overview

Copyright 2005, Society of Petroleum Engineers This paper was prepared for presentation at the 2005 SPE Annual Technical Conference and Exhibition held in Dallas, Texas, U.S.A., 9–12 October 2005. This paper was selected for presentation by an SPE Program Committee following review of information contained in a proposal submitted by the author(s). Contents of the paper, as pre- sented, have not been reviewed by the Society of Petroleum Engineers and are subject to cor- rection by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for com- mercial purposes without the written consent of the Society of Petroleum Engineers is prohib- ited. Permission to reproduce in print is restricted to a proposal of not more than 300 words; illustrations may not be copied. The proposal must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435. Abstract This paper represents an integration of artificial intelligence and lean sigma techniques to achieve large field production optimization. The first part of the methodology (detailed in SPE 90266 “Zonal Allocation and Increased Production Op- portunities Using Data Mining in Kern River”1) involves data management and predictive data mining for increased produc- tion opportunity identification. It utilizes a set of data mining tools including clustering techniques and neural networks to identify new candidates for clean-outs, perforating, sidetracks, deepening, and other types of workovers. Furthermore, the expert system was used to predict the estimated production in- crease for these candidates. The second part of the methodol- ogy optimizes the implementation and post-workover follow up of the opportunities identified in part one. It involves the use of lean sigma tools such as value stream mapping, level loading, continuous flow production, standard operating pro- cedures, and kanbans which optimize execution cycle time, peak oil production, decision making process, cost, and safety2. This approach was successfully applied and executed in the Kern River field.

Introduction With over 8,600 active producers averaging 10 BOPD each and a limited staff, streamlining the well optimization process in the Kern River field is critical to take advantage of a large and dynamic portfolio of relatively low oil gain opportunities. It is essential to effectively identify, prioritize, and implement a high number of these opportunities, which typically range from 2 to 8 incremental barrels of oil per day.

As detailed in SPE paper 902661, a significant production increase opportunity was discovered in the lower sands through the use of artificial intelligence tools after observing that some wells in the field have high production, while nearby neighbor wells are very low producers. A pilot pro-

gram was implemented and following its success, the study was extended across the entire field. After identifying the field-wide opportunity, a significant workover program was launched.

A lookback on the pilot program indicated several proc- esses, including candidate selection, were successful and would continue to be used “as is” in the execution of the field wide effort. The post-workover follow up and put on produc- tion (POP) processes, however, were identified as weaknesses and were highlighted as areas of improvement. Lean Sigma techniques were selected to optimize and streamline these processes.

Background This paper represents an integration of artificial intelligence and lean sigma techniques to improve workflow processes and execution of a large field optimization project in Kern River.

Reservoir Description. The Kern River field, located in Kern County, California, is a heavy oil reservoir consisting of nine productive sand intervals and many more individual sand lobes or flow units within the Kern River series. The field is 4 miles by 5 miles in areal extent and has over 8,600 active pro- ducing wells and 1,200 steam injectors. Producers are co- mingled with very little individual zone production test data available. The field is currently produced by steam injection with varying degrees of thermal maturity in each of the sands. The primary production mechanism is gravity drainage with extremely low average reservoir pressure of 20 psi in the oil sands, requiring pumps to be set at or below the bottom-most oil sand and pumped off to effectively produce.

The northeast half of the field has little to no water im- pacted sands, while the lowermost sands in the central portion of the field are water/aquifer impacted. The water impacted sands are found progressively higher moving southwest, down structure, across the southwest half of the field. Higher pres- sures exceeding 50 psi are found in these sands.

If open, the water/aquifer impacted sands produce excess water and often cause the fluid level to rise above the pump. As a result, oil production from the sand lobes immediately above is oftentimes reduced due to increased backpressure on the oil sand. There is typically little vertical separation (10 to 30 feet) between the oil productive and wet sands. If the well is plugged back to shut off water, oil zones in close vertical proximity may also be plugged off or the plug may prevent

SPE 97247

Integration of Artificial Intelligence and Lean Sigma for Large-Field Production Optimization: Application to Kern River Field A. Popa, SPE, R. Ramos, A. Cover, SPE, and C. Popa, SPE, Chevron

2 SPE 97247

setting the pump low enough to drawdown the oil sand to ef- fectively drain it.

The zones targeted in this study are the lowermost in the reservoir and referred to as the R sands. These workovers in- volve some combination of deepening the total depth or plug- back total depth, adding perforations, and almost always lowering the pump setting depth. In some cases, water shut- off is employed to allow the well to pump off and affect an oil gain by lowering the working fluid level relative to the oil productive zones. It is crucial to determine the oil versus the wet zones as closely as possible to assure workover success, particularly in the southwest half of the field.

Large Field Optimization. Significant optimization efforts at Kern River are generally done in a field wide programmatic approach. Past results of other optimization efforts at Kern River have indicated that this is the best way to effectively identify, prioritize, and implement field wide opportunities. It also provides a better means of incorporating learning and im- provements in a timely manner across the entire field wide opportunity base. The large areal extent, high number of wells, high number of similar opportunities, and large amounts of data in Kern River create economies of scale within the processes involved in these optimization efforts if done on a field wide basis.

Although most production engineers at Kern River are as- signed specific areas of the field to work within, significant optimization efforts at Kern River are more efficient if ap- proached in a field wide programmatic approach with dedi- cated technical staff, workover rig crews, and supervision, while still engaging the production engineers for each specific area to assure capture of their expert knowledge of that area. There are also many functional groups in a large field like Kern River that can be better coordinated by one team rather than numerous individual production engineers generating and implementing work for each of their specific areas. Some of the functional groups at Kern River typically involved with this type of effort are the workover group, operations group, engineering group, facilities group, development group, meas- urement group, and numerous others.

Artificial Intelligence. Artificial neural networks and fuzzy logic are the two primary artificial intelligence techniques util- ized in the data mining portion of this work. Artificial neural networks (ANNs) are excellent tools for pattern recognition and non-linear, multidimensional interpolation. Complex rela- tionships existing between input and output parameters are re- vealed through the use of distributive parallel processing. Fuzzy logic is a technique by which an object can belong to multiple sets in varying degrees, unlike traditional set theory where an object either entirely belongs to a set or it does not. It can process information using natural language and thus provides maximum efficiency in using imprecise information. Detailed information regarding neural networks and fuzzy logic can be found in the referenced publications3-11. Lean Sigma. Six Sigma is a highly disciplined process that helps organizations focus on delivering products at lower cost, with improved quality and reduced cycle time. Six Sigma theory was first applied to processes in the manufacturing in-

dustry, but has since spread to the aerospace, pharmaceutical, heavy manufacturing, and transactional service industries, and now the oil industry. The core notion behind Six Sigma is the measurement of the number of defects in a process in order to systematically figure out how to eliminate them, thus reducing the number of defects and getting as close to zero as possible.

Sigma is a statistical term that measures how far a given process deviates from perfection. A Six Sigma quality process must produce less than 3.4 defects per million opportunities. Six Sigma is a vision many organizations strive toward and a philosophy that is part of their business culture. Some of the most common tools used in Six Sigma projects are: Process Flow Mapping Cause and Effect Diagrams Input Process Output (IPO) Diagrams Pareto Charts Histograms Regression Analysis Root-cause Failure Analysis Failure Mode Effect Analysis Lean Tools

The systematic utilization of these tools in the five-step DMAIC process improves the quality of projects. These steps are: 1) Define, 2) Measure, 3) Analyze, 4) Improve, and 5) Control.

Lean sigma utilizes the t-test as a formal statistical test to detect a shift in average. The interpretation of the resulting P- value indicates that if the P-value is less than .05, the result is considered to be significant. In other words, the change in process made a significant impact on shifting the average and there is a (1 – P)*100% confidence that the change in average did not happen by random chance, but instead because of a change in the process.

The F-test is another statistical test used to detect a shift in standard deviation. Similar to the t-test, the interpretation of the resulting P-value indicates that if the P-value is less than .05, the result is considered to be significant. In other words, the change in process had a significant impact on shifting the standard deviation and there is a (1 – P)*100% confidence that the change in standard deviation did not happen by random chance, but instead due to a change in the process.

Lean Sigma is a process improvement methodology used in world-class organizations to eliminate waste from company processes and deliver exceptional products and services to their customers. The relentless pursuit of elimination of waste is the main concept behind a lean system. Waste is typically found in the following seven categories: Over production Waiting for the next process step Transporting materials unnecessarily Non-value added processing Inventory that is more than minimum Motion that is unnecessary Defects and rework

Lean Sigma has spread out to many industries successfully and continues to develop and grow as it is combined with Six Sigma. The combination of Lean and Six Sigma creates Lean Six Sigma, which helps companies improve quality and elimi- nate waste.

SPE 97247 3

Methodology The methodology is described in two parts: First, the artificial intelligence predictive model, which de-

tails the candidate selection process and oil production prediction, particularly the enhancements introduced fol- lowing the publication of SPE paper 902661.

Second, the lean sigma execution model, explaining the overall steps in the workflow analysis and the implementa- tion of lean sigma techniques in various processes. Following the successful pilot work in 2004, all work

processes were analyzed for strengths and areas of opportunity. Many processes, such as candidate selection, were identified as fortes, while other processes, such as post wellwork optimization, had significant gaps. As a result, a formal team was chartered to address the findings in a structured and focused manner. This led to an increased effort to identify and implement R sands workovers across the entire Kern River field at a heightened pace during 2005.

Post wellwork optimization is not an extremely technically complex process or one that was not well understood in the last program. However, it is a process that requires consistent and reliable coordination of multiple repetitive activities. This is particularly due to the high number of workovers that must be followed up and the large degree of interaction and coordination between the many functional groups involved. Post wellwork optimization of the R sands workovers involves: Assuring the wells have been cyclic steam stimulated as

needed to achieve the expected oil gain. Assuring the well is pumped off to optimize drawdown of

the newly targeted oil zones for maximum oil gain. Follow up remediation to shut-off water in the case of sig-

nificant water production gain without prohibitively costly larger pumping equipment. This was often done by partial plugback.

Miscellaneous sand cleanup workovers, mechanical re- pairs, and gauging repairs as needed.

Data Availability. In the first publication regarding this pro- ject, detailed in SPE 902661, it was demonstrated that the po- tential of the Kern River deeper sands could be estimated using basic, commonly available data such as completion, zone markers, and production data. The complete list of pa- rameters can be seen in Table 1.

Since the original study was performed in a small part of the reservoir, an expanded application of the methodology was desired for the entire field. Due to the increased complexity of the reservoir, additional reservoir data was brought into the analysis. Thus, various parts of the field (with potential water encroachment, areas separated by faults, and areas with low production) were required to be treated separately. As a result, data was queried for each of the specific areas, and then studied in an identical manner to that presented in SPE 902661.

For the Lean Sigma optimization, the following parameters were recorded and entered into a tracking tool database to measure the efficiency of the process: Time to POP (Put On Production): the number of days be-

tween the wellwork completion date and the date of first valid gauge. Conditions that would increase this time in-

clude immediate pump problems, sand problems, lead line failure, and large elapsed time to gauge due to miscommu- nication between operating parties. In some cases after wellwork, the installation of a larger pumping unit with new electric connections was required.

Time to Peak: the number of days from the date of well- work completion to the date of peak oil production. This parameter suggests when the well is fully optimized. A well is considered optimized when it has visible fluid pound and has been cyclic steamed.

Incremental Oil at Peak: the peak production after well- work minus the average production prior to wellwork.

Artificial Intelligence Predictive Model. The predictive model utilizes a set of data mining tools (including regression, neural networks, and fuzzy logic) to identify candidates for remedial work and the corresponding production increase ex- pected. SPE 902661 describes this module in detail and out- lines how it was applied in a 500 well test area of Kern River field.

First, clustering techniques were used to identify production increase potential. The clusters define the optimized system (wells performing at maximum potential) and the system in need of optimization (wells that do not perform as expected). In addition, the clustering techniques also indicated zonal allocation for the area studied. Next, the integration of artificial neural networks and fuzzy logic determined the candidate selection technique.

Neural networks were used to predict the expected oil production for currently underperforming wells if wellwork was completed on them to be optimized. The neural network proved to be a strong correlation system between completion data, sand markers, and production potential. The neural network system was trained using the data from optimized wells and then applied to predict production for the currently low producer wells.

After expanding the study to the entire Kern River field, training one neural network is neither efficient nor accurate due to the complexity of the reservoir (water zones, faults, dip, etc). Reservoir information, namely saturation maps, was built at different sand levels to identify high potential areas field wide. As a result, multiple neural networks were trained for certain areas of the field to better predict the expected oil production in that area.

In addition to predicting the expected oil production, the neural networks were trained to also predict the water production. This parameter was found to be extremely important in the second part of the process during implementation and optimization. The expected water production allowed proactive sizing for the lifting capacity (pump size) and installation of the correct pumping unit.

The fuzzy expert system is the second piece of the wellwork candidate selection methodology. Two of the three parameters (total expected daily oil production and increased oil production over current production) used in the fuzzy system are derived from the neural network prediction. The third parameter is the risk index and is meant to incorporate the chances of wellwork success as a result of the impact of the current completion.

4 SPE 97247

This system was used for candidate selection and provided a strong, unbiased selection.

Lean Sigma Execution Model. The first step in leaning out the execution process was to define and understand the current process. This was accomplished by creating a current state map outlining how materials and communications flow throughout the process. All steps were captured, including value added and non-value added steps that are required. The knowledge and understanding gained from the development of the current state map then became the foundation for the fu- ture state map.

The Kern River R sands process was mapped and the fol- lowing items were captured as opportunities: Insufficient detail in scope for activities after wellwork. Complicated communication links between activities and

activity owners. Excessive wait time or idle time due to poor delivery of

pumping units. Conflicting priorities among operations staff resulting in

well underperformance. These opportunities were incorporated into the future state map.

The future state map has lean flow designed into the proc- ess; the success of which depends on the quick adoption of the proposed process. A critical key is the assignment of respon- sibility for implementation to various process managers that have the capability to make changes happen across functional and departmental boundaries. The future state map for the lean R sands process was fairly straightforward – it concen- trated a majority of efforts on improving communication. The process flow diagram in Figure 1 outlines the process used in this project.

The new processes were designed to consistently and ef- fectively communicate information to all parties involved in the process. Each activity in the process followed a desig- nated sequence, and all team members were assigned desig- nated roles and responsibilities in the process.

The first step in the newly defined lean R sands process was the identification of workover opportunities for the entire field by area. This portion of the process had the least amount of opportunities to improve. Previous work during the pilot was successful in quickly identifying the most successful can- didates by utilizing the neural network and fuzzy logic meth- odology in the artificial intelligence predictive model.

Since the project evaluated opportunities throughout the Kern River field, the second step (as depicted in Figure 1) was an engagement meeting between lean R sands project team and the production area owners. The main purpose of this meeting was to obtain buy-in and gather knowledge from the area owners to be used during candidate selection.

Next, the lean R sands team identified and selected the candidate wells with the highest probability of success. This step was also completed using the artificial intelligence pre- dictive model. Wells on the probable wellwork candidate list were then scoped out in detail and an economic package was assembled with the following information: Detailed completion, production, and reservoir data Proposed workover program with estimated cost Detailed well file history results and risks

Proposed surface facilities upgrades and new installations, such as pumping unit upgrades

This level of detail in the scoping phase was extremely impor- tant in simplifying the communication and material flow proc- ess for all activities.

Economic justification and execution of wellwork were the next two steps in the process outlined in Figure 1. These steps were also greatly simplified in terms of communication as a result of having a detailed scope of work and realistic cost estimates.

Once the well workover was completed and returned to production, the well was monitored and officially transferred back to area owners. The well was considered to be optimized once the following has occurred: No repeat failure after work is completed (sanding, pump

failures, etc). Well has had a dyno within 7 days of the POP date and

dyno is current (within 7 days of handing the well over to area owner).

Well has been pumped off or considered to be optimized with fluid over pump condition by area owner.

Well has been considered for a cyclic steam job within 7 days of the well being pumped off.

If the well still was not producing as expected, proactive work was completed by lean R sands team to stimulate production (mechanical stimulation, packer steam, etc). In the event of unsuccessful workover or corresponding unsuccessful efforts to optimize the well, the project team returned the well back to the area owner in the best possible condition.

The last step of the newly developed lean R sands process was to conduct a lookback in order to identify best practices and lessons learned and quickly apply them in the process. This final step highlights the lean objective of continuous im- provement and the relentless pursuit of elimination of waste. Results

Due to the success of the pilot detailed in SPE 902661, this study was extended across the entire Kern River field with the addition of lean sigma techniques for implementation optimi- zation. As a result, put-on-production cycle time, time to peak, and incremental oil gain were improved.

Figure 2 shows the 2004 and 2005 incremental production from the R sands remedial program. Workover programs for 80 wells were completed in 2004, and 38 wells were com- pleted through May 2005. From the beginning of the program in April 2004, a 14% decline was arrested in the well group of completed workovers. As seen in Figure 2, a total incremental production of over 700 BOPD has been achieved to date with the trend still ascending. The successful lean R sands program was able to restore production to a level equivalent of that which was seen five years earlier.

Put on Production Cycle Time. Figure 3 clearly shows a re- duction in the average number of days a well was taking to be put on production following wellwork. Figures 4 and 5 show the capability of the process to meet specifications as it relates to the number of days to POP a well. The defined expectation for POP of a well is 6 days. In 2004, more than 50% of wells exceeded the 6-day specification. On the other hand, in 2005, 96% of all wells met the 6-day requirement.

SPE 97247 5

Not only was the average POP cycle time reduced, but the standard deviation decreased as well, enhancing the consis- tency in delivering the defined specification. The POP t-test value of 0.001123 indicates that there was an overall change to the process with 99.98% confidence. The POP F-test value of 0.0 confirms with 99.999% confidence that there was a change to the process.

Time to Peak. Similarly to put on production cycle time, Figure 6 also shows the process capability before and after the changes implemented through lean sigma. The defined speci- fications for time to peak were set a 60 days. In 2004, more than 50% of wells exceeded that criterion, while in 2005, 89% of wells met the 60-day target. Figures 7 and 8 show the CPK analysis for time to peak in 2004 and 2005 respectively.

Incremental Oil Gain. There is a visible increment in the av- erage amount of oil gained from 2004 to 2005, as seen in Fig- ure 9; however, the standard deviation increased slightly, as seen in Figures 10 and 11. The lean sigma efforts in this pro- ject did not target average oil production response since it is difficult to justify such a process improvement inherently re- lated to reservoir characteristics. Furthermore, the artificial intelligence based candidate selection process continued to consistently predict the incremental oil gain on each individual well with accuracy. It was not deemed necessary to have an overall incremental oil gain specification, since the overall multi-well economic packages were based on average incre- mental oil gain rather than minimums. Despite this, it is safe to say that the utilization of lean sigma techniques improved the project implementation and more barrels were captured quicker as a result of the reduced POP and time to peak cycle times.

Conclusions The combination of artificial intelligence tools and lean sigma complement each other well. Artificial intelligence tools pro- vide the best possible production incremental opportunities utilizing data mining and programming algorithms, while lean sigma improves the quality and efficiency of the implementa- tion process by removing waste, establishing consistent steps and creating continuous flow of communication and material from start to finish.

The data mining process utilizing artificial intelligence tools, presented in SPE 902661 and improved upon in this study, continues to be successfully used for field wide candi- date selection and optimization. Multiple neural networks were used for different field areas due to reservoir complexity.

The use of lean sigma tools has demonstrated a significant improvement in the implementation and post wellwork proc- ess leading to shortened put-on-production time, higher peak oil gains, and faster time to peak.

This methodology has proven successful for large fields and can be adapted and utilized for similar fields.

Acknowledgments The authors would like to thank Chevron for allowing the pub- lication of this work. Special thanks to Mr. Art Lewis and Mr. Greg Emerson who championed this project and encouraged the efforts of finding potential production increase and opti-

mizing execution efficiency using lean sigma tools. Thanks also to Kathy Pollock, Alan Townsend, Bobby Carey, and Ge- rald Dodson for their contributions.

References 1. Popa, C., Popa, A., Cover, A.: “Zonal Allocation and Increased

Production Opportunities Using Data Mining in Kern River”, SPE 90266, Proceedings, 2004 SPE Annual Technical Confer- ence and Exhibition, 26-29 September, Houston Texas.

2. “Six Sigma Specialist.” Air Academy Associates, Colorado Springs, Colorado, 2000.

3. Popa, A., Mohaghegh, S., Gaskari, R., Ameri, S.: “Identification of Contaminated Data in Hydraulic Fracturing Databases: Appli- cation to the Codell Formation in the DJ-Basin”, SPE 77597, Pro- ceedings, 2003 SPE Western Regional/ AAPG Pacific Section Joint Meeting, 19-24 May, Long Beach, California.

4. Mohaghegh, S., Popa, A., Gaskari, R., Ameri, S., and Wolhart, S.: “Identification of Successful Practices in Hydraulic Fracturing Using Intelligent Data Mining Tool; Application to the Codell Formation in the DJ-Basin”, SPE 77597, Proceedings, 2002 SPE Annual Technical Conference and Exhibition, 29 September-2 October, San Antonio, Texas.

5. Mohaghegh, S., Popa, A., Mohamad, K., and Ameri, S.: “Per- formance Drivers in Restimulation of Gas Storage Wells”, SPE 57453, Proceedings, 2003 SPE Eastern Regional Conference and Exhibition, 21-22 October, Charleston, West Virginia.

6. Mohaghegh, S., Reeves, S., and Hill, D.: “Development of an In- telligent System Approach for Restimulation Candidate Selec- tion”, SPE 59767, Proceedings, 1999 SPE Gas Technology Symposium, 3-5 April, Calgary, Canada.

7. Hosn, N., Popa, A., Popa, C.G.: “A New and Realistic Approach to Pumping Unit Optimization through the Use of Intelligent Sys- tems”, SPE 72360, Proceedings, 2001 SPE Eastern Regional Conference and Exhibition, 17-19 October, Canton, Ohio.

8. He, Z., Yang, L., Yen, J., and Wu C.: “Neural-Network To Pre- dict Well Performance Using Available Field Data”, SPE 68801, Proceedings, 2001 SPE Western Regional Meeting, 26-30 March, Bakersfield, California.

9. Coste, J.F., Valois, J.P., Mouret, Cl., Guittard, M., and Daniel, O.: “Data Mining Techniques For Optimizing Fast Track Re- engineering of Mature Fields”, SPE 78333, Proceedings, 2002 13th European Petroleum Conference, 29-31 October, Aberdeen, U.K.

10. Ali, J.K.: “Neural Networks: A New Tool for the Petroleum In- dustry?”, SPE 27561, 1994 European Petroleum Computer Con- ference, Aberdeen, U.K., 15-17 March.

11. Ross, T.J.: “Fuzzy Logic with Engineering Applications”, McGraw Hill Inc., 1995.

6 SPE 97247

Tables

Category Parameter Description WellName Commonly used well name API API number BottomLatitude Latitude of the well at the bottomhole

Well Identifier

BottomLongitude Longitude of the well at the bottomhole TopR Measured depth to the top of the R sand, ft TopR1 Measured depth to the top of the R1 sand, ft TopR2 Measured depth to the top of the R2 sand, ft hR Thickness of the R sand, ft hR1 Thickness of the R2 sand, ft

Zone Data

hR2 Thickness of the R2 sand, ft hpR Total footage perforated in the R sand, ft hpR1 Total footage perforated in the R1 sand, ft hpR2 Total footage perforated in the R2 sand, ft TD Total depth of the well measured during original drilling, ft ED Current effective depth of the well, ft PumpDepth Current pump depth intake, ft CasingSize Diameter of the production casing, in LinerSize Diameter of the slotted liner (if present), in InnerlinerSize Diameter of the innerliner (if present), in

Well Completion Data

Damage Minimum depth of recorded casing damage, ft LastOil Measured oil from last well test, BOPD LastWater Measured water from last well test, BWPD 30DayOil Average oil production over last 30 days, BOPD 30DayWater Average water production over last 30 days, BWPD 6MonthOil Average oil production over last 6 months, BOPD

Production Indicators

6MonthWater Average water production over last 6 months, BWPD Table 1: Input Data

SPE 97247 7

Figures

Figure 1: Integrated Artificial Intelligence and Lean Sigma Process Flow Chart

A rtificial Intelligence M

odule

8 SPE 97247

Figure 2: Incremental Production for R Sands Program

Time, years

April 2004

P ro

du ct

io n,

B O

P D

SPE 97247 9

Kern River Lean R-sands

POP Cycle Time Chart

CEN=6.6 CEN=2.4

-20

-10

0

10

20

30

40

50

Ap r-0

4 Ju

n-0 4

Ju l-0

4

Oc t-0

4

Ja n-0

5

Ap r-0

5

Figure 3: Put on Production Cycle Time Improvement using Lean Sigma

Cpk Analysis 2004 POP Cycle Time

-3 0

-2 5

-1 9

-1 4 -8

-2 .5

3. 01

8. 51 14

19 .5 25

30 .5 36

41 .5

In spec

Out spec right

USL

Mean = 6.6156 StdDev = 8.1394 USL = 6 LSL is Not Defined Sigma Level = -.0756 Sigma Capability = 1.4244 Cpk = -.0252 Cp is not available DPM = 530,145 N = 80

Figure 4: CPK Analysis 2004 Put on Production Cycle Time

Cpk Analysis 2005 POP Cycle Time

-7 .4 -6

-4 .5

-3 .1

-1 .7

-0 .2

1. 19

2. 62

4. 05

5. 47 6.

9

8. 33

9. 76

11 .2

In spec

Out spec right

USL

Mean = 2.125 StdDev = 2.1123 USL = 6 LSL is Not Defined Sigma Level = 1.8345 Sigma Capability = 3.3345 Cpk = .6115 Cp is not available DPM = 33,293 N = 38

Figure 5: CPK Analysis 2005 Put on Production Cycle Time

10 SPE 97247

Kern River Lean R-sands

Time to Peak Cycle Time Chart

CEN=62.43

CEN=37.11

-100

-50

0

50

100

150

200

Ap r-0

4

Ap r-0

4 Ju

l-0 4

Au g-

04

No v-0

4

Fe b-

05

Ap r-0

5

Figure 6: Time to Peak Cycle Time Improvement using Lean Sigma

Cpk Analysis 2004 Time to Peak

-1 15 -9

1 -6

7

-4 4

-2 0

3. 61

27 .3

50 .9

74 .5

98 .2

12 2

14 5

16 9

19 3

21 6

In spec Out spec right USL

Mean = 62.425 StdDev = 39.342 USL = 60 LSL is Not Defined Sigma Level = -.0616 Sigma Capability = 1.4384 Cpk = -.0205 Cp is not available DPM = 524,575 N = 80

Figure 7: CPK Analysis 2004 Time to Peak Cycle Time

Cpk Analysis 2005 Time to Peak

-4 6

-3 5

-2 4

-1 3

-1 .8

9. 38

20 .5

31 .7

42 .8 54

65 .1

76 .3

87 .4

98 .5

11 0

In spec Out spec right USL

Mean = 37.105 StdDev = 18.544 USL = 60 LSL is Not Defined Sigma Level = 1.2346 Sigma Capability = 2.7346 Cpk = .4115 Cp is not available DPM = 108,487 N = 38

Figure 8: CPK Analysis 2005 Time to Peak Cycle Time

SPE 97247 11

Kern River Lean R-sands

Incremental Oil Gain per Job

CEN=6.558 CEN=8.982

-15

-10

-5

0

5

10

15

20

25

30

35

Ap r-0

4

Ap r-0

4 Ju

l-0 4

Au g-

04

No v-0

4

Fe b-

05

Ap r-0

5

Figure 9: Incremental Oil Gain per Job Improvement using Lean Sigma

Cpk Analysis 2004 Incremental Oil Gain

-2 1

-1 7

-1 4

-1 0

-6 .5 -3 0. 6

4. 15

7. 69

11 .2

14 .8

18 .3

21 .9

25 .4 29

32 .5

In spec Out spec left LSL

Mean = 6.5575 StdDev = 6.0557 USL is Not Defined LSL = 3.5 Sigma Level = .5049 Sigma Capability = 2.0049 Cpk = .1683 Cp is not available DPM = 306,816 N = 80

Figure 10: CPK Analysis 2004 Incremental Oil Gain per Job

Cpk Analysis 2005 Incremental Oil Gain

-1 9

-1 5

-1 2 -8

-4 .4

-0 .8

2. 89

6. 53

10 .2

13 .8

17 .4

21 .1

24 .7

28 .4 32

35 .6

In spec Out spec left LSL

Mean = 9.0 StdDev = 6.2098 USL is Not Defined LSL = 3.5 Sigma Level = .8857 Sigma Capability = 2.3857 Cpk = .2952 Cp is not available DPM = 187,891 N = 35

Figure 11: CPK Analysis 2005 Incremental Oil Gain per Job