Python and phyphox - Revision history

Mol: /* Departure WAW-TLL: Linearize the data */

2025-06-16T14:52:25Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 17:52, 16 June 2025
Line 245:		Line 245:
	ys = np.empty( np.shape(xs) )		ys = np.empty( np.shape(xs) )
	for i, time in enumerate(xs):		for i, time in enumerate(xs):
	X = df.loc[ int(time): ]		X = df_rolled.loc[ int(time): ]
	ys[i] = X.iloc[0].ax		ys[i] = X.iloc[0].ax
	res = my_pwlf.fit( np.shape( xs )[0], xs, ys )		res = my_pwlf.fit( np.shape( xs )[0], xs, ys )

2025-06-16T14:49:55Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 17:49, 16 June 2025
Line 220:		Line 220:
	[[File:Phyphox Flight waw-tll draw5d piecewiseLinearFit simple.png\|thumb\|Hand-made linear fit with no optimization. Works rather well.]]		[[File:Phyphox Flight waw-tll draw5d piecewiseLinearFit simple.png\|thumb\|Hand-made linear fit with no optimization. Works rather well.]]
	<syntaxhighlight lang="python">		<syntaxhighlight lang="python">
	df_rolled = df.ax.rolling(window=1200).mean()		df_rolled = df.ax.rolling(window=1200).mean().dropna()
	xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])		xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])
	ys = np.empty( np.shape(xs) )		ys = np.empty( np.shape(xs) )

2025-06-16T14:48:29Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 17:48, 16 June 2025
Line 206:		Line 206:
	<syntaxhighlight lang="python">		<syntaxhighlight lang="python">
	import pwlf		import pwlf
	df_rolled = df.ax.rolling(window=1200).mean()		df_rolled = df.ax.rolling(window=1200).mean().dropna()

	xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])		xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])

2025-06-16T14:42:23Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 17:42, 16 June 2025
Line 206:		Line 206:
	<syntaxhighlight lang="python">		<syntaxhighlight lang="python">
	import pwlf		import pwlf
			df_rolled = df.ax.rolling(window=1200).mean()

	xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])		xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])

	my_pwlf = pwlf.PiecewiseLinFit(df.index.values, ~~df.ax~~.values )		my_pwlf = pwlf.PiecewiseLinFit(df_rolled.index.values, df_rolled.values )
	my_pwlf.fit_with_breaks(xs)		my_pwlf.fit_with_breaks(xs)
	xHat = np.linspace(0, 160, num=10000)		xHat = np.linspace(0, 160, num=10000)

2025-06-16T14:39:00Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 17:39, 16 June 2025
Line 215:		Line 215:
	plt.plot(xHat, yHat, '-')		plt.plot(xHat, yHat, '-')
	</syntaxhighlight>		</syntaxhighlight>


			[[File:Phyphox Flight waw-tll draw5d piecewiseLinearFit simple.png\|thumb\|Hand-made linear fit with no optimization. Works rather well.]]
			<syntaxhighlight lang="python">
			df_rolled = df.ax.rolling(window=1200).mean()
			xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])
			ys = np.empty( np.shape(xs) )
			for i, time in enumerate(xs):
			X = df_rolled.loc[ int(time):]
			ys[i] = X.iloc[0]
			for i in range(1, np.shape(xs)[0]):
			plt.plot( [xs[i-1], xs[i]], [ys[i-1], ys[i]], 'r' )
			</syntaxhighlight >

2025-06-16T14:11:12Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 17:11, 16 June 2025
Line 224:		Line 224:
	res = my_pwlf.fit(8)		res = my_pwlf.fit(8)
	</code>		</code>



			'''Piecewise linear fit''' with forcing through data points does not converge. It says only <code> LinAlgWarning: Ill-conditioned matrix (rcond=7.2053e-35): result may not be accurate</code> but tries to optimize.
			<syntaxhighlight lang="python">
			ys = np.empty( np.shape(xs) )
			for i, time in enumerate(xs):
			X = df.loc[ int(time): ]
			ys[i] = X.iloc[0].ax
			res = my_pwlf.fit( np.shape( xs )[0], xs, ys )
			</syntaxhighlight>

2025-06-16T12:13:10Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 15:13, 16 June 2025
Line 216:		Line 216:
	</syntaxhighlight>		</syntaxhighlight>


			[[File:Phyphox Flight waw-tll draw5c piecewiseLinearFit fulloptimization.png\|thumb\|Automatically find the points.]]

			'''Piecewise linear fit''' with optimization, does not work very well. This should be with 8 segments but only five is seen.

			<code>
			res = my_pwlf.fit(8)
			</code>

2025-06-16T11:39:36Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 14:39, 16 June 2025
Line 197:		Line 197:
	=== Departure WAW-TLL: Linearize the data ===		=== Departure WAW-TLL: Linearize the data ===

	The acceleration data is almost linear, so lets do a small regression analysis to get a bit simpler model. The complication is that the regression model should be piecewise continuous. There are automatic methods (eg linear tree, see https://gist.github.com/ruoyu0088/70effade57483355bbd18b31dc370f2a) but we will use semi-manual system. One option would be to assign a piecewise function in numpy, and then fit the data into the function, but there is an easier solution, by using pwlf library.		The acceleration data is almost linear, so lets do a small regression analysis to get a bit simpler model. The complication is that the regression model should be piecewise continuous. There are automatic methods (eg linear tree, see https://gist.github.com/ruoyu0088/70effade57483355bbd18b31dc370f2a) but we will use semi-manual system using the <code>plwf</code> library https://jekel.me/piecewise_linear_fit_py/examples.html. One option would be to assign a piecewise function in numpy, and then fit the data into the function, but there is an easier solution, by using pwlf library.

2025-06-16T11:38:26Z

Departure WAW-TLL: Linearize the data

← Older revision		Revision as of 14:38, 16 June 2025
Line 200:		Line 200:


	[[File:Phyphox Flight waw-tll draw5b piecewiseLinearFit.png\|thumb\|Piecewise linear fit ~~with~~ <code>plwf</code> library~~. This~~ is easy to use.		[[File:Phyphox Flight waw-tll draw5b piecewiseLinearFit.png\|thumb\|Piecewise linear fit]]


			'''Piecewise linear fit''' (<code>plwf</code> library) is easy to use, but somehow it doesn't follow all the points, as is shown on the Figure.
	<syntaxhighlight lang="python">		<syntaxhighlight lang="python">
	import pwlf		import pwlf
Line 214:		Line 217:


	~~'''Piecewise linear fit'''~~

	[[File:Phyphox Flight waw-tll draw5a piecewise.png\|thumb\|Simple piecewise nut not continous function]]		[[File:Phyphox Flight waw-tll draw5a piecewise.png\|thumb\|Simple piecewise nut not continous function]]

2025-06-16T11:36:06Z

Departure WAW-TLL: Linearize the data

@@ Line 199: / Line 199: @@
 The acceleration data is almost linear, so lets do a small regression analysis to get a bit simpler model. The complication is that the regression model should be piecewise continuous. There are automatic methods (eg linear tree, see https://gist.github.com/ruoyu0088/70effade57483355bbd18b31dc370f2a) but we will use semi-manual system. One option would be to assign a piecewise function in numpy, and then fit the data into the function, but there is an easier solution, by using pwlf library.
 [[File:Phyphox Flight waw-tll draw5a piecewise.png|thumb|Simple piecewise nut not continous function]]
@@ Line 204: / Line 220: @@
 '''Piecewise''' function is not continous. This is a failed attemp. The following code assumes that the piecewise points are known. The <code>linear_model</code> needs the data in a strange format, and thus the <code>reshape</code> command is used.
 <syntaxhighlight lang="python">
 xs = np.array([0, 99, 104, 109, 112.5, 116, 117, 118.5, 120, 127, 130, 135, 145, 160])