASS" or "ROBUSTNESS_FAIL"
+
+VALIDATION (hard thresholds):
+- z_diff_inf < 1e-4
+- tte_rel_err < 0.01
+All z0 must pass or verdict is FAIL.
+
+OUTPUT FORMAT:
+CSV, then JSON, then verdict line. No prose.