Session 6.1: Sensor Library

Session 6 takes place in the PHM Module. Prognostics & Health Monitoring (PHM) is used to:

• Understand coverage of functional flows in a system by sensors
• Understand how built-in sensors on legacy systems monitor functional flows
• Use algorithms to generate sensor sets to provide optimal coverage

Session 6.1 Outline

• 6.1.1: Sensor Library
• 6.1.2: Access Sensor Library
• 6.1.3: Sensor Library Editor
• 6.1.4: Create a Sensor Library
• 6.1.5: Sensor Library Categories
• 6.1.6: Sensor Detail Sections
• 6.1.7: Creating Sensors
• 6.1.8: ROC Curves
• 6.1.9: Sensor Selection Optimizer

Discussion 6.1.1: Sensor Library

• Contains both sample sensors & user-defined sensors in multiple categories.
• Sensors can be defined based on:
  • Physical attributes
  • Performance parameters
  • Operating environment
  • Flow types sensed
  • Any other custom properties

Exercise 6.1.2: Access Sensor Library

1. Select Modeling > Sensor Library from the main menu
2. Alternatively, select the sensor library icon from the icon toolbar

Discussion 6.1.3: Sensor Library Editor

The editor is divided into 5 sections:

• Sensor Libraries table — shows a list of libraries, categories and sensors
• Sensors Details — shows information on sensor ID, dimensions, costs, and reliability data
• Sensed Elements — shows the applicable flows for a sensor or sensor category
• Parameters — shows detailed sensor performance details such as POD, specificity, true/false rates, likelihood ratios etc.
• Additional Sensor Information — shows additional sensor information such as environmental constraints and custom properties

Exercise 6.1.4: Create a Sensor Library

1. Right-click on the open space within the Sensor Libraries table
2. Select New Library (alternatively, select the add icon next to the Sensor Libraries section)
3. Complete the Library Details:
   • Name: New User Library
   • Description: User Library containing sensors for the Training model.

Discussion 6.1.5: Sensor Library Categories

• Users can create multiple levels of sensor categories for a user library.
• Each library can contain specific sensed elements (flows, flow properties, symptoms).

Exercise 6.1.5: Sensor Library Categories

1. Select New User Library from the sensor libraries list
2. Select the add icon, or right-click New User Library and select New Category
3. Enter details for the Sensor category:
   • Name: Sample Category 1
   • Description: Sample Sensor category for New User Library
4. Select Flow Properties to be considered for this category by selecting checkboxes (for this category select Material — Liquid)

Create 5 additional Sensor categories:

Name	Description	Sensed Flow
Sample Category 2	Category for rotational energy flow	Energy, Mechanical — rotational
Sample Category 3	Category for linear energy flow	Energy, Mechanical — linear
Sample Category 4	Category for gas flow	Material, Gas
Sample Category 5	Category for liquid flow	Material, Liquid
Sample Category 6	Category for continuous signal flow	Signal, Continuous

Note: Sensors created in the categories will only be able to sense the selected sensed flows.

Discussion 6.1.6: Sensor Detail Sections

For a created sensor there are 5 sections:

• Sensor Libraries table — list of libraries, categories and sensors
• Sensors Details — sensor ID, dimensions, costs, and reliability data
• Sensed Elements — applicable flows for a sensor or sensor category
• Parameters — POD, specificity, true/false rates, likelihood ratios etc.
• Other — additional sensor information such as environmental constraints

Exercise 6.1.7: Create a New Sensor

1. Select Sample Sensor Category 1
2. Select the sensor icon or right-click and select New Sensor
3. Select Sensed Elements section
4. Uncheck all Sensed Elements and select only Liquid Dynamic pressure and Liquid Static pressure
5. Enter sensor details:
   • Name: Pressure Sensor
   • Part Number: PS1
   • Vendor: Sensor Company Inc.
   • Vendor Number: PS1-A
   • Dimensions: 50.0 (H), 20.0 (W), 20.0 (D)
   • Dimension Units: Millimetres
   • Weight: 100.0 grams
   • MTTF: 100,000 hours
   • MTTR: 5 minutes
   • Acquisition Cost: $100
   • Replacement Cost: $50
   • False Alarm Cost: $10
   • Detection Cost: $2
   • Operational Cost: $0.01 (per hour)
   • Testing Cost: $0.02 (per hour)
6. Enter sensor Parameters:
   • Estimated POD (Sensitivity): 0.98 (likelihood that sensor correctly detects failure condition)
   • Estimated Specificity: 0.90 (likelihood that sensor will not suffer a false alarm)
   • Verify additional fields: False Negative Rate 2%; False Alarm Rate 10%; Balanced Accuracy 94%; Likelihood Ratio Positive 9.80; Likelihood Ratio Negative 0.02
7. Enter Additional Sensor Details:
   • Description: COTS Fuel Tank Pressure Sensor
   • Detection Method: Passive
   • Signal Type: Digital
   • Operating Environment: Dry environment
   • Constraints: External temperature not exceeding 80 degrees Celsius

Create several more sensors:

Category	Flow Property	Name	Dimensions (H×W×D)	Weight	Replacement Cost	Detection Cost
Sample Category 2	Torque	Rotary Transformer	15, 10, 30 mm	10 g	$3500	$0.35
Sample Category 2	Angular Velocity	Angular Velocity Sensor 1	10, 10, 10 mm	60 g	$6000	$0.30
Sample Category 4	Angular Velocity	Angular Velocity Sensor 2	5, 15, 20 mm	120 g	$2000	$0.60
Sample Category 6	Angular Velocity	Angular Velocity Sensor 3	15, 15, 15 mm	75 g	$3500	$0.55
—	Linear Velocity	Optical Tracker and Velocimeter	30, 40, 10 mm	90 g	$10000	$0.30
—	Mass Flow Rate (Gas)	Tube Anemometer	4, 10, 8 mm	75 g	$1200	$2.00
—	Flow Rate (Liquid)	Orifice Plate	10, 20, 20 mm	300 g	$3000	$0.80
—	Continuous Amplitude	Amplitude Detector	5, 10, 15 mm	15 g	$1500	$0.10

(Table alignment across category/flow-property/sensor columns is degraded by PDF text extraction; sensor names and cost values above are transcribed as-is from the source. Cross-check exact category assignment in the live tool if precision matters.)

Discussion 6.1.8: Receiver Operating Characteristic (ROC) Curves

ROC curves use sensor failure data to plot the relationship between True Positives and False Positives observed based on a sensor's test threshold. This aids in:

• Determining the POD and Specificity for a sensor
• Selecting the most appropriate threshold based on the above

Exercise 6.1.8: Receiver Operating Characteristic (ROC) Curves

To import ROC setting:

1. Download the spreadsheet attached to this slide
2. Select the Tube Anemometer from Sensor Category 4 and select the ROC icon from the Parameters section
3. In the new window click the import button
4. Select Input Type as Direct Input
5. Browse for the ROC Curve Import Data.csv
6. To validate the .csv file, select Validate
7. Once validated, select Import

ROC Curve will be graphed and allows the user to determine the FPR Threshold that defines the Sensitivity and Specificity of the sensor:

• Input 0.33 as the FPR threshold and check that the numbers match, then confirm
• The slider can also be moved to select the threshold

Discussion 6.1.9: Sensor Selection Optimizer (SSO)

• Sensor Selection Optimizer aids the selection and allocation of a specific sensor from the sensor library to a location based on selected parameters.
• Uses the Analytical Hierarchy Process (AHP) to compare otherwise disparate parameters (e.g. weight, cost, and MTTF).

Exercise 6.1.9: Sensor Selection Optimizer (SSO)

1. Select the New User Library and select the add icon under Sensor Selection Optimizers
2. Enter the Name: Training SSO
3. Select the button and from the pop-up select checkboxes for: Weight, Cost, Cost Of Detection

Discussion note: The Analysis Type column indicates what the data is — Numeric or Text — automatically assigned when selecting the property. For this analysis, all Analysis Type is set to Quantitative (-ve), meaning lower is better.

To rank the properties:

1. Select the Criteria Ranking tab
2. Enter values into the Sensor Property Ranking matrix (only fields above and to the right of the principal diagonal require entry — yellow cells; hover over the input to verify importance as a tooltip):

	Weight	Cost	Cost of Detection
Weight	1	0.75	2.00
Cost	1.33	1	4.00
Cost of Detection	0.5	0.25	1

3. Select Calculate and select the Analysis Results tab

The results show that when selecting sensors, the priority should be minimizing in the following order: Cost, Weight, Cost of Detection. The Normalized Result displays the weightings that will be used when automating the allocation of sensors.

Session 6.1 Summary

• 6.1.1: Sensor Library
• 6.1.2: Access Sensor Library
• 6.1.3: Sensor Library Editor
• 6.1.4: Create a Sensor Library
• 6.1.5: Sensor Library Categories
• 6.1.6: Sensor Detail Sections
• 6.1.7: Creating Sensors
• 6.1.8: ROC Curves
• 6.1.9: Sensor Set Optimizer