Tony Sileo Opvantek is exploring development of a risk model to help optimize cross bore inspection programs based on data collected from various interesting data sources We will provide an update on this activity and also solicit presentations or discussion from across the user group on your own ID: 626059
Download Presentation The PPT/PDF document "Moderator:" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Moderator: Tony SileoOpvantek is exploring development of a risk model to help optimize cross bore inspection programs based on data collected from various interesting data sources. We will provide an update on this activity and also solicit presentations or discussion from across the user group on your own experiences and requirements related to the cross bore threat.ObjectivesIdentify data sources that can be used to find indications of legacy cross bore repairs.Review and gather input on other factors that make the likelihood of a cross bore more or less likely.Share initial results from a cross bore risk model developed in collaboration with one of our customers to help prioritize a legacy cross bore inspection program.
Risk Model for Cross-Boring Slide2
GOAL – risk score to prioritize legacy cross bore inspection programCustomer had built a preliminary model (keyword data mining)Gathered SME input on potential data sourcesExtracted legacy work orders and customer dispatch orders from data warehouseIncorporated enhanced GIS data from Optimain DSBuilt and reviewed 10 iterations of the model Building the Cross Bore Risk ModelSlide3
Gas Main Features (GIS and Optimain DS)Service Line Records (GIS, Optimain DS and Service Card System)Repair Work Orders (WMS)Pipe Install/Replace Work Orders (WMS)Customer Service Orders (DIS)Cross Bore Logs – workbooks by stateCross Bore Inspected Areas – MS Word documentsSummary of Data Used Slide4
Work Order Job TypesSlide5
Search comment fields for word patterns that might indicate a cross-bore was repaired10 million records … 876 possible cross boresEstablish HistoricMapId’sCross Bore Repair KeywordsSlide6
Install/Replace Pipe Work OrdersSearch for Trenchless Technology keywords in commentsEstablish HistoricMapId’sPossible Trenchless Tech Install Work OrdersSlide7
Also found 192 possible cross bore repairs from nearly 900,000 customer dispatch ordersCustomer Service x-Bore RepairsSlide8
Customer Svc Work Orders with Trenchless Tech KeywordsCustomer Service Trenchless Tech Install IndicationsSlide9
Cleansed and aligned with Municipality and StreetName values from GIS (as captured on GIS Main Features in Optimain DS)Cross Bore Log DataSlide10
Computed for two different geographic areasHistoricMapId (“Map”)HistoricMapId + StreetName (“StreetMap”) For each area with at least Y x-Bores, how many have at least one more? How many more did each area have?Sum over all areas – use to establish a-priori probability algorithmsProbability of at least one more x-Bore in the same areaExpected number of additional x-Bores in the same area (“EV”)X-Bore Probability AlgorithmsSlide11
Probability and Expected # Xbores by Map (v10)Slide12
# of Mains Possibly Installed with Trenchless Tech (vintage, mat’l)# of Mains with XBores on Install Job# of Services Possibly Installed with Trenchless Tech (vintage)# of Services with Long Company Side# of Services Initially Installed Since 1970# of Services Replaced Since 1970# of Services with Work Order indicating Trenchless Tech InstallAll factors normalized to percent of maximum# in this area / maximum from any areaProbability FactorsSlide13
Factor Weights optimized to maximize correlation of Prob Score with Actual Cross Bores FoundProbability Scores (Map or StreetMap)Prob Score
Factor 1 Weight
Factor 1 Score
Factor 2 Weight
Factor 2 Score
Factor n Weight
Factor n Score
Sum of all Factor WeightsSlide14
Weights tuned manually to use the StreetMap level EV and ProbScores to rank risk among all Streets that are on Maps with greater than the average number of expected additional cross bores. Does NOT use the Map ProbScore in the final algorithm since we’re only looking at maps that already have a number of reported cross bores somewhere on the map.Final Risk Algorithm - High EV StreetMaps
Final Risk Score
Probability Score
Consequence Score
Probability Score
MapEV
MapEV
Weight (3.0)
StreetMapEV
StreetMapEV
Weight (100.0)
StreetMap
ProbScore
StreetMap
ProbScore
Weight (10.0)
Consequence Score
Pop Density Factor
Max Pop Density This
StreetMap
Avg
Pop Density All Census Block Groups in StateSlide15
Weights established manually to use only Map and StreetMap ProbScores to rank risk among all Streets that are on Maps with fewer than the average number of expected additional cross bores, but with higher than average Map Prob Scores.Does NOT use the EV’s since these are all in Maps (and thus StreetMaps) with very few prior reported cross bores (thus, all EV’s are low or 0).Final Risk Algorithm - High FactorScore StreetMaps
Final Risk Score
Probability Score
Consequence Score
Consequence Score
Pop Density Factor
Max Pop Density This
StreetMap
Avg
Pop Density All Census Block Groups in State
Probability Score
Map
ProbScore
Map
ProbScore
Weight (10.0)
StreetMap
ProbScore
StreetMap
ProbScore
Weight (20.0)Slide16
Example Scoring Illustration2nd Ave3rd Ave
Newkirk Dr (EV=0.38, Factor Score=70.6)
Kingsdale
Blvd (EV=0, Factor Score=57.6)
Ridgewood Dr (EV=0.38, Factor Score=84.3)
1
st
Ave
Map 7436580I
MapEV
= 1.51
Factor Score = 257
PopDensity
= 2.8
Cross Bore Found
Possibly
Xbored
Mains: 4
Mains w/
xB
On Install Job: 2
Possibly
Xbored
Svcs
: 42
Long Side
Svcs
: 35
Svcs
Inst Since ‘70: 0
Svcs
Repl
Since ‘70: 42
TT Inst
Svcs
: 0
Possibly
Xbored
Mains: 39
Mains w/
xB
On Install Job
: 6
Possibly
Xbored
Svcs
: 32
Long Side
Svcs
: 7
Svcs
Inst Since ‘70: 32
Svcs
Repl
Since ‘70: 0
TT Inst
Svcs
: 0
Possibly
Xbored
Mains: 2
Mains w/
xB
On Install Job
: 2
Possibly
Xbored
Svcs
: 68
Long Side
Svcs
: 34
Svcs
Inst Since ‘70: 67
Svcs
Repl
Since ‘70: 1
TT Inst
Svcs
:
1Slide17
Positively map every cross bore indicationRelate every xBore indication that was due to a service line to the current service at that address (by premise ID)Relate every xBore indication that was due to a main line to the current main or abandoned main feature at that address (by GISID). If more than one main, relate to the one that was xBored, or that replaced the one that was xBored (if no abandoned main found).Positively identify every area that has been inspected or reviewed and deemed to be safe (office cleared)Polygon feature on map Capture well-defined StreetName, Municipality, HistoricMapID, FromStreetName, ToStreetNameEstablish a way to positively mark or track any historic work order that indicates a cross-bore occurred (Yes, Maybe, No, Not Reviewed).
Recommended Data Improvements