louis
5888b4ccce
- 读取 D700（阀后流量）、D710（阀后压力）、D720（阀前压力）， 读用功能码 03，0~32000 线性映射到各自量程（300 SLM / 400 kPa / 300 kPa） - 输出 D730（电机位置，0~10V），写用功能码 06，行程仍为 0~1000、死区 800 - 复用 MT2 骨架，保留 get_flow/get_pressure/set_motor_position 接口， channel 参数改为直接 Modbus 地址 - config.py 新增 PLC 设置，保留 MT2AM8 设置为死代码（MT2AM8Client 类保留） - 同步更新 README.md 中的硬件接线、配置默认值与模块职责描述
423 lines
14 KiB
Python
423 lines
14 KiB
Python
"""从开环扫点 CSV 拟合 A_eff(x) 表，输出 valve_model.json。
|
||
|
||
用法：
|
||
python identify_valve.py --csv open_loop_data/open_loop_xxx.csv --out valve_model.json
|
||
python identify_valve.py --csv "open_loop_data/*.csv" --out valve_model.json --plot
|
||
python identify_valve.py --plot # 省略 --csv 时默认合并 open_loop_data/ 下全部 CSV
|
||
|
||
流程：
|
||
1. --csv 支持多个路径与 glob 模式，全部 CSV 合并使用（文件名含时间戳，按名称
|
||
排序即按实验先后合并）；省略时默认 open_loop_data/*.csv；
|
||
2. 每个 CSV 按 step_index 分组，每组取末尾 --tail 个采样点平均，得到稳态
|
||
(x, Q, P1, P2)，并按开度访问顺序标记逼近方向（上升/下降/该文件起始步）；
|
||
3. 对每个稳态点反解 A_eff = Q / (P1_abs * F(r))，同一行程 x 的多个点
|
||
（来自不同次实验）先平均再构表；
|
||
4. 按行程升序构表，做单调性检查；
|
||
5. 用 ValveModel 存成 JSON，并打印摘要（可选画诊断图，实测点按逼近方向着色）。
|
||
"""
|
||
|
||
import argparse
|
||
import csv
|
||
import glob
|
||
import math
|
||
import os
|
||
from pathlib import Path
|
||
|
||
import config
|
||
from valve_model import CRITICAL_RATIO, ValveModel, abs_pressure, f_ratio
|
||
|
||
DEFAULT_TAIL = 20
|
||
|
||
# 逼近方向：按开度访问顺序，当前开度大于/小于上一开度。
|
||
DIRECTION_UP = "up" # 从更小开度升到当前开度
|
||
DIRECTION_DOWN = "down" # 从更大开度降到当前开度
|
||
DIRECTION_ORDER = [DIRECTION_UP, DIRECTION_DOWN, None]
|
||
DIRECTION_LABELS = {
|
||
DIRECTION_UP: "ascending",
|
||
DIRECTION_DOWN: "descending",
|
||
None: "start (unknown)",
|
||
}
|
||
DIRECTION_COLORS = {
|
||
DIRECTION_UP: "#1565C0",
|
||
DIRECTION_DOWN: "#D84315",
|
||
None: "#757575",
|
||
}
|
||
|
||
|
||
def _to_float(text):
|
||
"""把 CSV 单元格解析成有限浮点数；空/非法/非有限返回 None。"""
|
||
if text is None:
|
||
return None
|
||
text = str(text).strip()
|
||
if text == "":
|
||
return None
|
||
try:
|
||
value = float(text)
|
||
except (TypeError, ValueError):
|
||
return None
|
||
return value if math.isfinite(value) else None
|
||
|
||
|
||
def _average(values):
|
||
"""返回非 None 数值的平均；没有有效值时返回 None。"""
|
||
valid = [value for value in values if value is not None]
|
||
return None if not valid else sum(valid) / len(valid)
|
||
|
||
|
||
def load_steady_points(csv_path, tail=DEFAULT_TAIL):
|
||
"""提取单个 CSV 每个 step_index 的稳态均值与逼近方向。
|
||
|
||
返回按 step（访问顺序）排列的 ``(x, q, p1, p2, direction)`` 列表。
|
||
direction 为 ``"up"``（开度上升逼近）/ ``"down"``（下降逼近）/
|
||
``None``（该文件第一步，无上一开度可比较）。
|
||
"""
|
||
rows_by_step = {}
|
||
opening_by_step = {}
|
||
with open(csv_path, "r", newline="", encoding="utf-8-sig") as file:
|
||
for row in csv.DictReader(file):
|
||
step_text = (row.get("step_index") or "").strip()
|
||
if step_text == "":
|
||
continue
|
||
try:
|
||
step = int(float(step_text))
|
||
except (TypeError, ValueError):
|
||
continue
|
||
rows_by_step.setdefault(step, []).append(row)
|
||
if step not in opening_by_step:
|
||
opening_by_step[step] = _to_float(row.get("opening_pct"))
|
||
|
||
points = []
|
||
prev_opening = None
|
||
for step in sorted(rows_by_step):
|
||
opening = opening_by_step.get(step)
|
||
if prev_opening is not None and opening is not None:
|
||
if opening > prev_opening:
|
||
direction = DIRECTION_UP
|
||
elif opening < prev_opening:
|
||
direction = DIRECTION_DOWN
|
||
else:
|
||
direction = None
|
||
else:
|
||
direction = None
|
||
if opening is not None:
|
||
prev_opening = opening
|
||
|
||
rows = rows_by_step[step]
|
||
rows.sort(key=lambda r: _to_float(r.get("time_s")) or 0.0)
|
||
tail_rows = rows[-tail:] if tail > 0 else rows
|
||
|
||
x = _average(_to_float(r.get("motor_position")) for r in tail_rows)
|
||
q = _average(_to_float(r.get("flow_after_slm")) for r in tail_rows)
|
||
p1 = _average(_to_float(r.get("pressure_before_kpa")) for r in tail_rows)
|
||
p2 = _average(_to_float(r.get("pressure_after_kpa")) for r in tail_rows)
|
||
|
||
if None in (x, q, p1, p2):
|
||
continue
|
||
points.append((x, q, p1, p2, direction))
|
||
return points
|
||
|
||
|
||
def compute_area_table(steady_points):
|
||
"""把稳态点换算成 (x, A_eff) 单调表，并返回逐点测量值供诊断图着色。
|
||
|
||
返回 ``(table, dropped, stats, area_points)``：
|
||
|
||
- ``area_points``：每个实测点的 ``(x, A_eff, direction)``，按 x 升序，未平均；
|
||
- ``table``：同一行程 x 的多个 A_eff 先平均，再按 x 升序取最长单调连续段；
|
||
- ``dropped``：因非单调被丢弃的 ``(x, A_eff)``；
|
||
- ``stats``：阻塞/亚声速点统计（按原始测量点计数）。
|
||
"""
|
||
area_points = []
|
||
choked = subsonic = 0
|
||
for x, q, p1, p2, direction in steady_points:
|
||
p1_abs = abs_pressure(p1)
|
||
p2_abs = abs_pressure(p2)
|
||
if p1_abs <= 0.0:
|
||
continue
|
||
r = p2_abs / p1_abs
|
||
f = f_ratio(r)
|
||
denom = p1_abs * f
|
||
if denom <= 0.0 or q < 0.0:
|
||
continue
|
||
if r <= CRITICAL_RATIO:
|
||
choked += 1
|
||
else:
|
||
subsonic += 1
|
||
area_points.append((x, q / denom, direction))
|
||
|
||
area_points.sort(key=lambda item: item[0])
|
||
|
||
# 合并多份 CSV 后同一行程会有多个点：先按 x 分组平均（CSV 以 6 位小数
|
||
# 记录行程，round 用于吸收极小浮点差异）。
|
||
by_x = {}
|
||
for x, a, _ in area_points:
|
||
by_x.setdefault(round(x, 6), []).append(a)
|
||
merged = [(float(x), sum(vals) / len(vals)) for x, vals in sorted(by_x.items())]
|
||
|
||
ys = [a for _, a in merged]
|
||
if _is_monotonic(ys):
|
||
table = merged
|
||
dropped = []
|
||
else:
|
||
start, end = _longest_monotonic_run(ys)
|
||
table = merged[start:end + 1]
|
||
dropped = merged[:start] + merged[end + 1:]
|
||
|
||
stats = {"choked": choked, "subsonic": subsonic, "total": choked + subsonic}
|
||
return table, dropped, stats, area_points
|
||
|
||
|
||
def _is_monotonic(ys):
|
||
"""序列是否单调（允许相等，但不允许中途反向）。"""
|
||
direction = 0
|
||
for i in range(1, len(ys)):
|
||
delta = ys[i] - ys[i - 1]
|
||
if delta == 0:
|
||
continue
|
||
sign = 1 if delta > 0 else -1
|
||
if direction == 0:
|
||
direction = sign
|
||
elif direction != sign:
|
||
return False
|
||
return True
|
||
|
||
|
||
def _longest_monotonic_run(ys):
|
||
"""返回最长单调连续段的闭区间下标 ``(start, end)``。"""
|
||
best_start = best_end = 0
|
||
for start in range(len(ys)):
|
||
direction = 0
|
||
end = start
|
||
for j in range(start + 1, len(ys)):
|
||
delta = ys[j] - ys[j - 1]
|
||
if delta == 0:
|
||
end = j
|
||
continue
|
||
sign = 1 if delta > 0 else -1
|
||
if direction == 0:
|
||
direction = sign
|
||
end = j
|
||
elif direction == sign:
|
||
end = j
|
||
else:
|
||
break
|
||
if end - start > best_end - best_start:
|
||
best_start, best_end = start, end
|
||
return best_start, best_end
|
||
|
||
|
||
def identify(csv_paths, tail=DEFAULT_TAIL, out_path=None):
|
||
"""从一份或多份开环扫点 CSV 端到端辨识。
|
||
|
||
返回 ``(model, steady_points, table, dropped, stats, area_points)``。
|
||
``csv_paths`` 为路径列表，其稳态点合并后统一拟合。
|
||
"""
|
||
steady = []
|
||
for path in csv_paths:
|
||
steady.extend(load_steady_points(path, tail))
|
||
table, dropped, stats, area_points = compute_area_table(steady)
|
||
if len(table) < 2:
|
||
raise ValueError(
|
||
f"有效稳态点不足（{len(table)} 个），无法构表。请确认 CSV 为 "
|
||
"open_loop.py 产出（需含 step_index 列），且 "
|
||
"pressure_before_kpa / pressure_after_kpa / flow_after_slm "
|
||
"列读数合理。"
|
||
)
|
||
model = ValveModel(
|
||
table,
|
||
config.MOTOR_OPEN_POSITION,
|
||
config.MOTOR_CLOSED_POSITION,
|
||
)
|
||
if out_path:
|
||
model.save(out_path)
|
||
return model, steady, table, dropped, stats, area_points
|
||
|
||
|
||
def resolve_csv_paths(items=None):
|
||
"""把 CLI 传入的路径/glob 展开为去重、按名称排序的 CSV 路径列表。
|
||
|
||
``items`` 为空时默认使用工程目录下 ``open_loop_data/*.csv``（文件名含
|
||
时间戳，排序后即按实验先后合并）。
|
||
"""
|
||
if not items:
|
||
default_dir = Path(__file__).resolve().parent / "open_loop_data"
|
||
items = [str(default_dir / "*.csv")]
|
||
paths = []
|
||
for item in items:
|
||
pattern = str(item).replace("\\", "/")
|
||
if glob.has_magic(pattern):
|
||
paths.extend(sorted(glob.glob(pattern)))
|
||
else:
|
||
paths.append(pattern)
|
||
# 去重：glob 结果与显式路径的斜杠方向可能不同，用 normcase 归一化比较
|
||
# （Windows 下同时统一大小写）。
|
||
unique = []
|
||
seen = set()
|
||
for path in paths:
|
||
key = os.path.normcase(path)
|
||
if key not in seen:
|
||
seen.add(key)
|
||
unique.append(path)
|
||
return unique
|
||
|
||
|
||
def _count_directions(steady_points):
|
||
counts = {DIRECTION_UP: 0, DIRECTION_DOWN: 0, None: 0}
|
||
for point in steady_points:
|
||
counts[point[4]] += 1
|
||
return counts
|
||
|
||
|
||
def _scatter_by_direction(axis, points):
|
||
"""按逼近方向分色画散点；``points`` 为 ``(x, y, direction)`` 列表。"""
|
||
by_direction = {d: [] for d in DIRECTION_ORDER}
|
||
for x, y, direction in points:
|
||
by_direction.setdefault(direction, []).append((x, y))
|
||
for direction in DIRECTION_ORDER:
|
||
pts = by_direction.get(direction, [])
|
||
if not pts:
|
||
continue
|
||
axis.scatter(
|
||
[p[0] for p in pts],
|
||
[p[1] for p in pts],
|
||
color=DIRECTION_COLORS[direction],
|
||
label=DIRECTION_LABELS[direction],
|
||
s=18,
|
||
alpha=0.85,
|
||
)
|
||
|
||
|
||
def create_diagnostic_plot(steady_points, table, dropped, area_points, image_path):
|
||
"""画 x vs A_eff 与 x vs Q_ss 两张子图，实测点按逼近方向着色。
|
||
|
||
同一行程下上升/下降两团点明显分离，说明阀门迟滞显著，前馈表应考虑
|
||
分方向处理；完全重叠则迟滞可忽略。
|
||
"""
|
||
import matplotlib
|
||
|
||
matplotlib.use("Agg")
|
||
import matplotlib.pyplot as plt
|
||
|
||
xs = [x for x, _ in table]
|
||
areas = [a for _, a in table]
|
||
|
||
figure, axes = plt.subplots(2, 1, figsize=(10, 8), constrained_layout=True)
|
||
figure.suptitle("Identified valve characteristic")
|
||
|
||
axes[0].plot(xs, areas, color="#263238", linewidth=1.2, label="fitted A_eff (per-x mean)")
|
||
_scatter_by_direction(axes[0], area_points)
|
||
if dropped:
|
||
axes[0].scatter(
|
||
[x for x, _ in dropped],
|
||
[a for _, a in dropped],
|
||
color="#B71C1C",
|
||
marker="x",
|
||
label="dropped (non-monotonic)",
|
||
)
|
||
axes[0].set_ylabel("A_eff (slm/kPa)")
|
||
axes[0].set_title("Effective area vs stroke")
|
||
axes[0].legend(loc="best")
|
||
axes[0].grid(True, alpha=0.3, linestyle="--")
|
||
|
||
_scatter_by_direction(
|
||
axes[1], [(p[0], p[1], p[4]) for p in steady_points]
|
||
)
|
||
axes[1].set_ylabel("Flow (SLM)")
|
||
axes[1].set_xlabel("Motor stroke x")
|
||
axes[1].set_title("Steady flow vs stroke")
|
||
axes[1].legend(loc="best")
|
||
axes[1].grid(True, alpha=0.3, linestyle="--")
|
||
|
||
figure.savefig(image_path, dpi=config.PLOT_DPI, bbox_inches="tight")
|
||
plt.close(figure)
|
||
|
||
|
||
def parse_args(argv=None):
|
||
parser = argparse.ArgumentParser(
|
||
description="从开环扫点 CSV 拟合 A_eff(x) 阀特性表（支持多 CSV/glob 合并）"
|
||
)
|
||
parser.add_argument(
|
||
"--csv", nargs="+",
|
||
help="一个或多个 CSV 路径或 glob 模式；省略时默认合并 open_loop_data/*.csv",
|
||
)
|
||
parser.add_argument("--out", default="valve_model.json", help="输出 JSON 路径")
|
||
parser.add_argument(
|
||
"--tail", type=int, default=DEFAULT_TAIL,
|
||
help=f"每组末尾用于平均的采样点数（默认 {DEFAULT_TAIL}）",
|
||
)
|
||
parser.add_argument("--plot", action="store_true", help="生成诊断图 PNG")
|
||
args = parser.parse_args(argv)
|
||
if args.tail < 1:
|
||
parser.error("--tail 必须 >= 1")
|
||
return args
|
||
|
||
|
||
def main(argv=None):
|
||
args = parse_args(argv)
|
||
csv_paths = resolve_csv_paths(args.csv)
|
||
if not csv_paths:
|
||
print("未找到任何 CSV：请确认 open_loop_data/ 下有开环扫点数据，")
|
||
print('或用 --csv 指定路径或 glob 模式（如 --csv "open_loop_data/*.csv"）。')
|
||
return 2
|
||
for path in csv_paths:
|
||
if not Path(path).exists():
|
||
print(f"CSV 不存在：{path}")
|
||
return 2
|
||
|
||
try:
|
||
model, steady, table, dropped, stats, area_points = identify(
|
||
csv_paths, tail=args.tail, out_path=args.out
|
||
)
|
||
except (ValueError, OSError) as exc:
|
||
print(f"辨识失败：{exc}")
|
||
return 2
|
||
|
||
if len(csv_paths) <= 8:
|
||
print("合并的 CSV：")
|
||
for path in csv_paths:
|
||
print(f" {path}")
|
||
else:
|
||
print(f"合并的 CSV：{csv_paths[0]} 等共 {len(csv_paths)} 个文件")
|
||
directions = _count_directions(steady)
|
||
print(
|
||
f"读取稳态点 {len(steady)} 个：上升逼近 {directions[DIRECTION_UP]}、"
|
||
f"下降逼近 {directions[DIRECTION_DOWN]}、起始点 {directions[None]}；"
|
||
f"阻塞 {stats['choked']} 个，亚声速 {stats['subsonic']} 个，"
|
||
f"合计 {stats['total']} 个。"
|
||
)
|
||
print(
|
||
f"有效行程范围 {table[0][0]:.1f} ~ {table[-1][0]:.1f}"
|
||
f"（{len(table)} 点，同一行程已平均）。"
|
||
)
|
||
|
||
if dropped:
|
||
print(
|
||
"警告：A_eff(x) 非单调，以下点被丢弃"
|
||
"（建议缩小扫点范围到单调段重扫）。"
|
||
)
|
||
print("行程 x -> A_eff：")
|
||
for x, a in table:
|
||
print(f" x={x:8.3f} A_eff={a:.6f}")
|
||
if dropped:
|
||
print("被丢弃的点：")
|
||
for x, a in dropped:
|
||
print(f" x={x:8.3f} A_eff={a:.6f} [DROPPED]")
|
||
|
||
print(f"已保存阀特性模型：{args.out}")
|
||
|
||
if args.plot:
|
||
image_path = Path(args.out).with_suffix(".png")
|
||
try:
|
||
create_diagnostic_plot(steady, table, dropped, area_points, image_path)
|
||
print(f"诊断图已保存：{image_path}")
|
||
except Exception as exc:
|
||
print(f"生成诊断图失败（不影响 JSON）：{exc}")
|
||
|
||
return 0
|
||
|
||
|
||
if __name__ == "__main__":
|
||
import sys
|
||
|
||
sys.exit(main())
|