Compressor/backend/indicator_features/make_trend_dashboard.py

#!/usr/bin/env python3
"""Render actual sampled waveforms for selected events without changing the original page."""
from __future__ import annotations

import argparse
import csv
from datetime import datetime, timedelta
from pathlib import Path

import numpy as np

from make_event_dashboard import html
from extract_from_db import connect, fetch_cycles
from indicator_features.indicator_features import (
    CYLINDER_BORE_MM, _plateau_levels, smooth, volume_from_angle,
)
from new_unit_events import collect_events

EVENTS = [
    ("7号 2026-05-21", "7", "2026-05-21", "高密度特征_7号_2026.csv", 30, 30),
    ("8号 2026-03-29/04-01 合并事件", "8", "2026-03-30", "高密度特征_8号_2026.csv", 30, 30),
    ("8号 2025同期（后段延长至60天）", "8", "2025-03-30", "高密度特征_8号_2025同期.csv", 30, 60),
    ("7号 2025-08-08（7月前段 / 11月后段）", "7", "2025-08-08", "高密度特征_7号_2025气阀事件.csv", 30, 110),
    ("7号 2025-10-07（7月前段 / 11月后段）", "7", "2025-10-07", "高密度特征_7号_2025气阀事件.csv", 90, 50),
    ("7号 2025停机维修合并过程（4~7月回溯 / 11月恢复）", "7", "2025-07-28", "高密度特征_7号_2025维修前回溯.csv", 120, 120),
]
BAD = {("7", "四", "盖侧"), ("7", "六", "盖侧"), ("8", "六", "盖侧")}


def build_data(root: Path):
    bins = 90
    data = {"angle": (np.arange(bins) * 360 / bins).tolist(), "machines": {}}
    connection = connect()
    cycle_cache = {}
    try:
        for label, unit, date, filename, pre_days, post_days in EVENTS:
            t0 = datetime.fromisoformat(date)
            with (root / filename).open(encoding="utf-8-sig", newline="") as handle:
                rows = list(csv.DictReader(handle))
            if filename in ("高密度特征_7号_2025维修前回溯.csv", "高密度特征_7号_2025气阀事件.csv"):
                with (root / "高密度特征_7号_2025事件后11月.csv").open(encoding="utf-8-sig", newline="") as handle:
                    rows.extend(csv.DictReader(handle))
            rows = [r for r in rows if
                    t0 - timedelta(days=pre_days) <= datetime.fromisoformat(r["sample_time"]) < t0 + timedelta(days=post_days + 1)
                    and float(r.get("p_span") or 0) >= 0.3
                    and (unit, r["cylinder"], r["side"]) not in BAD]
            points = sorted({(r["cylinder"], r["side"]) for r in rows})
            # Select one cylinder per event. A dropdown exposes both chambers without
            # presenting unrelated cylinders as peers or duplicate fault samples.
            cylinder = "一" if any(c == "一" for c, _ in points) else "三"
            chosen = [r for r in rows if r["cylinder"] == cylinder]
            if filename == "高密度特征_7号_2025维修前回溯.csv":
                chosen = [r for r in rows if (r["cylinder"], r["side"]) in
                          {("三", "盖侧"), ("三", "轴侧"), ("四", "轴侧"), ("五", "盖侧") }]
            ids = [int(r["wave_file_id"]) for r in chosen]
            missing = [i for i in ids if i not in cycle_cache]
            if missing:
                cycle_cache.update(fetch_cycles(connection, missing, 100))
            event = {"label": label, "sides": {}}
            for cylinder, side in sorted({(r["cylinder"], r["side"]) for r in chosen}):
                samples = []
                for row in chosen:
                    if (row["cylinder"], row["side"]) != (cylinder, side):
                        continue
                    signal = cycle_cache.get(int(row["wave_file_id"]))
                    if signal is None or len(signal) < 60 or not np.all(np.isfinite(signal)):
                        continue
                    p = smooth(signal, 21)
                    suction, discharge, _ = _plateau_levels(p)
                    span = discharge - suction
                    if span < 0.3:
                        continue
                    indices = np.arange(bins) * len(p) / bins
                    absolute = np.interp(indices, np.arange(len(p)), p)
                    profile = (absolute - suction) / span
                    volume = np.interp(indices, np.arange(len(p)), volume_from_angle(len(p), CYLINDER_BORE_MM[cylinder]))
                    time = datetime.fromisoformat(row["sample_time"])
                    samples.append({
                        "date": row["sample_time"], "day": (time.date() - t0.date()).days,
                        "phase": "修前" if time < t0 else "修后",
                        "profile": profile.round(6).tolist(), "absolute": absolute.round(6).tolist(),
                        "volume": volume.round(6).tolist(),
                    })
                samples.sort(key=lambda s: s["date"])
                if not samples:
                    continue
                pre = [s for s in samples if s["phase"] == "修前"]
                post = [s for s in samples if s["phase"] == "修后"]
                reference = pre or post
                if filename == "高密度特征_7号_2025维修前回溯.csv":
                    reference = [s for s in samples if s["date"][:7] == "2025-04"] or reference
                template = np.median([s["profile"] for s in reference], axis=0)
                for sample in samples:
                    sample["distance"] = round(float(np.linalg.norm(np.asarray(sample["profile"]) - template)), 5)

                def median(items, key):
                    return np.median([s[key] for s in items], axis=0).round(6).tolist() if items else []

                event["sides"][f"{cylinder}缸{side}"] = {
                    "samples": samples, "counts": {"pre": len(pre), "post": len(post)},
                    "preMedian": median(pre, "profile"), "postMedian": median(post, "profile"),
                    "preAbsoluteMedian": median(pre, "absolute"), "postAbsoluteMedian": median(post, "absolute"),
                    "trend": [{"date": s["date"], "day": s["day"], "distance": s["distance"]} for s in samples],
                }
                if filename == "高密度特征_7号_2025维修前回溯.csv":
                    # Compare the last running period with restart, but measure the
                    # lookback trend against the earliest available April period.
                    july = [s for s in samples if s["date"][:7] == "2025-07"]
                    event["sides"][f"{cylinder}缸{side}"]["preMedian"] = median(july, "profile")
                    event["sides"][f"{cylinder}缸{side}"]["preAbsoluteMedian"] = median(july, "absolute")
            coverage = sorted({s["date"][:10] for d in event["sides"].values() for s in d["samples"]})
            if coverage:
                event["label"] += f" · 实际功图 {coverage[0]} 至 {coverage[-1]}"
            data["machines"][label] = event
        data["machines"].update(collect_events(connection, per_day=6, bins=bins))
    finally:
        connection.close()
    return data


def make_html(data):
    page = html(data)
    # Reuse the style and actual-waveform plots without editing the source dashboard.
    page = page.replace("一缸 P-V 功图事件前后对比", "事件功图扩展对比")
    page = page.replace("图例在图外，可逐时间点勾选", "图例在图外，按四类勾选")
    page = page.replace("P-V图例可按时间点单独勾选", "功图按前段/后段及中位线分组勾选")
    page = page.replace("<div id=\"app\"></div>",
                        '<label>选择事件：<select id="event-select"></select></label>'
                        '<p class="note">每次只展示一个选项。两个旧选项保留7月与11月的三缸两侧对照，'
                        '使用7月模板计算形状距离；10月选项的前段实际也是7月，不是10月采样。'
                        '两个旧选项共用同一批周期，仅横轴相对日期不同，不能作为独立维修证据。'
                        '7号2025年的维修合并为一次停机过程，7月28日仅作末次运行后的时间分界，不是维修日期。'
                        '该过程粗蓝线为7月末运行段中位，红线为11月恢复段中位；细蓝线包含4~7月回溯周期。'
                        '回溯选项的形状距离以4月模板为参照，其他选项仍以事件前模板为参照，数值不可混比。'
                        '新增4号和6号事件先扫描全部有效测点；本页保留全部测点，候选排名按扣除全机共同变化后的局部残差计算。'
                        '6号4月22日至26日视为同一维修过程，过程内周期不进入前后模板。'
                        '5号因当前数据库没有压力功图而暂不展示。展示测点并非已确认异常缸，缺测期间不推断变化发生时刻。</p>'
                        '<div id="app"></div>')
    page = page.replace("<script>renderPage(DATA);</script>", """<script>
renderPage(DATA);
const sections=Array.from(document.querySelectorAll('.machine'));
const selector=document.getElementById('event-select');
Object.keys(DATA.machines).forEach((name,index)=>{const option=document.createElement('option');option.value=index;option.textContent=name;selector.append(option)});
function selectEvent(){sections.forEach((section,index)=>section.hidden=index!==Number(selector.value));requestAnimationFrame(()=>charts.forEach(chart=>chart.resize()))}
selector.addEventListener('change',selectEvent);selectEvent();
</script>""")
    return page


def main():
    parser = argparse.ArgumentParser(description="生成实际周期的事件功图扩展网页")
    parser.add_argument("--root", default="cache/indicator")
    parser.add_argument("--out", required=True)
    args = parser.parse_args()
    data = build_data(Path(args.root))
    out = Path(args.out)
    out.parent.mkdir(parents=True, exist_ok=True)
    out.write_text(make_html(data), encoding="utf-8")
    print(f"对比网页 -> {out}")
    for event in data["machines"].values():
        print(event["label"], {point: d["counts"] for point, d in event["sides"].items()})


if __name__ == "__main__":
    main()