Case Studies

Case study: Load Cell Calibration


NIST info: This example illustrates the construction of a linear regression model for load cell data that relates a known load applied to a load cell to the deflection of the cell. The model is then used to calibrate future cell readings associated with loads of unknown magnitude. Full description can be found here.

Data: The data collected in the calibration experiment consisted of a known load, applied to the load cell, and the corresponding deflection of the cell from its nominal position. Forty measurements were made over a range of loads from 150,000 to 3,000,000 units. The data were collected in two sets in order of increasing load. The systematic run order makes it difficult to determine whether or not there was any drift in the load cell or measuring equipment over time. Assuming there is no drift, however, the experiment should provide a good description of the relationship between the load applied to the cell and its response. Full dataset is available here.

In this example we will follow the suggested steps, listed at NIST pages. As suggested, we will use linear and quadratic polynomial to perform fitting and estimating.

1) Read in the data: For regression we will use Polynoms PolyFit routine. First we have to create necessary vectors and populate them.

2) Plot the y=y(x) chart to get an idea about the fit model. For this, you can use the MtxVecTee.DrawValues method. These two steps can be acomplished by the following code:

project d32

Uses MtxVec, Polynoms, Math387, MtxVecTee, MtxExpr;

procedure FitModel;
var  x,y, residuals, coeffs, ycalc: Vector;
  rss: TSample;
  nlr := TMtxNonLinReg.Create(nil);

project cpp

#include "MtxVecCpp.h"
#include "Polynoms.hpp"
#include "Math387.hpp"
#include "MtxVec.hpp"
#include "MtxVecTee.hpp"

void FitModel(void);
  Vector x,y,residuals,coeffs,ycalc;
  Chart1.AddSeries(new TLineSeries(Chart1));

project cs

using Dew.Math;
using Dew.Math.Units;
using Dew.Stats.Units;
using Dew.Stats;

namespace Dew.Tests
   private void FitModel()
      Vector x = new Vector(0);
      Vector y = new Vector(0);
      Vector residuals = new Vector(0);
      Vector ycalc = new Vector(0);
      Vector coeffs = new Vector(0);
      tChart1.Series.Add(new Steema.TeeChart.Styles.Line(tChart1.Chart));
         MtxVecTee.DrawValues(x, y, tChart1[0],0,1,false);

Plotting the data indicates that the hypothesized, simple relationship between load and deflection is reasonable. The plot indicates that a straight-line model is likely to fit the data. It does not indicate any other problems, such as presence of outliers or nonconstant standard deviation of the response.

3) Perform simple LQ linear regression on data: With just few lines of code, least squared linear regression is applied to x,y data:


// Perform linear regression
 // Calculate residuals
 rss := residuals.SumOfSquares;


// perform linear regression
 // calculate residuals
 double rss = residuals->SumOfSquares();


// Perform linear regression
 // Calculate residuals
 double rss = residuals.SumOfSquares();

Now that we have estimated y(x) values and residuals, we can use MtxVecTee.DrawValues method to check the goodnes of out model selection. Plotting several charts reveals the linear model is not a very good choice. Next we try to fit a higher order polynomial to the data (quadratic polynom). We only have to change the PolyFit(x,y,coeffs,1) to PolyFit(x,y,coeffs,2) and repeat the procedure.

Additional things to try :

  • Plot data, fitted data and residuals by using MtxVecTee.DrawValues routine.
  • Plot normal probability plot of residuals.
  • Plot histogram of residuals.
  • Compare the results with quoted results at NIST pages.

© DewResearch 1997 - 2022 All Rights Reserved.

E-mail This email address is being protected from spambots. You need JavaScript enabled to view it..
Delphi & C++ Builder are registered trademarks of Embarcadero Corporation. All other brands and product names are trademarks or registered trademarks of their respective owners.