military, Operation Research A JOIRI1 OF TI IIITI mi Zll$Iwrri tar ed pulcrlae V1, N1 I LIARY O FRA C SOECTI A JOIIIII OF WF1 ~iI{W O9970218700 (;JlIIOi182 199708100 Military Operations Research A publication of the Military OperationsR esearch Society The Military Operations Research Society is a pro- tioners and users of military operations research. fessional society incorporated under the laws of The Society does not make or advocate official poli- Virginia. The Society conducts a classified sympo- cy nor does it attempt to influence the formation of sium and several other meetings annually. It pub- policy. Matters discussed or statements made in the lishes proceedings, monographs, a quarterly bul- course of MORS symposia or printed in its publica- letin, ,PHALANX, and a quarterly journal, Military tions represent the opinions of the authors and not Operations Research, for professional exchange and the Society. peer criticism among students, theoreticians, practi- Editor Publisher Peter Purdue Natalie Addison Department of Operations Research Military Operations Research Society Naval Postgraduate School 101 S. Whiting Street, Suite 202 Monterey, CA 93940 Alexandria, VA 22304 Associate Editors Marion Bryson, FS, Monterey, CA Yupo Chan, Air Force Institute of Technology, Dayton, OH Peter Cherry, Vector Research, Ann Arbor, MI Paul Davis, RAND, Santa Monica, CA Christine Fox, CNA, Alexandria, VA Dean Hartley, Martin Marietta, Oak Ridge, TN Wayne Hughes, FS, Naval Postgraduate School, Monterey, CA James Kays, US Military Academy, West Point, NY Robert LaRocque, National Simulation Center, Fort Leavenworth, KS John Matheme, ALMC, Fort Lee, VA Brian McEnany, SAIC, Alexandria, VA Daniel Nussbaum, NCCA, Washington, DC Steven Pollock, University of Michigan, Ann Arbor, MI Clayton Thomas, USAFSAA, Washington, DC Eugene Visco, FS, USAMISMA, Washington, DC Mark Youngren, Joint Chief of Staff (J8), Washington, DC EditorialA ssistant Michael Cronin, MORS, Alexandria, VA Note from the Publisher: This is the first issue of Military Operations Research, the Journal of the Military Operations Research Society. This and other issues published in 1994 will be sent without charge to the entire MORS mailing list. Subsequent issues of the Journal will be sent to MORS members as part of their membership privileges and to government employees covered by our Sponsors. If you are not a MORS member or are not included on the list provided to us by our Sponsors, there will be a subscription charge for additional issues. A lot of work has gone into planning and publishing this first issue of the Journal. We hope that you find this publication to be of value. As always, the Board of Directors and Staff ask for your comments on this and any of our other publications or programs. Natalie S. Addison, Publisher © Copyright 1994, Military Operations Research Society, 101 S. Whiting Street, Suite 202, Alexandria, VA 22304. ISSN Number applied for. ( Printed on Recycled Paper Introductory Remarks Table of Colonel Greg Parnell,P resident, MORS .......................................................... 2 Dean Hartley, MAS Chairman ........................................................................ 3Contents Peter Purdue, Editor, Military Operations Research ............................. 4 Managing Ship Performance of Naval Gunfire Support Volume 1,N umber 1 Using Statistical Process Control Summer 1994 M .Bailey, 1. Bowden, A .Callahan. ...................................................................... 5 Sortie Optimization and Munitions Planning G. Brown, D. Coulter, A. Washburn ................................................................. 13 Modeling and Simulation for the New Pacific Community: A USPACOM Perspective D. Ha ut, M .M cCurdy ........................................................................................ 19 Topological Network Designs and Analyses for the Defense Information System Network M. Fischer, V Sielen, G. Swinsky, S. Wang ..................................................... 31 Commentary: "Combat Science: An Organizing Study" W. Hu ghes, Jr .................................................................................................... 45 Statement of Editorial Policy and Subm ission of Papers ............................................................................ . 5 8 Copyright Agreem ent .............................................................................. . . 59 ur first issue of the Military Operations Research Journal is a direct result of the vision, leadership and dedication of several MORS leaders and our MORS ye:ive" Sponsors. For many years, the MORS members and the MORS Board of Directors Ct have discussed the concept of a journal dedicated to military operations research. In April 91 to June 92, Mr. E.B. Vandiver III, COL James L. Kays, and Professor Peter Purdue successfully proposed the Journal by convincing the Board of Directors, and Become the Jo urnal obtaining support from the Sponsors. Van led the Journal effort as Education Committee Chair, VP for Professional Affairs, and President. Jim was an early supporter and major organizer as he succeeded Van as Education Chair and VP for Professional Affairs. Peter ofUC hoice was selected as our Editor and has been responsible for selecting the associate editors, reviewing the papers, and coordinating the administrative details. Many others have made significant contributions. Without our Sponsors' enthusias- tic support (and funding) the Journal would not have been possible. Brian McEnany, our Gregory S. Parnell, current VP for Professional Affairs, has actively pressed for publication of the initial Col, USAF issue, and has provided guidance for the development of the business plan. Gerry McNichols developed the preliminary business plan. Dick Wiles, our Executive Director, President and the MORS staff (as they do on all MORS activities) have made important contribu- Military Operations tions in planning and publishing the Journal. Research Society Without these efforts, the first issue of our Military Operations Research Journal would not have been possible. We hope to publish the best military OR analyses and document the military OR art of the leaders of our profession. Simply put, our objective is to become the journal of choice for all military OR analysts! Page 2 Military Operations Research, Summer 1994 or several years we in the Military Applications Section (MAS) of the Operations Research Society of America (ORSA) have been concerned about the lack of articles useful to practitioners in the Operations Research (OR) journals. OR is not mathe- S 0lvi na matics, economics, computer science, or statistics. OR may use techniques from these or other technical disciplines - or it may not. What OR is, is the application of disciplined Real thought in the effort to solve operational problems. Together with the Military Operations Research Society (MORS), we decided to create this journal to publish arti- Problems cles of interest to the military OR practitioner. We intend to publish articles about real problems and real attempts to solve the prob- lems. Some of the attempts will be successful; some will involve significant compromis- es; and some will reflect failures. Some of the efforts described will involve sophisticated mathematics and some will focus on the politics of human nature. In each case the thrust Dean Harley will be in identifying for you the situation, thought processes of the OR practitioners, MAS Chairman techniques attempted, and the results. We want our articles to be readable to the major- ity of practitioners and students of OR. Where detailed and abstruse mathematics are absolutely required (rare), the articles should include an explanation in plain English of the meaning (e.g., this equation shows that such-and-such is true or that it is legitimate to transform the problem in the desired way). When you read an article, you should be able to compare the problem with those you may face. If you find an approach that is novel and appears useful, you can contact the authors (or research the techniques yourself) for details. If you have faced a similar prob- lem and found (or suspect there is) a better approach, write a note to us. We may pub- lish the note. Or we may ask you to expand the note into an article. Philosophically, we regard the application of OR to be an art. Even the most experi- enced can learn from others. We are assuming the authors were doing their best consid- ering time available, talents, resources, and restrictions. Suggesting an alternative approach does not denigrate their efforts. That alternative may not have been applicable to that particular instance, but could be useful in others. Or the authors may not have been aware of the applicability and will learn from your suggestion. This view of OR also leads to an appreciation of what mindsets are useful (cf., Zen and the Art of Motorcycle Maintenance).W e will publish some philosophical and historical pieces as they are available to illustrate the possibilities. Communication is the basic reason for the existence of a journal. Its publication sup- plies the link from the authors to the readers; however, the link in the other direction requires work on your part. If you have something to say, write me or send an e-mail message. I am the MAS Associate Editor and your ear on the staff. Tell me what you want to see in your journal. Military Operations Research, Summer 1994 Page 3 his is the first issue of a new journal dedicated to supporting the military operations research community. The journal's aim is to further the development and promo- tion of the science and practice of military operations research. We will provide the This is community with a peer-reviewed outlet for information on methodological develop- ments, new and creative applications of existing methodology, and the exchange of ideas Your on the practical and the speculative aspects of our field. This is your journal!! You are hereby invited to contribute to its health and wellbeing. Jo ur n l a The journal will publish original research results, review papers, case studies and notes on all aspects of military operations research. I encourage prospective authors to make their papers complete and interesting, and not to be overly concerned with the length of the paper. Unlike some other journals MOR will not render papers unreadable Peter Purdue by requiring their authors to abbreviate them to the point of no return. To facilitate Editor, MOR debate on controversial or provocative issues, I will occasionally invite "commentary" pieces from well-known contributors to the field, and allow an exchange of ideas to con- tinue into future volumes. In this way I hope to stimulate some very interesting, as well as technical, discussions in the pages of the journal. Recent major changes in the world political and military scene have generated many new challenges for the military decision maker. We no longer have the luxury of con- centrating on the central front scenario; the threats we face have become much more dif- fuse, our forces are called upon to play new, non-traditional roles, new technologies present great opportunities as well as challenges. But these uncertainties and frustrations have generated a climate in which the military operations research community has its greatest opportunity to shine since its creation in World War II. The community will have to both adapt some old and reliable concepts, and learn to exploit new ones, like dis- tributed simulation. This should be a very productive time for operations research prac- titioners, developers, and academicians, as well as for policy makers. It is my hope and desire that the journal will facilitate the fruitful exchange of ideas among all these par- ties. As the editor of this new journal, I want to be responsive to the needs and expecta- tions of my readers. So, I encourage you to write to me with any ideas you might have about the journal, and how it may best serve your needs. And, do not hesitate to submit papers for review! The journal will only be as good as the papers it attracts. Join me in this exciting new venture. Page 4 Military Operations Research, Summer 1994 INTRODUCTION power into thermal power in the firing pin of N the gun. The chain continues as thermal aval Gunfire Support (NGFS) is a critical power until the round propellent is exploded. M anaging capability of the U. S. Navy, and is used Hereafter, the power is mechanical - the pro- to protect amphibious forces. Often, the jectile is catapulting toward the target. execution of NGFS is time-critical. The accu- As the projectile nears the target, its deto- Ship racy of the shooting (called fall-of-shot) and nation sequence takes place. This sequence is the time it takes the gun crew to shoot the gun started by the fuze in the nose of the projectile. Perform ance (called the gun cycle time), are two critical The fuze can be mechanically timed, point attributes of a mission. If the rounds land on detonating, or work based on the proximity of of Naval or near the target, and if the gun crew can fire the earth. Fusing is determined by the round the gun quickly, the NGFS mission is most type and the nature of the target. Once the Gunfire likely to be successful. detonation is witnessed by spotters near the Both of these performance characteristics target, an aimpoint correction vector is calcu- Support vary from ship to ship, crew to crew, and mis- lated. This vector is relayed to the gun crew sion to mission. There is both bias and vari- through the command, control, and commu- U ability in the characteristics, and one or both nications network. Uing can suffer when some component of the gun The ship has been underway during the system deteriorates. There is immediate feed- flight of the round, the construction of the Statistical back within the system in that the crew is con- spotting vector, and the communications. The tinuously apprized of its fall-of-shot accuracy, gun crew uses a fire control computer and a Process and any crew member can observe the gun radar or other navigational system to stabilize cycle time. the old aimpoint, then corrects the aim of the Control In this work, we develop a novel applica- gun based on the new spotting vector. Often tion of statistical process control (SPC) to fall- the gun gets a second aimpoint change of-shot, and to gun cycle times. We will show because of the nature of the target. At this Michael Bailey how control charting can be used during the point, the process starts over again. execution of an NGFS mission, as well as dur- Statistical process control is a set of tech- Associate Professor ing NGFS training to increase the efficiency of niques used in managing a noisy system we Department of Operations the gun crew and to save costly ammunition wish to keep stationary in mean and variabil- Research and range time. ity. These techniques have been used widely Naval Postgraduate School The purpose of this work is in managing machinery in factories. The goal Monterey, California 93943 of many SPC techniques is to provide an auto- John Bowden * to describe how SPC techniques can be matic method for detecting drastic deteriora- effectively used to manage NGFS process- tion of the performance of the equipment, Operations Branch Manager es; where performance is measured via a set of Crane Division, Naval " to discuss the practical hurdles in applying key performance characteristics. Control Surface Warfare Center SPC to data coming from NGFS qualifica- bcohaurntsd sa roen smimepalseu rdeedv iccheasr auctseerdis ttiocs ,e sstoa btlhisaht CCrarnae,e InWdaiarfnaan tion exercises; to discuss how the use of the results of SPC when the bounds are exceeded, production Alexander J. Callahan ton pshould be suspended and the machine in Senior Engineer agement of these platforms and the assets question should be adjusted or repaired. COMARCO theyexendt hine sematormsnandethnsst In what follows, we show how SPC tech- Bloomfield, Indiana they expend in maintaining readiness. niques can be used to manage the gun crew's execution of the mission. In particular, we will THE PROCESS use control charts to examine the performance of gun systems, where the attributes mea- The process of executing an NGFS mission sured will be cycle time and miss distance. has as its kernel the process of firing a single round accurately. Starting in the gun maga- METHODOLOGY zine, where the rounds are kept, a round is retrieved and placed in an automatic loader. The loader system delivers the round to the In this section, we describe the data we will gun chamber, and the system starts its elec- use to control our process as well as the tronic firing sequence. The sequence is analo- sequence of steps we employed to character- gous to a series of links in a chain which ize the behavior of the NGFS target process- ultimately ends in the conversion of electrical ing system. Military Operations Research, Summer 1994 Page 5 MANAGING SHIP PERFORMANCE The Data ance from the spotters. Spotting for area targets is significantly different and less demanding than for The data we used came from ship exercises per- point targets. Hence, the data associated with each formed at Atlantic Fleet Weapons Training Facility round fired may come from one of four situations (AFWTF), Isle de Vieques, Puerto Rico. We collect- - spotting point targets, spotting area targets, fir- ed data from three (3) different ships, one of which ing for effect on point targets, and firing for effect we saw for two qualification exercises separated by on area targets. twelve (12) months. The other ships were moni- The upshot of the preceding discussion is that tored for only one qualification exercise. Hence, the system we described above is only fully real- we have four observation sets. We recorded the ized during spotting for point targets. Point target data at the AFWTF observation post, where the spotting rounds are the most complicated to deliv- exercises are managed. er and involve all of the components of the NGFS The exercises are made up of a collection of process. Only a small portion of the data collected engagements designed to be challenging, as well as were spotting rounds for point targets. The data representative of the different NGFS missions that we analyzed for miss distances included only these a ship might be asked to perform in battle. The rounds, while we were able to accept cycle times exercises are performed by the ships in a controlled from point target spotting rounds as well as some environment where their execution can be of the area target spotting rounds. As we will see, observed in detail, and are scored based on the this makes each one-exercise data set too small to overall performance of the ship. The ships' crews use in off-the-shelf control charts. carry these scores as credentials, so the crews are The U. S. Navy developed and maintains a set of very motivated to perform well. threat lists for different warfare areas, including The exercises involve area and point targets. NGFS. The NGFS warplan establishes require- Each target undergoes a spotting evolution fol- ments that ships performing NGFS be able to lowed by a fire-for-effect evolution. In the spotting destroy or disable a specified set of targets in a evolution the gun system is expected to locate the specified time interval. These war plans are used target and fire a round at it. The round's impact is to develop exercises for NGFS qualification, and witnessed by spotters and the gun's aimpoint is can be used to develop thresholds for miss distance adjusted. After the spotters are satisfied that the and cycle times. As the usual manufacturing appli- gun can reliably hit the target, the fire-for-effect cation has a specification for workstation perfor- evolution begins. During the fire-for-effect, the gun mance, NGFS platforms have threshold cycle times fires a salvo of many rounds with no further guid- and miss distances which are derived from the U. S. Navy war plan for NGFS. Evolving capabilities of NGFS platforms are taken into account when the war plan is formulated, thus the goals set out in the war plan are consistent with the ships' process's SHIP CHARACTERISTIC TOTAL OK AFTER capabilities. ROUNDS OBS FILTER 1 cycle time 72 34 32 We recorded two characteristics - cycle times 1 miss distance 72 26 24 and miss distances. We wish to know: 2 (1st) cycle time 82 25 24 2 (1st) miss distance 82 24 24 e Is the time between rounds in control? If in control, is the time between rounds low 2 (2nd) cycle time 48 28 28 * 2 (2nd) miss distance 48 14 14 enough or should the war plan requirements be 3 cycle time 115 65 60 altered to reflect realities seen in the charts? 3 miss distance 115 52 52 I Is the miss distance for each gun in control? Table 1: Samples for the different characteristics for the e If in control, is the miss distance for each gun different ships. OK OBSs were rounds which we con- low enough, or should the war plan require- sidered to come from the same distribution as the char- ments be altered to reflect realities seen in the acteristic we sought. The filter mentioned was for charts? outliers Page 6 Military Operations Research, Summer 1994 MANAGING SHIP PERFORMANCE utable causes were for shots fired at down-slope CY targets. These target geometries amplify usually o small miss distances for the same reason that a 0 CYround raindrop looks oblong on a slanted wind- o) shield. E0 Preliminary Control Charts 10- As mentioned above, we employed the short pro- S2(1 St) duction run SPC techniques as provided in the sta- tistical package MIN1TAB [8]. We employed the Figure 1: A boxplot for the first exercise cycle times of short-run moving average (MA) chart and the ship 2. Two outliers were identified for removal, moving range (MR) charts. A prototypical control charting technique starts with the data X, X,..., X.. We group the data into 1 2 groups G, G,..., G, each group being of size m. 1 2 A Priori Identification of Outliers We then take some measurement for each group Gi, eg. the range in disjoint groups G, G,...,Gg Because we were the data collectors during the 2 exercises, we were privileged to listen to ship-to- shore communications. We took notes during the Ri = max xeGi X -min Xeci X. exercises with an eye toward identifying data points which were outstandingly bad, and where there existed known attributable causes for this We establish upper (lower) control limit UCL(LCL) badness. Due to the nature of the operation, some parts of the process, usually hardware failures, would cause large delays in the delivery of rounds UCL = R+36R; or cause rounds with large miss distances. These conditions are regularly regarded as failures by all concerned, and are treated by repairing the system. LCL = R+38R We wished to exclude these data from consider- ation because we wished to know if the system was We diagnose the system using a plot of the series Ri in control when the crew thought the system was along with the control limits. The system is defi- operational. Many such data were not recorded, nitely out of control if the control limits are exceed- but some points found their way into our data. For ed. Other diagnostics can also raise alarms, eg. a each data set, we constructed a simple boxplot [4] long series of points which are all increasing or and possible outliers were identified, see figure 1. decreasing. The control chart we used here as an The historical account was then referenced to example is called an R chart, and tells whether the determine if an attributable cause existed to variability in the data is consistent throughout. remove the data point. Small sample control charts typically use small Each data set was filtered for outliers with overlapping groups. MA computes the moving attributable causes. The result was that several averages of the groups using a specified window points were removed from the cycle time data, width w, and uses the moving range as a surrogate while only two points were removed from the miss for the population standard deviation to produce distance data. This was because causes for most control limits. Let Gi = {X{(i-l)w + 11, X{(i-1)w + large miss distances often involve gun system 2},..., X{(i)w}}, then alignment, and only examination of the gun system itself could confirm misalignment. Furthermore, any misalignment would persist throughout the I XeG, X; exercise, so that the resulting large miss distances MAE-= were not outliers, but consistent. Our only attrib- g Military Operations Research, Summer 1994 Page 7 MANAGING SHIP PERFORMANCE MRi = max XeG, X -min X , X. practices, the ship continued to shoot its qualifica- tion missions, used a large number of expensive Our small sample MA chart has control limits rounds of ammunition, and received a low score given by: quality control. This chart is: for its qualification for NGFS. Clearly, no one ben- efits from such an experience. UCL = lA+AIMR; LCL = lA+AMR. Table 2: Preliminary control charting results. A, is gotten from tables in sources like [1], or any SHIP CHARACTERISTIC MA N MR IN text on quality control. This chart is reputed to CONTROL CONTROL have increased sensitivity as compared with other 1 cycle time Y Y short-rim charts, and serves the same purpose - to 1 distance _ Y _ Y identify when the measure of central tendency 2(1st) cycle time Y Y goes out of control [3]. 2(1st) miss distance Y N 2(2nd) cycle time Y N MR charts are used to track the variation of the 2(2nd) miss distance N N data. When the process is seen to be out-of-control 3 cycle time Y N using MR, the process is displaying an unusually 3 miss distance N N high variability. This can happen even when MA is in control. The chart is formed using control lim- its: Consolidating Data: Historical and UCL= R+DMR, Intership Consolidation 5L= Md LCL = MR+DMRMd. As noted, the sample sizes available from a single exercise were too small for proper application of where MRMd is the median value of MR, i = 1, 2,..., standard control charting techniques. Further- g, and D, and D, are found in tables. As seen in fig- more, it is clearly desirable to give control limits to ure 2, our cycle times for ship 3 remained in control ships before they arrive at the AFWTF range to for MA but went out of control for MR. The assign- expend some expensive ammunition and range able cause for the behavior we see here is that the time. ship had intermittent visual contact with its navi- We were inclined to combine the data into a gation reference point. Thus it was forced to hold single data set to use for all ships. However, we its fire for short but significant periods of time dur- realized that the data sets came from different ing some missions. As a result of the analysis of the control charts for the different ships, we were able to reach the conclusions shown in table 2. From figure 2 we see 0 (o that the cycle times for ship 3 displayed large vari- ) o ation {\it in its variability). The MR makes two 0- sojourns outside the control limits. An I chart was > o * \ *' also constructed for this case, and it showed one < ( of9 .1 sojourn outside the control limits, as well as seven ._M 0 . consecutive decreasing points. These are both sig- o C nals that assignable causes should be sought. In the recommendations we form below, ship 3 o should have left the firing line early in the qualifi- 0 5 10 15 20 25 30 cation process to examine the causes for these chart behaviors. According to current management sample number Page 8 Military Operations Research, Summer 1994