Calculating General Tools Effectiveness (OEE) with AWS IoT SiteWise

Introduction

This weblog put up represents the second entry in a collection on utilizing General Tools Effectiveness (OEE) with AWS IoT SiteWise. On this put up, we are going to do a deep dive on methods to calculate OEE utilizing AWS IoT SiteWise native capabilities to gather, retailer, rework, and show calculations as an end-to-end answer. We are going to have a look at a Baggage Dealing with System (BHS) situated at an airport as a use case as an example the method.  Please, first learn half 1 of this collection, Industrial General Tools Effectiveness (OEE) information with AWS IoT SiteWise, for extra context on the use case.

Moreover, we are going to present how one can automate OEE parts to streamline the implementation of this answer in lots of different use instances, like manufacturing manufacturing strains in pharmaceutical, meals and beverage industries. That will help you put into apply the ideas offered on this weblog, we additionally present a code repository that allows you to stream artificial knowledge to AWS IoT SiteWise to create an OEE dashboard utilizing the calculations offered right here.

Use case

Earlier than diving deep into the OEE calculations, let’s outline the instance we will probably be utilizing as a body of reference. Our instance is a BHS and the required knowledge factors for an OEE calculation are gathered from the put in {hardware} on the BHS within the carousel.  The {hardware} consists of 4 sensors: two vibration sensors for motor monitoring, one velocity sensor for conveyor surveillance, and one photograph electrical sensor counting the luggage throughput.

The structure of the answer is as follows:

Sensor knowledge is collected and formatted by way of CloudRail, an AWS Companion whose answer significantly simplifies the gathering and streaming of IIoT knowledge to AWS IoT SiteWise. This integration is configurable by means of the CloudRail administration Portal straight.  The structure consists of extra elements for making the sensor knowledge obtainable to different AWS companies by means of an S3 bucket.

AWS IoT SiteWise pre-requisites

Earlier than sending knowledge to AWS IoT SiteWise, you will need to create a mannequin and outline its properties.  As talked about earlier, we’ve got 4 sensors that will probably be grouped into one mannequin, with the next measurements (knowledge streams from tools):

Mannequin:Carousel

Asset Identify: CarouselAsset

Property {

Measurement: Picture.Distance

Measurement: Velocity.PDV1

Measurement: VibrationL.Temperature

Measurement: VibrationR.Temperature

}

Along with the measurements, we are going to add a number of attributes (static knowledge) to the asset mannequin. The attributes signify completely different values that we want within the OEE calculations.

Mannequin:Carousel

Asset Identify: CarouselAsset

Property {

Attribute: SerialNumber

Attribute: Picture.distanceBase

Attribute: Picture.distanceThold

Attribute: Velocity.max_speed_alarm

Attribute: Velocity.min_speed_alarm

Attribute: Vibration.max_temp_c_alarm

Attribute: Ideal_Run_Rate_5_min

}

Now, let’s go to the and create the Carousel mannequin and asset that signify the airport BHS.

Open the navigation menu on the left, select Construct, Fashions, after which select Create Mannequin to outline the attributes and measurements for this mannequin:

For extra info on creating asset fashions go to the documentation.

Calculating OEE

Let’s check out the OEE definition and its elements.

The usual OEE method is:

Let’s have a look at the parameter definition for the BHS. For a full description of OEE parameters please go to the documentation.

• Ideal_Run_Rate:  In our case, the best run fee is 300 baggage/hour, which is equal to 0.83333 baggage/second. This worth depends upon the system and ought to be obtained from the producer or primarily based on subject commentary efficiency.

Availability

Availability = Run_time/(Run_time + Down_time)

We’ve got 4 sensors on the BHS and we have to outline what measurements (temperature, vibration, and so forth.) from the sensors we need to embrace within the calculation.  The temperature coming from the 2 vibration sensors (in Celsius) and the velocity of the carousel coming from the velocity sensor (m/s) will dictate the provision state.

The appropriate values for proper operation are primarily based on the next attributes of the Asset Mannequin.

Vibration.max_temp_c_alarm = 50

Velocity.min_speed_alarm = 28

Velocity.max_speed_alarm = 32

Let’s outline Equipment_State, an information rework that gives the present state of the BHS in numerical code:
1024 – The machine is idle
1020 – A fault, like an irregular operation of the system, excessive temperature or a velocity worth not throughout the regular vary outlined
1000 – A deliberate cease
1111 – A standard operation

The idle state of the BHS shouldn’t be outlined on this simplified use case, nevertheless, it’s potential to combine different knowledge streams into AWS IoT SiteWise and register info coming from Programmable Logic Controllers (PLCs) or different programs the place a human operator dictates if the system is idle or not.

So as to add a rework, go to the mannequin on the AWS IoT SiteWise console and select Edit. Scroll to the rework definitions and supply a Identify, Information sort (Double) and enter the next method on the respective subject:

Equipment_state =

if((Velocity.PDV1>Velocity.max_speed_alarm) or (Velocity.PDV1<Velocity.min_speed_alarm) or (VibrationL.Temperature>Vibration.max_temp_c_alarm) or (VibrationR.temperature>Vibration.max_temp_c_alarm),1020).elif(eq(Velocity.PDV1,0),1000,1111)

The method ought to appear to be this as you enter it within the console. The UI will carry strategies so that you can choose attributes and measurements already outlined within the mannequin to construct the method.

As soon as Equipment_State is outlined, create the next derived transforms to seize the completely different states of the BHS. Transforms can reference different transforms.

Proceed to outline the next metrics to mixture machine knowledge over time. Maintain the identical interval for every metric.

Fault_Time = statetime(Fault) – The machine’s complete fault time (in seconds)

Stop_Time = statetime(Cease) – The machine’s complete deliberate cease time (in seconds)

Run_Time = statetime(Operating) – The machine’s complete time (in seconds) operating with out subject.

Down_Time = Idle_Time + Fault_Time + Stop_Time – The machine’s complete downtime

The metric definitions of the mannequin ought to appear to be this:

High quality

High quality = Successes / (Successes + Failures)

Right here, we have to outline what constitutes successful and a failure. On this case our unit of manufacturing is a counted bag, so how will we outline when a bag is counted efficiently and when it’s not?  we use the measurements and knowledge which can be obtainable from the BHS’s 4 sensors.

The baggage are counted by trying on the distance the photograph electrical sensor is offering, due to this fact when there’s an object passing the band, the sensor will report a distance that’s much less that the “base” distance. This can be a easy strategy to calculate the baggage passing, however on the identical time it’s susceptible to a number of circumstances that may affect the accuracy of the measurement.

We use these mannequin attributes on the standard calculation:

Picture.distanceBase = 108

Picture.distanceThold = 0.1

The Picture.distanceBase is the space reported by the sensor, when there are not any objects in entrance of it. This worth would possibly must be calibrated frequently and adjusted, components like vibration and misalignment can result in false constructive counts.
Picture.distanceThold is used for outlining a threshold for the way delicate is the sensor, in an effort to keep away from counting particles or small objects (like bag attachments or belts) as an everyday bag.

We then set up two transforms for bag depend:

Bag_Count = if(Picture.Distance < Picture.distanceBase,1,0)

Dubious_Bag_Count = if((gt(Picture.Distance,Picture.distanceBase*(1-Picture.distanceThold)) and lt(Picture.Distance,Picture.distanceBase*0.95)) or (Velocity.PDV1>Velocity.max_speed_alarm) or (Picture.Distance>Picture.distanceBase),1,0)

Bag_count will account for all baggage passing in entrance of the photograph electrical sensor, and Dubious_Bag_Count will depend the objects detected as baggage below two irregular circumstances:

1. The gap detected is throughout the vary of 95% and 90% of the bottom distance; accounting for small objects and really small variations within the measurements, indications of modifications attributable to vibration or a sensor not correctly hooked up.
2. Baggage counted when the velocity of the carousel is above the restrict outlined; below this situation the sensor can miss counting baggage which can be too shut collectively on the carousel.

NOTE: the above circumstances are easy guidelines and the correct values for distance base and thresholds must be reviewed and analyzed with subject knowledge for higher outcomes.

Let’s outline successes and failures as metrics:

Successes = sum(Bag_Count) – sum(Dubious_Bag_Count)

Failures = sum(Dubious_Bag_Count)

Lastly we are able to outline OEE Availability as a metric as effectively:

High quality = Successes / (Successes + Failures)

Bear in mind to make use of the identical metric interval as in all different metric definitions.

Efficiency

Efficiency = ((Successes + Failures) / Run_Time) / Ideal_Run_Rate

We’ve got Successes and Failures from our High quality calculation, in addition to Run_Time from Availability. Due to this fact, we simply want to make use of the Ideal_Run_Rate_5_min, which in our system is 300 baggage/hour = 0.0833333 baggage/second.

OEE Worth

Having Availability, High quality, and Efficiency we proceed to outline our final metric for OEE.

OEE = Availability * High quality * Efficiency

Simplify transforms and metric definitions

As a substitute, the OEE elements outlined as transforms and metrics may be outlined programmatically as an alternative of utilizing the AWS Console. That is significantly helpful when there are advanced formulation that contain a number of variables, just like the Equipment_State and Dubious_Bag_Count transforms; additionally, automated options are much less error inclined than handbook ones and may be configured persistently throughout a number of environments. Let’s check out how we are able to do it utilizing the AWS SDK for Python (Boto3).

First, determine the measurements and attributes property IDs that you can be referencing on the rework/metric calculation, in addition to the mannequin ID.

Then outline a JSON for the metric/rework. As an illustration, to create a brand new rework to calculate the Equipment_State of the BHS we want the next attributes:

Vibration.max_temp_c_alarm

Velocity.max_speed_alarm

Velocity.min_speed_alarm

And the next measurements:

VibrationL.Temperature

VibrationR.Temperature

Velocity.PDV1

Create a file following this construction. Bear in mind to interchange the propertyIds and reserve it as equipment_state.json:

{
     "identify": "Equipment_State",
     "dataType": "DOUBLE",
     "sort": {
       "rework": {
         "expression": "if((var_speedpdv1>var_speedmax_speed_alarm) or (var_speedpdv1<var_speedmin_speed_alarm) or (var_vibrationltemperature>var_vibrationmax_temp_c_alarm) or (var_vibrationrtemperature>var_vibrationmax_temp_c_alarm),1020).elif(eq(var_speedpdv1,0),1000,1111)",
         "variables": [
           {
             "name": "var_vibrationrtemperature",
             "value": {
               "propertyId": "b9554855-b50f-4b56-a5f2-572fbd1a8967"
             }
           },
           {
             "name": "var_vibrationltemperature",
             "value": {
               "propertyId": "e3f1c4e0-a05c-4652-b640-7e3402e8d6a1"
             }
           },
           {
             "name": "var_vibrationmax_temp_c_alarm",
             "value": {
               "propertyId": "f54e16fd-dd9f-46b4-b8b2-c411cdef79a2"
             }
           },
           {
             "name": "var_speedpdv1",
             "value": {
               "propertyId": "d17d07c7-442d-4897-911b-4b267519ae3d"
             }
           },
           {
             "name": "var_speedmin_speed_alarm",
             "value": {
               "propertyId": "7a927051-a569-41c0-974f-7b7290d7e73c"
             }
           },
           {
             "name": "var_speedmax_speed_alarm",
             "value": {
               "propertyId": "0897a3b4-1c52-4e80-80fc-0a632e09da7e"
             }
           }
         ]
       }
     }
}

The principle expression is as follows:

if((var_speedpdv1>var_speedmax_speed_alarm) or (var_speedpdv1<var_speedmin_speed_alarm) or (var_vibrationltemperature>var_vibrationmax_temp_c_alarm) or (var_vibrationrtemperature>var_vibrationmax_temp_c_alarm),1020).elif(eq(var_speedpdv1,0),1000,1111)

Acquire the script update_asset_model_sitewise.py and extra particulars on methods to stream knowledge to AWS IoT SiteWise by visiting this public repository.

Then, run the next script passing the mannequin ID and the identify of the file beforehand outlined.

#python3 update_asset_model_sitewise.py --assetModelId [Asset Model ID] --property_file [JSON File defining the new property] --region [AWS Region]

After the script returns a profitable response, the brand new property ID created may be obtained straight from the AWS Console as described earlier than or by utilizing the AWS CLI to question the up to date mannequin definition and the jq utility to filter the end result.

#aws iotsitewise describe-asset-model --asset-model-id [model ID] | jq .'assetModelProperties[] | choose(.identify=="Equipment_State_API")'.id

You may then repeat the method with the opposite transforms and metrics in an effort to create all of the required elements for the OEE calculation.

For extra info on updating an AWS IoT SiteWise asset mannequin please go to the API reference.

Conclusion

On this weblog put up, we explored how we are able to use sensor knowledge from a real-life state of affairs to calculate OEE and get insightful info from our bodily system by utilizing AWS IoT SiteWise native capabilities. We walked by means of the method of figuring out the obtainable knowledge and we outlined the weather that represent the primary OEE parts, Availability, High quality and Efficiency, to lastly take a deep dive into the calculations and the way we are able to automate them.

As a name to motion, we invite you to take the content material offered right here additional, making use of the OEE calculation course of to your personal use instances, in addition to utilizing the automation instruments supplied to simplify and streamline the creation of information that helps monitor your industrial programs with accuracy.

Within the occasion you don’t have obtainable knowledge to make use of, we encourage you to comply with the steps outlined on this public repository to simply strive AWS IoT SiteWise with artificial knowledge and uncover the insightful info OEE may give you.

In regards to the Authors