Methods & Tool Used to Plan the Rewind of the Fish Units at The Dalles Dam

Terry B. Armentrout 1
Robert W. Ford 2

The Fish Unit Problem

The east fish ladder is a critical habitat to species listed as endangered, the Snake River Sockeye, and the Snake River Chinook. Attraction water, used to draw fish into the entrances of the ladder, comes from the discharge from two fish attraction water Hydro Generation units. Each unit has a rating of 15 MW and combined they provide a total discharge of 4,800 CFS. Presently full load is necessary on each machine in order to supply the minimum amount of attraction water to keep the east fish ladder within the criteria designated in the current fish passage plan.

Both fish units began service in 1957. They run continuously for 9 or 10 months of the year. Since no alternate attraction water supply exists for the east fish ladder, when the ladder is out of service for maintenance and inspection, the Project performs yearly maintenance on the units. Cavitation damage repair, seal replacement, governor adjustment, and alarm testing are some of the maintenance procedures accomplished during the normal outages.

The stator windings have asphaltic insulation similar to the insulation in the first 14 units of The Dalles Powerhouse. These main units began service in 1957 just like the fish units. In the 38 years of service, 99 failures have occurred to the stator windings of powerhouse units. Just good luck has prevented a winding insulation failure in the fish units.

The Portland District of Corps of Engineers developed an attraction water replacement study to investigate Project Improvements for Endangered Species(PIES). This study determined that no cost effective or resource efficient alternative exists. The failure of one or both fish unit windings will cause the east fish ladder to become out of criteria and incidental "takes" of the species occur. The risk of a fish unit winding failure increases with each day of operation. In time, the failure is certain.

The best possible option is to rewind the fish units before a crippling failure occurs. If the Project can replace the winding during the normal east ladder outage, the fish ladder will meet the fish passage criteria and no incidental 'take' will occur to the endangered species. The worst possible scenario is for a winding to fail during the fish passage season. This would cause the fish unit will be out of service for an extended period, continually causing damage to the endangered Salmon runs.

Normal outage periods for the east and north ladders are from early November through February. This is a period of 120 days, plus or minus. "Can the Project with supplemented forces, rewind a fish unit during that period?" That is the question that was asked at The Dalles Dam. The task of finding the answer to that question was the assignment to the four Project Maintenance Planners.

Fish Unit Rewind Study

The Assumptions:

  1. Rewinding the fish units will be the Project's top priority during the fishway outage period. The Project can forego or delay all other maintenance efforts in order to provide support to the rewind effort.
  2. We will assign the optimum crew to each task. The addition of one more craftsman will result in delay of the work. The subtraction of one craftsman will result in delay of the work.
  3. Work on the rewind will occur twenty-four hours a day, seven days a week, and through holidays.
  4. All supplies and materials will be on hand when necessary for the process.
  5. Assume a Chi Square skewed probability distribution. Having all go well will result in a small gain. Having things go badly will result in a large delay.
  6. 90% confidence level will determine the optimistic duration and the pessimistic duration. This assumption results in a 95% confidence the Project can complete the work in the pessimistic time or less.

Repair Alternatives:

The Planners considered these alternatives:

  1. Remove the stator from the generator, leaving the rotor in place.
  2. Remove several rotor pole pieces, rewind unit with rotor installed minus the removed poles
  3. Remove rotor exposing the full stator bore
  4. Remove rotor, bearing bracket, and exciter assembly in one lift, disassemble rotor, bearings, exciter as an off line task.

Study Method:

The study consists of analyzing alternative methods of disassembly and repair. The planners surveyed Project personnel who had disassembled units at either John Day or The Dalles. Each respondent gave his independent opinion on the number of man hours necessary to complete tasks with which he was familiar. From the responses, the planners calculated average duration, standard deviation and variation of the survey responses. The planners accomplished these calculations using the Microsoft Excel Program.

All work has a consistent or non variable portion. All work also has a variable component. So we can determine the limits of variation using standard statistical analysis. We have chosen to use a skewed probability distribution to analyze the variability and investigate the practical limits of each task. The Chi Squared probability distribution for 1 degree of freedom meets this requirement. For this distribution, the lower limit for the 90% confidence level occurs at the average minus 0.451 times the standard deviation. The upper limit for the 90% confidence level occurs at the average plus 3.39 times the standard deviation. Since our interest is only in the maximum duration of the task, the upper limit actually represents a 95% confidence we can complete the task at the pessimistic time or less. This skewed distribution fits our original assumption of things going well will result in a small gain. Things going poorly will result in a significant delay.

Planners applied the statistical methodology to the data supplied by the expert craftsmen who have first hand knowledge of unit assembly and disassembly. The appendix contains a compilation of that data on an Excel work sheet. The average duration, and the variations are the input for Microsoft Project. Project calculates all the network variables, determines the critical path, and allows resource loading for each of the separate tasks. The appendix includes a sample portion of the Project network chart.

Since each separate task's duration is independent of other task duration, we can add the variations. Independence in this case means the time to do task B does not depend on the time to complete task A. Hence the variation for the critical path is the sum of the variations for each separate task. The standard deviation is the square root of the variation . Thus we can use the lower limit and upper limit calculations described above to calculate the optimistic time and the pessimistic time for the entire project. and

Resources will vary from with the activity. Disassembly and reassemble will require a flurry of activity. Installation of a new winding will require less skilled resources. With the GANT and PERT charts drawn, the resource schedule is relatively easy to develop.

Analysis of Options:

Remove the Stator from the Rotor:

The fish unit stator is relatively small compared to a main unit. The stator is a single component, delivered and installed as a single piece. The gain from removing the stator would be to reduce the disassembly required. The less time spent on disassembly would mean more time available for the actual rewind work. Initially, removing the stator originally looked feasible. But, as the investigation progressed it was discovered that the thrust bearing bracket and the oil head assembly of the unit rests on the stator super structure. The manufacturer welded all these components directly to the stator frame. Consequently removing the stator from the rotor would mean disassembly of all components except for removal of rotor itself. This option left little gain in that removing the stator would disturb the alignment of the unit. Indeed, removing the stator might be more time consuming than removing the rotor because of the alignment considerations and the myriad of electrical connections associated with the stator which are not present for the rotor.

A sub option was to purchase a second stator with the winding already installed. Rewinding work would be outside the fish outage period and consequently relieve the time pressure during the outage. Work during the outage period would only consist of exchanging full stator assemblies. The draw back to this option is the necessary time for budgeting, approval, and fabrication. The normal budget cycle, if the item is indeed successful, is at least three years and includes feasibility studies, economic analysis, and considerable expenditure. If a new stator item would successfully clear the budget to program hurdle, an additional two years are necessary for procurement. This puts repair of the unit at least five years away. Repair would conflict with The Dalles major rehabilitation process adding complication and competition for scarce resources. This option presents high cost and high risk.

Remove Several Pole Pieces :

The intent of this option would be to minimize disassembly by removing a portion of the rotor pole pieces. The fish units are relative small, about half the size of a regular unit. Regular unit pole pieces are about 18"-24" thick and fish unit pole pieces are only 12"-15" thick. Having several men work in such a confined space severely limits productivity.

Windings span about 4 stator slots. Removing or replacing windings requires a cascading of individual windings. The limited space provided by pole piece removal would severely hamper removal. However, removal is possible because we can cut apart the old winding for removal. The space limitation might preclude installation because the bars have to go into place as whole assemblies. Indeed sufficient care in handling is necessary to prevent insulation damage. Using this method presents a high risk of a low quality repair.

Having a 12"-15" work space also severely limits the work force used in the rewinding effort. It also prevents an assembly line kind of effort where several activities can occur in parallel. For instance with the stator bore completely open, old winding removal, new winding placement, wedging, and electrical connecting can all take place simultaneously. In the pole removal method, each is a serial activity and can occur only after the preceding one is complete. Although this method results in saving the time for unit disassembly, the flow of rewinding work and the space restraint will quickly erode the savings. This option has high risk of poor quality and unacceptable time efficiency.

Traditional Rotor Removal:

This method consists of removing components above the stator allowing removal of the rotor. This option includes disassembling each component in turn. The disassembly work will be serial tasks of removing component A then component B, etc. Final disassembly task will be removal of the rotor and exposing the full stator bore. With the stator bore open, rewinding can receive the maximum effort. Removing, installing, wedging, and electrical connecting can be parallel work. Work crews can direct maximum assets toward the central focus of the repair, the rewinding.

Although this method sacrifices efficiency in disassembly at the expense of rotor exposure, stator work is the primary task. Disassembly is a means to that end. The efficiency of stator work will more than compensate for the inefficiency in disassembly. This option is worthy of a more detailed PERT analysis.

Removal of Rotor, Thrust Bearing and Exciter as a Single Component:

This procedure is to remove as much of the structure above the rotor -- shaft connection as a single piece. This method would allow quick exposure of the stator bore and give maximum outage time to the primary task. Perhaps a special jig or fixture is necessary to hold the rotor and attached components. Any disassembly, maintenance of the exciters or bearings and reassemble could be a parallel activity to the rewind.

Considerable thought needs to be given to the placement of the rotor and all attachments. Weight and stability are key considerations. Perhaps special rigging to lift all off in a single operation is also necessary. Existing rigging for lifting the rotor suspends the rotor with a single cable entwined through the rotor lifting device. This arrangement allows the rotor some freedom to rotate about a horizontal axis. Perhaps improving this rigging to include four slings would result in a more controllable rotor lift. This sling improvement could also be a tooling upgrade for the previous method. Significant analyses and planning effort are necessary to perform this procedure successfully and efficiently. This method contains substantial risk of delay because of unforeseen complications which will arise in a new and untried method. After the major lift of rotor out of the bore, all these possible complications will be away from the primary focus. This work will be a parallel effort to the most important work. This option has enough potential to warrant a more detailed PERT analysis.

PERT Analysis:

Data for each of the above options is in appendix A and appendix B respectively. Each appendix has:

  1. original estimates of the knowledgeable craftsmen
  2. Statistical manipulation of those estimates
  3. calculated average, optimistic, pessimistic times for each task
  4. PERT chart with the expected, optimistic, and pessimistic overall project times.

The expected time for completion of the rewind is 31 days by the "tried and true" method and 30 days for the new proposal. If all goes well with luck falling in the Project's favor, we can expect about a two day improvement over the expected time. If complications occur, an additional 12 to 13 days are necessary. The hard conclusion is that the augmented Project force can with 95% confidence rewind a fish unit and return it to service within in 45 days. With the normal fish passage outage period lasting 60-90 days as a repair window, Our confidence is even greater.

The time for each of the two methods differs less than a day. Although first appearances seem to indicate craftsman could expose the stator bore in substantially less time by lifting the exciter, bearing, and rotor as a single piece, the analysis does not support that conclusion. The expected times are essentially the same. Lifting the upper assemblies as a single piece requires substantial pre planning, jig and sling fabrication, and the introduction of risk into the whole project. So the rhetorical question arises: "Why introduce more work and risk for no increased benefit?"

The Enumeration of Unforeseen Complications

The original assumption of using a skewed probability distribution is to account for unseen events which can delay the rewind. Attempting to understand some of these possibilities and at least having a plan of action for that particular situation will mitigate their scope and effect. Planning for the unlikely situations is outside the focus of the primary task. The priority for planning to handle this situation is only to enumerate and make a cursory definition of alternatives. The Project can more completely define alternatives for possibilities having a disastrous effect on total job. Certainly we cannot enumerate all eventualities, however we can have a running start on work around options.

We have broken process complications into three categories: Those which would affect the critical path and thus delay the project; those which would affect work parallel to the critical path; and those which have little consequence. 12 possibilities affect the critical path. These eventualities deserve more attention. 7 problems affect the parallel work and have some potential to alter the outcome. The size and complexity of items in this category will determine what over all effect occurs. 7 snarl-ups are inconsequential and cannot delay the return of the unit to service.

Of the 12 critical path interfering problems the most vexing is a damaged bearing. Either newly discovered operational damage or damage during the maintenance effort would have a disastrous affect. Our experience on having new babbit placed in bearings indicates delivery time of months. Even in semi-emergency conditions such as the rehabilitation of John Day thrust bearing is taking much more time than expected. Discovering a bearing problem will be disastrous and have the same effect as a winding failure outside the normal outage period.

Obtaining a complete set of spare bearings is essential. Under the BPA critical spare parts effort the Project and the District needs to procure a set. Having a set on hand when the rewind activity begins, will give the Project the capability of working around a calamitous possibility.

The Rewind Plan

The first order of business is to lower the risk of the effort by learning more about the existing condition. We have learned much about assembly of the unit elements by reviewing shop drawings and by reconnaissance of the unit. However some important details require a closer examination. Questions like, "Is asbestos present in the ring bus connections?" and "What is the status of the surge ring?" need answers before a time constrained outage begins. Consequently, we will disassemble FU2 during the '95-'96 outage season to the point of removing the rotor. We have previously planned to remove FU2 exciter for dip tank cleaning in Portland. Continuing further to the disassembly will allow us to answer questions about the details of the construction. If we can obtain more accurate information about the thrust bearing cooler , we can have a replacement in hand for the full repair. We can also improve on our estimates for assembly and disassembly. We can remove some of the risk from doing the tasks during the '96-'97 season.

We can also treat the rewind of the two fish units as special North Pacific Division tasks. Just as the Ice Harbor lock gate replacement is a special maintenance activity which will alter maintenance schedules for the entire Columbia River, focus the North Pacific maintenance forces on this special 60 to 90 day activity. Beginning at the Project level, require participation inversely to the organizational distance from the Project. For instance have Portland District Project sacrifice the most people to the effort with Seattle and Walla Walla each contributing in a smaller fashion since they are organizationally more distant.

Testing The Data

The data was compared to actual data collected during the recent rewind of Fish Unit 2 at The Dalles Dam. The planning, disassembly and reassembly was performed by Corps of Engineer forces already stationed on Project. The rewind of the stator core was accomplished by Tennessee Valley Authority forces on site at The Dalles.

The disassembly of the unit commenced on November 18, 1996. The unit was disassembled and ready for TVA forces on December 2, 1996. The stator was rewound and ready for Corps reassembly on January 21, 1997. The unit was ready to return to service on February 3, 1997. The Unit was still undergoing last minute adjustments and modifications unrelated to the rewind at the time this report was written. The unit is expected to be returned to service on February 13, 1997.

Looking At The Data

Looking at the data table in appendix A, the data from the study can be directly compared to actual data collected during the rewind process. The table 1 below summarizes the predicted crew and man hours, determined by the study, for the overall process:

Study Data
Optimistic Average Pessimistic
Total Man Hours 7670 8365 13577
Total Crew Hours 867 968 1723

From the data collected from the actual rewind process, we can validate the study. For the overall rewind operation, 9,945 total man hours were necessary to complete the process. Looking above to table 1, the optimistic prediction was 7,670 man hours and the pessimistic prediction was 13,577 man hours. As we can readily see the actual data falls safely into the area between the pessimistic and optimistic expectations. Another quick look shows the same conclusions with respect to the crew hours expectation. With the optimistic and pessimistic predictions at 867 and 1723 crew hours respectively, the actual crew hours fall nicely inside the range at 1257 crew hours. Indeed, both data sets, from the study compare well when measured against real data!

Conclusions

The processes used in this study can be used to plan and predict the process schedule of most any process in which a historical data base can be compiled and manipulated. In fact, with the addition of innovative software such as Excel and Project, the statistical analysis is no longer a task relegated to high level staff. Using these tools, experience, imagination and little common sense even the most difficult and complex effort can be estimated and planned easily and accurately. These tasks can be given to on-site planners and coordinators to be used as tools for their daily planning and coordination.

1 Manager, The Dalles/John Day/Willow Creek Project, Portland District, US Army Corps of Engineers

2 Electrical Engineer, /John Day/Willow Creek Project, Portland District, US Army Corps of Engineers