energy savings in foundries

118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA

Computerization in the Foundry Industry (Panel 14-132)

Energy Savings and Productivity Improvements using Computer Controlled,

Remotely-Monitored, Smart Sensors

• James Wiczer, Sensor Synergy Inc., Barrington, IL• Brad Tidd, North Vernon Industry Corp., North Vernon, IN• Michael Wiczer, Sensor Synergy Inc., Barrington, IL

• Case studies described in this presentation were funded by the American Foundry Society under R&D project #12-13#03 and by the Collaborating Foundries including NVIC


Today’s Talk

• Background on Energy Savings & Monitoring

• Project 1 – V-Process Pumps• Project 2 – Shakeout Table & Dust

Collector• Project 3 – Induction Furnaces• Analysis Tools• Conclusions


What we want to talk about• Real-world application of Technology to

Foundry Issues

• Millions of Measurements

• Data shared in the Cloud & the Foundry

• Application of High-Tech Software Tools to Analyze Lots of Data

• Create Actionable & Meaningful Recommendations


Project #1 - Vacuum Pumps for V-Process Molds – Energy Saving Opportunities?

• Up to 4 Large Electric Motors Driving Vacuum Pumps to Supply Vacuum to the Pouring Floor and Shakeout area– Motors Include 250hp, 250hp, 250hp & 200hp

• Monitor Power Used by Motors and Vacuum Pressure Measured at Pouring Floor and Shakeout Floor


Measure Electric Power Consumed by 3 Connected

Motors –

Pouring Floor Vacuum Pressure Monitor


Unexpected Motor Failure During Measurements Revealed Excess Capacity


Vacuum System Energy Savings Opportunities

DatesHours “On” Past

Ideal Turn Off time

Approx. Extra kW-hr usage

Potential Savings if turn-

off time optimum

Total Potential Savings from 6 weekends

10,280 kW-hr $1028

Total Potential Savings from

1 year (50 wks) if these 6 weekends are

typical

85,570 kW-hr $8,570

Potential Savings by Reducing

Extra Vacuum

Capacity

No Capital Expenditure

Expenditures per

Building ($12,000)

Estimated Annual Failure Costs Due to

Reduced BackUp Vacuum Capacity

Measured in Bldg. 2 $80,000 $100,000 to

$130,000 $0 / $7,500

Estimated Total for both Bldg. 1 & 2

$160,000 $200,000 to $260,000 $0 / $15,000


Results from Vacuum Pump Monitoring

• Excess Capacity for “Design Margin” Can Be Costly in the long term

• Identified $50,000/Yr to $90,000/Yr at Each Facility Extra Capacity Energy Costs

• Effort Needed to Identify Extra Capacity (if any) and to Achieve Savings

• Cost vs. Savings Trade-Off


Project #2 – Monitor Dust Collector and Shakeout Table

• Monitor the Synchronization of the Shakeout Table and a Dedicated Dust Collector

• Opportunities for Savings with Better Synchronization

• Use 2-second Measurements to Monitor Operations of these two Power Hogs


Project #2 Conclusions

• Improved Synchronization of Shakeout Table and Dedicated Dust Collector Drive Motor can Save– $38,000/year in Electricity


Project #3 – Induction Furnaces #4 & #5 @ NVIC

• Install Power-Use Measurement Equipment on 2, 4MW induction furnaces ---- #4 and #5

• Make 1-Measurement/second, 24/7 for the duration of the project

• Send Data to Cloud Server to Share Data with all Interested Stake-Holders

• Correlate Power-Use Measurements with Activities at Foundry


Key Features of Induction Furnace Monitoring Project

• Over 16 Million Power Measurements during the first 3-months on 2 Furnaces

• Correlated Weight of Metal Heated During Each Furnace Heat with the Amount of Electricity Used for a Portion of the 3-month Study Period

• Correlated Run Sheet “Time of Day” Information with Power Measurements Time of Day Data


Equipment Installation

PC Data Viewer located in Foundry


Web Browser Data Viewer on the Internet

Power measurements for the same time duration during normal operations and during problem operations

Red area shows when

furnace is using

electrical power

White area shows when

furnace is not using electrical

power

Furnace Electrical Power

Usage is as expected – Start at ½ power then go to full power and finally off while tapping

Furnace Electrical Power Usage Profile is

unexpected – Long periods of “Off Time” while

equipment is repaired and other issues.


Cost of Electricity/Heat per Ton

12:00 AM 4:48 AM 9:36 AM 2:24 PM 7:12 PM 12:00 AM 4:48 AM0

20

40

60

80

100

120

140

160

Electricity Costs/Heat per 2,000 LB --- Costs Measured for Duration from Tap to Tap time)

Cost per 2,000 LB

Electricity Cost per Ton per HeatAverage $53.15Std. Dev $19.29Coefficient of Variation 0.363

Associating the amount of electricity used (costs) for a single heat with the amount of metal heated does not result in a strong correlation. In fact the Coefficient of Variation is Greater when we tried to compensate for the different amount of metal being heated during each furnace run.

Average Cost/Ton

$53.15

Lawanda

Brad SlideAbility to look at cost per ton Avg $53.15.Variation due to operator and equipment issues


Observations from Project #3

• Power Data is a High Resolution Indicator of Induction Furnace Activity

• Knowing the Costs of “Issues” may help a foundry determine which ones to Remedy – Equipment in need of Frequent Repair– Shift Change Issues– Hold Power Usage

• Power Setting During Hold Times• What Events Cause a Shift to Hold Power• Total Energy Expenditures after Ready Time


Costs with Worst 25%

9/4/13 12:00 AM 9/4/13 12:00 PM 9/5/13 12:00 AM 9/5/13 12:00 PM 9/6/13 12:00 AM 9/6/13 12:00 PM 9/7/13 12:00 AM 9/7/13 12:00 PM$0.00

$50.00

$100.00

$150.00

$200.00

$250.00

$300.00

$350.00

$400.00

Tap to Tap Costs/Heat ($'s) for 3 Day Sample / Orange - Lower 75% / Lt. Blue = Upper 25%

Tap to Tap CostsUpper 25% of Tap to Tap Costs

Date and Time of Day

Tap

-to

-Tap

Co

sts

($'s

)

Average Costs

including Lt. Blue Values

Average Costs Excluding Lt. Blue Values


If we could do this, then ….

• The average electricity costs/heat for operating the induction furnaces would decrease from

•$200.18 … to … $178.11

• This would create and Annual Electricity Savings of …

•$466,400


Data Analysis Tools

• Build tools to study data specifically for foundry applications

• Goal: Extract the best information for those most knowledgeable about the operations & demands of a particular foundry– Experts + data can learn the most about

improving their operations


Generating Furnace Operation Logs

• Let’s focus on one example of a data analysis tool we’re working on– Problem inspired by this project with

NVIC• In this project, we focused on

correlating energy with operations.– This required accurate logging that is

difficult to achieve over extended periods of time


Generating Furnace Operation Logs (cont.)

• Challenges to getting accurate operations logs / run sheets:1. Manual logging is time-consuming

• Inaccuracies may result from the many tasks performed by furnace operators at tap time

2. Additional instrumentation for electronic detection of melt start / end times is often complex and costly


• Use the power data we collect to estimate when operation events (e.g. tap time, melt ready time, etc) occur

• But first…

Furnace Operation Logs – A Third Option


Some preparation for a little bit of computer talk

• Algorithm: A calculation or procedure a computer program uses

• Parse: Splitting up data into pieces• Score: The algorithm’s evaluation of

some data.• Sigma (σ): Standard deviation. The

“spread” of some data.


Software identification of melt start & stop events

• Computer prediction can sacrifice some accuracy in favor of ease-of-implementation– Algorithmic approach– Get lots of data and treat them

statistically– Enable operators to focus on core

duties


How the algorithm works• Manually log times during

“training” period– We used 3 days & 82 melts for this

example• Learning based on Principle

Component Analysis (PCA)– Most notably used in face

recognition algorithm “eigenfaces”– Think of a 15 min window around

each event as a “face”(Photo: AT&T Labs Cambridge)


Sliding windowPower measurements

Prediction score

Window (prediction input)

Evaluate a window and predict how likely start of tap event is at the center.

15 Minutes


Sliding window scoresPower measurements

Prediction score

The power curve gets a corresponding curve for likelihood of tap start.


Algorithm processes scores and generates melt start times

Power measurements

Predicted event times

Take the highest scores in a neighborhood and get discrete tap time predictions.


Repeat for “melt ready” events

Apply the same algorithm to melt ready times and combine the results.

Power measurements

Predicted start times

Predicted ready times


Exploratory Analysis

Tap to Ready Time (min)

Ave

rage

Pow

er (

KW

)


Exploratory Analysis

Tap to Ready Time (min)

Ene

rgy

Con

sum

ed (

KW

Hr)


Improved Metrics

Mean = 42.1 minσ = 12.4 min

Mean 11.7 minσ = 4.3 min


Improved Metrics

• Analyze furnace usage from “Ready time” to “End-of-Tap time”:– Average energy: 24 KW Hr– % of Tap-to-Tap energy: 2.4%– Average time: 11.8 min– % of Tap-to-Tap time: 23.7%


Conclusions• These Projects Improved Energy Costs and

Operational Efficiencies by Applying High-Tech Monitoring Hardware & Software

• Foundry Operations Experts & Energy Analysts Exploit these Tools to Identify and Quantify Issues

• In each Case, Monitoring & Analysis Generated Actionable Recommendations


For additional information:

James (Jamie) Wiczer, PhD Sensor Synergy, Inc. – Barrington, IL

[email protected] / Ph. 847-353-8200

Brad Tidd, Senior Kaizen Manager North Vernon Industry Corporation, North Vernon, IN

Brad.Tidd@NVIC-CWT / 1-812-346-8772 x 1288

Michael Wiczer, Senior Software Engineer

Sensor Synergy, Inc. – Barrington, [email protected] / Ph. 847-226-0687