DEAP User Guide and Tutorial
Overview
The design of experiments to identify improvements in the performance of a product that is assembled from manufactured components does not readily fit into conventional methods of designing experiments. The DEAP software allows experiments to be designed for these problems where some, or all, of the factors of interest are a) not easily set to prescribed values and b) dependent on a combination of properties of several components, and where components cannot be re-used in an experiment. The methodology involves taking a sample of each type of component, measuring the relevant features and then finding a design that specifies a set of assembled products for experiment that gives the most information on the factors of interest. These experiments are called semi-controlled. In an experiment, a response is measured that is related to product performance for each assembled product, where it is assumed that a low order polynomial model is a reasonable description of how the factors influence the response. The results may be analysed using standard regression analysis.
Designs for two types of semi-controlled experiment are available through DEAP.
· Batch experiments, where all the components required for an experiment are available at one time and the experiment can be completed in a single stage.
· Sequential experiments, where only a few components can be acquired at one time and the experiment is executed by building one product at a time. A further product is not built until the used components have been replaced from production. Sequential experiments are useful for products that are produced in small numbers and where it is essential to minimise interruption to the manufacturing process.
Designs for experiments that involve factors of some, or all of, the following three types may be found using DEAP.
· A measured parameter of a component, for example, the length of a part.
· A derived factor, formed as a function of two or more measured parameters. Two kinds of derived factor are available:
a) a linear combination of two or more measured parameters.
b) a user-defined special function for more complicated combinations of measured parameters.
· Preset factors: conventional factors which can be set in an experiment to either 2 or 3 levels.
The structure of DEAP is in four parts:
· Product Management, where the product to be considered is defined. Here the various types of component and their parameters are entered into the system. The product for experimentation is defined by selecting components from those listed. The experiment is then designed for that particular product. The structure of DEAP allows information on many different components to be stored, that are relevant to a variety of products. A particular product for investigation is then defined, through the software, in terms of those components to be investigated in an experiment.
· Component Management, where a unique identifier or serial number is assigned to each individual measured component, together with the parameter measurements on that component. As designs are developed, the availability of these components is monitored and can be viewed at any time.
· Design of Experiment Management, where the experiment designs are produced via an appropriate search algorithm and their properties are examined. The search criterion used to find designs for a batch experiment is D-optimality. For the sequential experiments a space-filling criterion is used for the initial runs and a switch to D-optimality is made when the size of the experiment has reached the same size as the number of terms in the model.
· Design exploration and management, where the properties of the competing designs found from the algorithm can be examined.
This document gives a detailed description of the use of the screens that allow the above functionalities. Example inputs will explain the use of the screens. These example inputs will be shown in grey boxes. Tips are shown in yellow boxes.
A tutorial takes the user through a simple example of a batch experiment in Section 9 and in Section 10 a simple sequential experiment is detailed. These examples are based on a larger experiment for a sound capsule which is presented in Anthony et al (2003). The full data is available in the file HBL.exp and can be imported into DEAP through the Open dialog off the File menu. Two other example files, SS1.exp and SS3.exp, are provided. These relate to experiments on hydraulic gear pumps, the details of which are given in Sexton, Lewis and Please (2001). SS1 is for a small experiment of 11 pumps to investigate three derived factors and SS3 for a larger experiment of 44 pumps, investigating four derived factors, three pre-set factors. An advanced batch experiment tutorial is provided which goes through the SS3 example in detail and illustrates the more advanced features of DEAP. The data in all three of the example files has been scaled for reasons of confidentiality.
Authors
D.P. Dupplaw, C.J. Sexton, S.M. Lewis and C.P.Please
of the University of Southampton
Acknowledgements
This research was an outcome of the work undertaken on the Engineering and Physical Sciences Research Council grant GR/N16754. The project received financial and technical support from Jaguar Cars, Hosiden Besson and Goodrich Engine Control Systems.
Contents
Overview 1
1. Main overview screen 5
2. Product Management 7
1 2.1 Define Component List 7
2 2.2 Define Component Parameters 8
3 2.3 Define Product and Parts List 10
4 2.4 Define Manufacturing Processes 12
5 2.5 Define Process Parameters 12
3. Component Management 13
6 3.1 Add Components 13
7 3.2 Input Measurements 14
8 3.3 Explore Measurements 15
9 3.4 Update Usage Stats 22
4. Experiment Management 23
10 4.1 Setup Factors: measured parameter, derived factor or preset factor 23
11 4.2 Explore Factor Data 28
12 4.3 Define Models 29
13 4.4 Other Experiment Definitions 31
14 4.5 Summarise and Run 33
5. Manage Built Products 37
6. Results Exploration and Management 40
7. Creating Plugins for Special Functions 40
15 7.1 The Special Function Plugin Interface 40
16 7.2 Creating a Special Function (Visual C++) 42
8. References 47
9. Tutorial 1: Example of Batch Experiment 48
17 9.1 Introduction 48
18 9.2 Product Management 49
19 9.3 Component Management 53
20 9.4 Design of Experiment Management 56
10. DEAP: Sequential Experiments 66
10.1 Introduction 66
10.2 Product Management and Component Management. 66
10.3 Design of Experiment Management 67
10.4 The Sequentially Designed Experiment 69
10.5 Incrementing the Sequential Experiment Design 74
10.6 Stopping the sequential experiment. 78
1. Main overview screen
On activating the software, you are presented with a blank window as shown in Figure 1 . This is how the main screen appears when there is no project in memory.
Figure 1 - DEAP Software Main Screen when no project is loaded into memory.
To open a project that already exists, use the Open dialog off the File menu or, where available, the previous file listed at the bottom of the File menu, as shown in Figure 2.
Figure 2 - Open an Existing Project
To begin a new project, select New from the File menu, as shown in Figure 3. There are buttons in the toolbar along the top of the window which are shortcuts to these functions.
Figure 3 - Create a New Project
When you create a new project, the main screen will show the current state of that project. A new project will look like the screen shown in Figure 4.
Figure 4 - The Main Screen for a New Project
There are three sections to the main screen – Product Management, Component Management and Experiment Management. These sections allow you to input the information required for the experiment planning software to run correctly. The sections are designed to be traversed in “newspaper reading” order – that is, if you were to visit each screen (by pressing the buttons) in each column from top to bottom, and left to right by column, you will end up with a complete dataset. Each of these sections in described in the following sections of this manual.
2. Product Management
The product management section on the main screen allows you to set up the product with which you wish to experiment. It is used to define which types of components and manufacturing processes are used to make particular types of products.
The statistics at the top of the product management section show information about the data that has been entered. At the top of a new project the values will all read zero, as in Figure 4, and some of the buttons will be unavailable (“greyed-out”) because in order to use the functions which they launch, other information must first be entered. For example, on a new project the button “Define Component Parameters” is unavailable until at least one component has been defined.
1 2.1 Define Component List
Figure 5 - Define Component List Dialog
This screen is used to define the types of components from which a product for an experiment may be created. A name for a component is entered in the text box on the right, and the Add Component button allows you to confirm that entry as a component.
Tip: As soon as you type in the text box the Add Component button becomes the default button meaning you can just type a name and hit Enter to confirm it as a new component.
Selecting a component from the list on the left-hand side, will activate the Delete Selected Component button, which, when pressed, will remove the component from the list, and it will therefore not be available from other screens.
Tip: To rename a component you can click twice on the component name, in the list on the left-hand side (double click will not work – click twice slowly).
To complete the entry of your components, click the OK button to dismiss the dialog box.
Example: In this example we have entered three of the components required to make a gear pump – a bearing, a drive gear and a driven gear.
2 2.2 Define Component Parameters
Figure 6 - Define Component Parameter Dialog
This screen is designed to allow easy input of the parameters associated with each of the components that have been entered in the Define Components dialog. These components are shown in the list down the left-hand side, and are also available in the drop down Component list on the right-hand side. To select a component to add a parameter to, either click on it in the list on the left-hand side, or select it from the drop down Component list. Enter a name for the parameter in the Parameter Name text box.
Parameters can be of two types: preset or measured. If a parameter is preset, then you must be able to make components with specified values of that parameter for the experiment. If it is not preset, the parameter value is whatever measurement it happens to be for the individual component that is taken from production. A parameter can be designated as preset by ticking the preset checkbox. If a parameter is not preset, the range of values expected for that parameter can be entered in the Value Range text box. In Figure 6 the parameter gear width on the drive gear component has been entered and the measured parameter shaft diameter on the drive gear is about to be entered.
Figure 7 - Preset Level Definitions
In the case of preset parameters, values are entered in the Levels list which defines the values of the parameter which you are to use in an experiment. The Level Definition text box becomes available when the Preset checkbox is ticked. Type into the Level Definition text box the level which you would like to add, and click the Add Level button. The Clear Levels button will clear all of the levels in the list. See Figure 7, where the preset factor, bore diameter, on the bearing block component, is being entered with three levels.
Tip: To copy the data from a previously entered parameter, select the parameter from the list on the left-hand side and change its name, and, if necessary, select the component with to which itis will be associated in the entry boxes on the right-hand side. Click the Add Parameter button to add it as a new parameter. This will copy all the data.
To delete a parameter, select the parameter from the list on the left-hand side. This will activate the Delete Selected Parameter or Component button underneath the list. Click this button to remove the selected parameter.
Tip: You can also delete the whole component this way, rather than returning to the Define Component screen.
To alter the attributes of a parameter, select the parameter from the list on the left-hand side and its attributes will appear in the entry boxes on the right. Alter the incorrect value, and click the Update Parameter button which will now be available.
Tip: To change the name of a parameter, you can simple click it twice in the list on the left-hand side (don’t double click).
Example: We want to experiment which considers the clearance of the gear shafts in the bearing bores of a pump. The gears cannot be made with the shafts to exact pre-specified diameters. The shafts can be accurately measured and therefore they are entered as measured parameters with a range of 17.425-17.435mm. The bearing bore diameter can be precisely machined to the specified values so is entered as a preset factor with 3 levels, 17.5mm, 17.6mm, 17.7mm.
3 2.3 Define Product and Parts List
Figure 8 - Define Product Dialog
The Define Product dialog box allows you to see which products for experimenting have been defined from the components you have entered, and which parts each product uses(see Figure 8). When you initially enter this dialog box with a new project, it will be empty). Type in the name of the item you are constructing in the Product Name text box and click the Add New Product button. Here the product name is given as “pump”. Another dialog will appear, similar to that in Figure 9.