Python 做曲线拟合和求积分的方法
这是一个由加油站油罐传感器测量的油罐高度数据和出油体积,根据体积和高度的倒数,用截面积来描述油罐形状,求出拟合曲线,再用标准数据,求积分来验证拟合曲线效果和误差的一个小项目。 主要的就是首先要安装Anaconda python库,然后来运用这些数学工具。
###最小二乘法试验### import numpy as np import pymysql from scipy.optimize import leastsq from scipy import integrate ###绘图,看拟合效果### import matplotlib.pyplot as plt from sympy import * path='E:\PythonProgram\oildata.txt' lieh0 =[] #初始第一列,油管高度 liev1 =[] #初始第二列,油枪记录的体积 h_median =[] # 高度相邻中位数 h_subtract =[] #相邻高度作差 v_subtract =[] #相邻体积作差 select_h_subtr =[] #筛选的高度作差 ######## select_v_subtr =[] #筛选的体积作差 VdtH=[] #筛选的V 和 t 的 倒数。 def loadData(Spath,lie0,lie1): with open(Spath,'r') as f0: for i in f0: tmp=i.split() lie0.append(float(tmp[0])) lie1.append(float(tmp[2])) print ("source lie0",len(lie0)) def connectMySQL(): db = pymysql.connect(host='10.**.**.**', user='root', passwd='***', db="zabbix", charset="utf8") # 校罐 cur = db.cursor() try: # 查询 cur.execute("SELECT * FROM station_snapshot limit 10 ") for row in cur.fetchall(): # print(row) id = row[0] snapshot_id = row[1] DateTime = row[13] attr1V = row[5] attr2H = row[6] print("id=%d ,snapshot_id=%s,DateTime=%s,attr1V =%s, attr2H=%s ", (id, snapshot_id, DateTime, attr1V, attr2H)) except: print("Error: unable to fecth data of station_stock") try: cur.execute("SELECT * FROM can_stock limit 5"); for row in cur.fetchall(): # print(row) stockid = row[0] stationid = row[1] DateTime = row[4] Volume = row[5] Height = row[8] print("stockid=%d ,stationid=%s,DateTime=%s,Volume =%f, Height=%f ", (stockid, stationid, DateTime, Volume, Height)) except: print("Error: unable to fecth data of can_snapshot") cur.close() db.close() def formatData(h_med,h_subtr,v_subtr): lh0 = lieh0[:] del lh0[0] print("lh0 size(): ",len(lh0)) lh1 =lieh0[:] del lh1[len(lh1)-1] print("lh1 size() : ",len(lh1)) lv0 =liev1[:] del lv0[0] #print (liev1) print ("Souce liev1 size() : ",len(liev1)) print ("lv1 size() :",len(lv0)) """ lv1 =liev1[:] del lv1[len(lv1)-1] print("lv1 size(): ",len(lv1)) """ h_med[:] =[(x+y)/2 for x,y in zip(lh0,lh1)] ###采样点(Xi,Yi)### print("h_med size() : ", len(h_med)) h_subtr[:] = [(y-x) for x,y in zip(lh0,lh1)] print("h_subtr size() : ", len(h_subtr)) # v_subtr[:] = [(y-x) for x,y in zip(lv0,lv1)] v_subtr[:] = lv0 print("v_subtr size() : ", len(v_subtr)) def removeBadPoint(h_med,h_sub,v_sub): for val in h_sub: position=h_sub.index(val) if 0.01 > val > -0.01: del h_sub[position] del h_med[position] del v_sub[position] v_dt_h_ay = [(y/x) for x, y in zip(h_sub, v_sub)] return v_dt_h_ay def selectRightPoint(h_med,h_subtr,v_dt_h_ay): for val in v_dt_h_ay: pos = v_dt_h_ay.index(val) if val > 20 : del v_dt_h_ay[pos] del h_med[pos] del h_subtr[pos] for val in v_dt_h_ay: ptr = v_dt_h_ay.index(val) if val < 14: del v_dt_h_ay[ptr] del h_med[ptr] del h_subtr[ptr] def writeFile(h_mp, v_dt_h): s='\n'.join(str(num)[1:-1] for num in h_mp) v='\n'.join(str(vdt)[1:-1] for vdt in v_dt_h) open(r'h_2.txt','w').write(s) open(r'v_dt_h.txt','w').write(v) print("write h_median: ",len(h_mp)) # print("V_dt also is (y-x) : ",v_dt_h,end="\n") print("Write V_dt_h : ",len(v_dt_h)) # file=open('data.txt','w') # file.write(str(h_mp)); # file.close def integralCalculate(coeff,xspace): vCalcute =[] x = Symbol('x') a, b, c, d = coeff[0] y = a * x ** 3 + b * x ** 2 + c * x + d i=0 while (i< len(xspace)-1) : m = integrate(y, (x, xspace[i], xspace[i+1])) vCalcute.append(abs(m)) i=i+1 print("求导结果:",vCalcute) print("求导长度 len(VCalcute): ",len(vCalcute)) return vCalcute ###需要拟合的函数func及误差error### def func(p,x): a,b,c,d=p return a*x**3+b*x**2+c*x+d #指明待拟合的函数的形状,设定的拟合函数。 def error(p,x,y): return func(p,x)-y #x、y都是列表,故返回值也是个列表 def leastSquareFitting(h_mp,v_dt_hl): p0=[1,2,6,10] #a,b,c 的假设初始值,随着迭代会越来越小 #print(error(p0,h_mp,v_dt_h,"cishu")) #目标是让error 不断减小 #s="Test the number of iteration" #试验最小二乘法函数leastsq得调用几次error函数才能找到使得均方误差之和最小的a~c Para=leastsq(error,p0,args=(h_mp,v_dt_hl)) #把error函数中除了p以外的参数打包到args中 a,b,c,d=Para[0] #leastsq 返回的第一个值是a,b,c 的求解结果,leastsq返回类型相当于c++ 中的tuple print(" a=",a," b=",b," c=",c," d=",d) plt.figure(figsize=(8,6)) plt.scatter(h_mp,v_dt_hl,color="red",label="Sample Point",linewidth=3) #画样本点 x=np.linspace(200,2200,1000) y=a*x**3+b*x**2+c*x+d integralCalculate(Para,h_subtract) plt.plot(x,y,color="orange",label="Fitting Curve",linewidth=2) #画拟合曲线 # plt.plot(h_mp, v_dt_hl,color="blue", label='Origin Line',linewidth=1) #画连接线 plt.legend() plt.show() def freeParameterFitting(h_mp,v_dt_hl): z1 = np.polyfit(h_mp, v_dt_hl, 6) # 第一个拟合,自由度为6 # 生成多项式对象 p1 = np.poly1d(z1) print("Z1:") print(z1) print("P1:") print(p1) print("\n") x = np.linspace(400, 1700, 1000) plt.plot(h_mp, v_dt_hl, color="blue", label='Origin Line', linewidth=1) # 画连接线 plt.plot(x, p1(x), 'gv--', color="black", label='Poly Fitting Line(deg=6)', linewidth=1) plt.legend() plt.show() def main(): loadData(path, lieh0, liev1) connectMySQL() # 读取oildata数据库 formatData(h_median, h_subtract, v_subtract) # 去除被除数为0对应的点,并得到v 和 h 求导 值的列表 VdtH[:] = removeBadPoint(h_median, h_subtract, v_subtract) print("h_median1:", len(h_median)) print("VdtH1 : ", len(VdtH)) # 赛选数据,去除异常点 selectRightPoint(h_median, h_subtract, VdtH) print("h_median2:", len(h_median)) print("h_subtract: ", len(h_subtract)) print("VdtH2 : ", len(VdtH)) h_mp = np.array(h_median) v_dt_h = np.array(VdtH) writeFile(h_mp, v_dt_h) # 最小二乘法作图 leastSquareFitting(h_mp, v_dt_h) # 多项式自由参数法作图 freeParameterFitting(h_mp, v_dt_h) if __name__ == '__main__': main()