# Understanding Basic Principles Optimization Regulation and ... Understanding Basic Principles...

date post

12-Aug-2020Category

## Documents

view

1download

0

Embed Size (px)

### Transcript of Understanding Basic Principles Optimization Regulation and ... Understanding Basic Principles...

1

Understanding Basic Principles

Optimization

Regulation and Adjustment

The Beginners Guide

In the beginning there is balance

2

Preface This booklet is designed to serve as an easy to understand and digest guideline to adjustment and optimization technique. The basic (mechanical) principles of balance or generalized equilibration are described starting with simple means. It is our target to achieve transparency and ease of understanding. We shall not concentrate on sophisticated scientific explanations In the beginning there is balance (equilibrium) This almost philosophical statement shall be found in some basic attitudes of modern adjustment and optimization technology. Its application is reflected within the inverse balancing that comes along as a strategy to judge and evaluate parameter estimation with so called indirect observation We shall explain and investigate

- the calculation of means (balanced position) - the generalized calculation of means and their different targets - the inverse balanced position and the necessary re-weighting - visible and hidden restrictions or crisp restrictions

Within adjustment and optimization there is a strong impact of statistical concepts for parameter estimation and determination We shall provide a variety of tools that supplement each other in order to serve most appropriate and reliable results that are significant for the use of data evaluation within Manufacturing Excellence Control (MEC). Georg Gottfried Kampmann Stockholm - February 13, 2009

3

Content

1 Basics about different means and targets (4-24) 1.1 About the mechanical background (balance) of the simple arithmetic mean 1.2 The data evaluation strategy from different means 1.3 The multiple method approach (for means) 1.4 Different target function for corrections

2 Generalisation of common application of different means (25-32) 2.1 Extension of the means 2.2 Introduction to leverages, restrictions and balancing

3 Inverse balancing and its shape, the Inner Reference (33-39)

4 Additional (40-46)

- Terms and notation to judge data evaluation results - Recommended Literature

4

1 Basics about different means and targets

1.1 About the mechanical background (balance) of the simple arithmetic mean

Consider some mechanical guys taking time to determine the actual width of a vise.

Initially they use a simple ruler as seen on the picture above. And – as a matter of fact –

these guys start a discussion how to make it best.

Idea: Get an impression about the mechanical properties of the ruler. A screw driver is attached to the vise to build an apex. Its top is taken to balance the ruler and – no surprise – there is a single location (almost exactly in the middle of the ruler body) that

comes along with balance.

5

Idea: How does this peak point on the ruler change if some equal weights are placed arbitrarily on the ruler surface?

The mechanics immediately recognize a real change in the balancing position of the ruler. And pretty soon they realize they do not have to test out this individual point of balance, they may determine the screwdrivers location numerically by just adding the weights positions (location on the ruler) and dividing the result by the number of the weights.

This is the so called arithmetic mean AM which is well known. It is simple to calculate, for example AM = (l1 + l2 + l3 + l4) / 4 In this simple formula the l1 , l2 , l3 , l4 denote the individual positions of the sphere weights on the ruler and 4 denotes their number (you may recognize 4 sphere weights on the picture).

The mechanics repeat their experiments using the arithmetic mean several times and they realize it is correct unless some “mistakes” are made namely

a) Miscalculating of the AM b) Misreading the spheres position

From all the different experiments they even recognize another interesting fact, namely

the repetition number. This term comes from placing multiple sphere weights at the same ruler location as can bee seen from the picture below (three identical sphere weights almost at the same position l2)

6

From the computation of the AM there is no difference. Consider that the multiple position l2 can be expressed as three different portions l2A, l2B and l2C then the AM for the peak balance position can be expressed as AM = (l1 + l2A + l2B + l2C + l3 + l4) / 6 The mechanics quickly realize the ordinary arithmetic mean might be extended to the

arithmetic mean with repetitions, or generally the weighted arithmetic mean WAM. Consider the individual 4 positions (l) attached to their repetition number (p), say (l1 and p1 = 1) , (l2 and p2 = 3) , (l3 and p3 = 1) , (l4 and p4 = 1)

In this case the weighted arithmetic mean WAM can be expressed as WAM = (l1 p1 + l2 p2 + l3 p3 + l4 p4) / (p1 + p2 + p3 + p4) This is nothing more than a numerical extension of the simple arithmetic mean. Nevertheless, this formula can still be used in the case of non integer weights (or repetition numbers). Repetition numbers in terms of equilibration or balance may be regarded as weights – hence may take any real positive number. They have to be positive since negative weights are not defined.

Summary: We have recognized the close relationship between the balanced position (center of gravity) and the arithmetic mean (weighted or equally weighted). When data evaluation comes to apply arithmetic means (or its generalization) there is a straight

correspondence and relationship towards mechanics, a so called duality. We shall not loose this attitude when working within data evaluation.

7

1.2 The data evaluation strategy from different means After having some discussion about center of gravity and its relationship towards the arithmetic mean the mechanic guys start to determine the width of the vise. What they already feel from their previous experiments

a) to determine the width of the vise there should be repeated measurements to account for the possibility that one was incorrect (wrong determination)

b) involve different persons (independent measurements) to avoid repeated errors – this comes along with the statistical term “no correlations” or independent observations

Five different persons now provide a single measurement result of the vises width using a caliper rule as the measurement device.

No.

1

2

3

4

5

Size

8.12

8.13

8.14

8.14

8.15

Figure: Series of five (direct) observations from with caliper rule The result from the simple arithmetic mean is

AM = (8.12+8.13+8.14+8.14+8.15) / 5 = 8.136

8

This value of the AM being derived from a sample of size 5 (number of observations) is

called parameter estimation. The parameter to be estimated is the width of the vise and the estimation technique is the (weighted) arithmetic mean.

Each measurements (observations) deviation from the AM is called correction and the numerical computation is quite easy to provide

Correction = AM - measurement Generally

Correction = Parameter (Estimation) Result - Observation

For example: first correction to the first observation yields 8.136 – 8.12 = 0.016. Hence the following results for the corrections are computed:

No.

1

2

3

4

5

Size

0.016

0.006

-0.004

-0.004

-0.014

Figure: Corrections from the AM caliper rule Remark: Summing up all computed corrections from the simple arithmetic mean

MUST be 0.0. This fact derived from theory may be consequently applied to check the proper computation of AM and the related corrections.

Remark: Often the corrections are called residuals, derived from the Latin word for remnant or leftover. Now the mechanic guys chose another device to determine the width of the vise. Instead of the calliper rule they apply a micrometer to “enhance the accuracy”. We shall see what accuracy is meant to be. It is plain to see: accuracy is related to

the choice of the measurement device and the number of measurements being executed. Both facts have to be taken into account when achieving pre formulated accuracies of the parameter estimation (width of the vise).

9

The picture shows the calliper rule, the micrometer and two different rulers. Here are the results from the measurements with the micrometer:

No.

1

2

3

4

5

Size

8.134

8.135

8.135

8.137

8.138

Figure: Series of five (direct) observations micrometer The result from the simple arithmetic mean of the data is

AM = (8.134+8.135+8.135+8.137+8.138) / 5 = 8.1358 And the correct

*View more*