Vous êtes sur la page 1sur 11

Journal of Ship Production, Vol. 18, No. 2, May 2002, pp.

105115

A Dimensional Engineering Process for Shipbuilding


Mark H. Spicknall* and Ramesh Kumar
*University of Michigan Transportation Research Institute, Marine Systems Division; and Business Performance Group LLC Dimensional Control Systems, Inc.

Dimensional engineering practices and tools used in other industries are highly evolved and broadly applicable to shipbuilding. However, because of relatively small product volume per design and relatively high variability in the dimensional attributes of intermediate products, detailed dimensional engineering of each ship design is not economically viable. This paper describes a viable dimensional engineering process for shipbuilding.

Introduction
DIMENSIONAL management and engineering practices, commonly referred to as accuracy control in shipbuilding (Yokata 1963, Levingston 1980, Storch 1984, IHI 1985), are fairly standard and well established in other industries such as automotive and aerospace. These practices include statistical process control (SPC) on the shop floor (Shewhart 1939, Grant 1996), and dimensional variation simulation analysis and geometric dimensioning and tolerancing (GD&T) (ASME 1994, Drake 1999) in detail/ production design. The combination of these practices is often referred to in other industries as a dimensional control system. This paper focuses primarily on dimensional variation simulation analysis practices, which are most often referred to as dimensional engineering. Dimensional engineering practices utilized within other industries are generally applicable to the shipbuilding industry. However, a ship design and production system differs from the systems that produce, say, automobiles in ways that will impact dimensional engineering practices. For example, very few products are typically produced per design within the shipbuilding industry, so a ship is most often a semi-custom or custom product. And although each ship is made up of a very large number of intermediate products (IPs), because of the semi-custom nature of ships the number of standard intermediate products is small. Therefore, to gain economies of scale, group technology is utilized to the greatest extent possible. Group technology exploits the similarities of common types of intermediate products to gain the production volume necessary to justify the creation of intermediate-productPresented at the Ship Production Symposium, Ypsilanti, Michigan, June 1315, 2001.
MAY 2002

type-specific process lanes and work centers. But still within each intermediate product type/group/family there will be some degree of variability in dimensional attributes. Given this variability both between and within products, to fully analyze the impact of IP dimensional variation on final product form and function would require a substantial dimensional engineering effort for each ship design. While such an effort is economically viable within the automobile industry because many products will be produced from a single design, such a dimensional engineering effort is not economically viable for a ship because of the few ships typically built per design (Kumar & Spicknall 2000). An alternative approach for applying dimensional engineering practices to shipbuilding is as follows: Model and simulate a representative sample of intermediate products of each type/group/family and then utilize standard parametric dimension and tolerance optimization results developed from those simulations to establish GD&T for all future intermediate products within that type/group/family, Model and simulate in detail only the portions of the ship that are both unique and likely to be dimensionally critical. The remainder of this paper describes this dimensional engineering process in detail using an example of a typical ships double bottom.

Dimensional engineering process for shipbuilding


Figure 1 shows the dimensional engineering process for shipbuilding. Each step of the process is described below. 1. Select build strategy alternativeThe build strategy identifies the product structure, including all intermediate products from
JOURNAL OF SHIP PRODUCTION 105

8756/1417/02/1802-0105$00.47/0

Fig. 1

major assemblies to piece-parts, and their assembly relationships. The build strategy should take into consideration the combination of product design features and process parameters that meet all program requirements and functional product objectives. It should balance major assembly, subassembly, and some detailed build requirements against total product requirements. Build strategy development must be coordinated with design and engineering. When deciding the appropriate build strategies, due attention must be paid to issues like:

Actual versus required production capabilities Cost Schedule Risk 2. Identify dimensionally critical intermediate products and determine if a GD&T standard exists for eachAs discussed earlier, the low number of ships built per design, the very high total number of intermediate products per ship, and the moderate to low

106

MAY 2002

JOURNAL OF SHIP PRODUCTION

number and generally nonstandard nature of intermediate products of any given type, make detailed dimensional modeling of every ship design economically nonviable. An alternative approach is to identify the dimensionally critical intermediate products in the total product structure, and utilize existing dimensioning and tolerance standards if they exist to generate the final GD&T scheme for the intermediate product (Step 5). If such a standard does not already exist, one must be created via detailed dimensional engineering analysis (Step 3). Eventually as these standards are generated and reused, detailed dimensional analysis for a new ship design will include only the portions of the product that are both dimensionally critical and unique. 3. Carry out detailed dimensional engineering analysis for dimensionally critical IPs for which no standards existIf a standard GD&T scheme does not exist for the type/group/family of a particular dimensionally critical intermediate product, it is first necessary to define the detailed production processes for this IP. Then critical dimensions are defined and functional design tolerances are allocated. Functional design tolerances describe in dimensional terms the allowable limits in build variation that will produce a product that meets overall requirements for fit and function. Functional design tolerances are acquired from the functional design/engineering group. After identifying the critical dimensions and assigning design tolerances, it is now time to apply production variances to the parts and assemblies. These data are acquired from a process performance standards database, which is continually being updated from shop-floor SPC data. Production variances data might also come from probabilistic structural engineering analyses for IPs that have never been built before and for which there is no statistical production data. The next step is to analytically determine if the product and processes as specified by design would meet overall dimensional build objectives. Conduct sensitivity analysis to show large contributors to variation. Identify and rank the effect critical characteristics will have on assembly functional requirements. Develop a corrective action for each root cause (i.e., product and/or process). Apply preliminary KPCs to identified tolerances based on level of contribution to overall dimensional build objectives. This is an iterative process required to achieve the overall dimensional build objectives. The general goal is for tolerances to be as wide as possible while still achieving final dimensional build objectives. Build strategy and design/process/tolerance requirements could possibly be refined in order to achieve a tolerance distribution that minimizes production cost while achieving overall dimensional build objectives. 4. Assess resources to meet requirementsOne of the most important steps in the process is the assessment of the availability of human resources and capability of production tools and equipment. This process essentially validates that production capabilities are sufficient to meet the dimensional requirements established. If shop floor variance data were used in the dimensional engineering analysis as described above, presumably the capabil-

ity of production tools and equipment is satisfactory. Does the organization have the human resources at its disposal to achieve the dimensional objectives? If the required human resources are not available or the capability of production tools and equipment is not satisfactory, it is time to reevaluate the design, process, tolerances or the build strategy. Any one of them or a combination of them might require change, modification, or refinement to achieve overall goals. Use simulation analysis to optimize tolerances. 5. Finalize GD&T and complete and issue production information for each dimensionally critical intermediate product (define and maintain GD&T standards by IP type)Release level GD&T is developed using either existing standards or the results of dimensional analyses. Then deliverable production drawings are developed for the dimensionally critical IPs. Design and production engineering should identify and maintain the configuration of the 3-D GD&T schemes. 6. Produce, measure, and analyze dimensional data to validate GD&T schemesThe most important step in the implementation of a dimensional control system is the validation of the final output and the analysis of the results. If the final product does not meet the desired overall dimensional specifications, the root causes must be investigated. Some of the causes are easily tackled and fixed whereas other might require detailed investigation. The following activities should be carried out to support fabrication, assembly, and erection processes. Critical data should be collected and statistically analyzed for further improvement in product quality. DesignDoes the design perform as intended? Tools/FixturesAre tools repeatable and user friendly? ProcessDoes the process appear to be capable? Validation of design Do parts within specifications build assemblies within specification? Do assemblies within specification build a product within specification? Do parts out of specification build assemblies within specification? Validation of Tools and Fixtures GR&Rgage repeatability and reproducibility Validation of process Can parts and assemblies within specification be processed in such a way that a completed product can be produced within specification? If not . . . Root cause analysis Are parts right? Are tools right and repeatable? Are datums valid (paper and physical coordination validation)? What does the data say (data quality validation)? If the above problems could not be fixed, this warrants revisiting the build strategy and redefining the objectives. After validating the final product with the design criteria, the data should be stored in a database for future use and update. These design and production data will be used to refine future processes.

MAY 2002

JOURNAL OF SHIP PRODUCTION

107

Definitions and examples


Portions of a ship double bottom with some pipe and machinery installed will be shown below to provide examples of some of the concepts discussed in the dimensional engineering process described. Build strategy A build strategy is represented by a physical product structure to the piece-part level, sometimes called an assembly tree. Figure 2 shows a portion of a product structure for a typical vessel. GD&T schemes Figure 3 shows a plate-to-fixture GD&T scheme. Figure 4 shows a plate-to-plate GD&T scheme. Figure 5 shows a stiffenerto-plate and fixture GD&T scheme. Figure 6 shows a stiffened panel GD&T scheme. Critical dimensions The critical dimension for this stiffened panel IP could be the transverse distance between the outboard plate edges, the measurement of each stiffener face from the edge of the plate (Fig. 7), and the flushness offset between the plates (Fig. 8). Dimensionally critical IP build/assembly process There are usually several ways to assemble any given IP. Each process of assembly might have unique consequences for dimensional variation. Following are two example alternative assembly processes for one of the stiffened panels in a double bottom. 1. An individual plate is brought to a station (fixture) and aligned to its datum. The next plate is brought to the same fixture and aligned to the joining face of the first plate. The first and second plates are joined together. A third plate is brought to the same fixture and its side is aligned to the joining face of the second plate. The third plate is then welded to the earlier welded plates. This subassembly is then moved to a different station (fixture) where it will be

stiffened. A stiffener fixture is used and is aligned on the assembled plate/station. The stiffener is then placed and aligned to the faces of the stiffener fixture and the plate and then fit and welded. In a similar fashion, the rest of the stiffeners are fitted and welded to the assembled plates. Figures 9 through 11 show this process. 2. An individual plate is brought on a station (fixture) and aligned to its datum. A stiffener fixture is brought and aligned on the plate/station. The stiffener is then placed and aligned to the faces of stiffener fixture/plates and fitted and welded. In a similar way another stiffener is brought and fitted and welded to the plate. On different stations, other individual plates are similarly stiffened. The first stiffened plate is then moved to a different station (fixture) where it is aligned to its datum. The second stiffened plate is brought to the same fixture and aligned to the joining face of the first stiffened plated. The first and second stiffened plates are joined together and so on. Figures 12 through 14 below show this process. Dimensional variation simulation analysis After assigning the design tolerances and production variances, dimensional variation simulation analysis is carried out. Example results are shown below. The purpose of simulation analysis is to determine the variation in process output with the intent of evaluating the ability of the process to consistently produce assemblies within the required specifications. The analysis simulates the assembly of a number of identical assemblies. The differences in assembly dimensions arise from the unique combination of various component and subassembly production variances from simulated build to simulated build. It is these different builds that generate the statistical results about the consistency of the assembly process. The output contains all the statistical data of the measurements associated with the assembly model. The sensitivity analysis shows whether a particular dimension is within the desired specification limits. It also shows various other statistical results (e.g., Std. Deviation, Ppk, Nominal, Mean, Median, Range, % Out). Sensitivity and GeoFactor analyses show the percentage contribution of each of the design/build tolerances that have affect on that particular di-

Fig. 2

Example of build strategy or physical product structure

108

MAY 2002

JOURNAL OF SHIP PRODUCTION

Fig. 3 Example plate-to-fixture GD&T scheme

mension. If the analysis results meet the build requirements, the final GD&T is created and production information is issued after assessment of the resources. Following are descriptions of many of the outputs from dimensional variation simulation analyses. Shown in Fig. 15 are the histogram and statistical results from the analysis of a particular critical dimension. Figure 16 shows the results of a sensitivity analysis, and Fig. 17 shows the results of a GeoFactor analysis. Histogram and associated statistics A histogram gives the distribution of measurements for a particular critical dimension for the number of builds simulated. The highest frequency value is displayed on top of the histogram as Max. 4 Highest Frequency. The Upper Specification Limit and the Lower Specification Limits are displayed as two lines on the histogram. The histogram bars outside the Upper and Lower Specification limits range are shown in red signifying that they are out of range. A histogram is generated for each critical dimension of the assembly model specified for measurement. RunsThis is the actual number of Monte Carlo simulation runs or builds that have been performed on the assembly. MeanThe critical dimension average value of the population generated by the simulation. NominalThe nominal value of the measurement is the value when all tolerances associated with the measurement are at their nominal values.

STDSTD denotes Standard Deviation (s ), which is a statistical measure of variation. Six sigma (6s )It is the Standard Deviation value multiplied by 6. This represents the 99.97% width of the normal curve, which is equivalent to a Cp of 1. Pp Pp is a process performance index, which measures the performance of the process. This index compares the variation of the process to the allowable variation that is set by the specification limits (USL and LSL). Pp is similar to Process Capability index, Cp . Outside of the context of Statistical Process Control (SPC) there is no difference between Cp and Pp , and only the term Cp is used. When sample data are in view, s is estimated by the familiar sample standard deviation that is technically denoted as s. It is within the context of SPC that the distinction between Cp and Pp becomes important. The difference is in the estimate of s that is used. In SPC, for Cp , s is estimated by s /d2 , which is the average range of the subgroups divided by a scale factor called d2 that is dependent on the subgroup size. For Pp , s is estimated by s. Since the 3DCS analysis software tool shown does not record measurements in numerous small subgroups, but rather one large group, Pp is more appropriate to use. The Pp index tells us about the capability of a process but does not tell us where the process lies within respect to its center. When the range between the Upper Design Limit (UDL) and Lower Design Limit (LDL) is the same as the 6s spread, the Cp index will be 1.0. The Pp index is calculated for a bilateral tolerance, there is no universally accepted way of defining Pp Index for a unilateral tolerance. Controlling factors for Pp are: design specifications, standard deviation, and sources of the deviation.

MAY 2002

JOURNAL OF SHIP PRODUCTION

109

Fig. 4

Example plate-to-plate GD&T scheme

Fig. 5
110 MAY 2002

Example of stiffener-to-plate and fixture GD&T


JOURNAL OF SHIP PRODUCTION

Fig. 6

Stiffened panel GD&T scheme

Fig. 7
MAY 2002

Some critical dimensions of stiffened panel


JOURNAL OF SHIP PRODUCTION 111

Fig. 8

More critical dimensions of stiffened panel

Fig. 9

Panel assembly process A

Fig. 10

Panel assembly process A

Pp k The Pp k indicator is a Process Performance index that represents how well the process is centered, or how distant the mean of the process is compared to its specification limit. The Pp k index can never be higher than the Pp index value. If the Pp k and Pp values are equal, the process is centered. If the Pp k value is 0, the mean of the process is centered about one of its specification limits. Controlling factors for Pp k are: design specifications, standard deviation, sources of the deviation, and central tendency of the process. MinThe minimum value of the measurement generated during a given set of process simulations for that particular measurement. MaxThe maximum value of the measurement generated dur112 MAY 2002

ing a given set of process simulations for that particular measurement. RangeThe range is the difference between the maximum value and the minimum value. LSLLSL is the Lower Specification Limit number set for that particular measurement. This comes from functional engineering requirements. USLUSL is the Upper Specification Limit number set for that particular measurement. This comes from functional engineering requirements. LoutLout is the actual percentage of measurement values falling below the Lower Specification Limit. RoutRout is the actual percentage of measurement values falling above the upper specification limit.
JOURNAL OF SHIP PRODUCTION

Fig. 11

Panel assembly process A

Fig. 13

Panel assembly process B

Fig. 12

Panel assembly process B

Fig. 14

Panel assembly process B

RngLoutRngLout is the actual percentage of statistical measurement values that fall between the Lower Specification Limit and -3STD. RngRoutRngRout is the actual percentage of statistical measurement values that fall between the Upper Specification Limit and +3STD. Sensitivity analysis Sensitivity or High-Low-Mean (HLM) analysis is executed to determine the sources of variation that contribute, in terms of percentage, to specific measurements/critical dimensions in an assembly model. The High-Low-Mean simulation varies each tolerance to its different levels, one at a time, while holding all other tolerances at their nominal values. The HLM tolerance level number governs the variation of each tolerancethe higher the number the greater the accuracy of the tolerance sensitivity for the particular measurement. A complete sensitivity report summarizes all the measurements associated with the assembly model and the different tolerances affecting each of these measurements in terms of percentage. GeoFactor analysis Similar to the sensitivity analysis, GeoFactor examines the effect of each tolerance on a given measurement. The distinguishing
MAY 2002

factor is that the GeoFactor analysis examines the tolerance contribution by geometry effect. To represent this effect the result is given as a GeoFactor. If a tolerance has a GeoFactor of less than one, then it has minimal contribution to the variation of a measurement point. If the value of the GeoFactor is greater than one, then it has a noticeable effect. In essence the GeoFactor number is the amplification of each tolerance in the model.

Conclusions
Dimensional engineering practices and tools used in other industries such as automotive and aerospace are highly evolved and also broadly applicable to shipbuilding. Therefore, U.S. shipyards have the opportunity to benefit from these well-established methods and tools. However, because of differences in product volume per design and variability in IP dimensional attributes, complete detailed dimensional engineering of each ship design is not economically viable. Alternatively, shipbuilding dimensional engineering practices should be focused primarily on developing parametric standards by IP type/group/family to be used in establishing the vast majority of GD&T for future ship designs. Detailed dimensional variation simulation analysis should be carried out only for those
JOURNAL OF SHIP PRODUCTION 113

Fig. 15 Example variance simulation analysis results for a single critical dimension

specific IPs that are determined to be both unique and dimensionally critical. At least one dimensional engineering company, Dimensional Control Systems, Inc., is evolving its analysis tools to allow for

easier modeling and simulation of the dimensional effects of weld shrinkage and distortion. This development work is being done to support the U.S. shipbuilding industry and is being partially funded by ONR through the SBIR program. Beta testers within the

Fig. 16
114 MAY 2002

Example sensitivity analysis results for a single critical dimension


JOURNAL OF SHIP PRODUCTION

Fig. 17

Example GeoFactor analysis for a single critical dimension KUMAR, R., AND SPICKNALL, M. 2000 Final Technical Report, DoD Navy N00-091, October. L EVINGSTON S HIPYARD AND I SHIKAWAJIMA-H ARIMA H EA VY I NDUSTRIES (IHI) 1980 Accuracy Control Planning for Hull Construction, U.S. Department of Transportation, Maritime Administration. SHEWHART, W. 1939 Statistical Methods from the Viewpoint of Quality Control, University of Washington, Department of Agriculture. STORCH, R. 1984 Accuracy control variation merging equations: a case study of their application in U.S. shipyards. National Shipbuilding Research Program, Ship Production Symposium Proceedings. YOKATA, T., MINAMIZAKI, K., HORI, S., SHIMOMURA, T., AND MIYAZAKI, M. 1963 A study of accuracy control in hull construction work. Japanese Society of Naval Architecture, 14, 242.

U.S. shipbuilding industry are being sought to support this development work.

References
ASME INTERNATIONAL 1994 GD&T Standard ASME Y14.5M. DRAKE, P. 1999 Dimensioning and Tolerancing Handbook, McGraw-Hill Professional Publishing. GRANT, E. AND LEAVENSWORTH, R. 1996 Statistical Quality Control, McGraw-Hill. I S H IK A W A JI M A -H A R IM A H EA V Y I N D U S TR IES (IHI) A N D C H I RI LLO , L. 1985 Process Analysis Via Accuracy Control, Revised, National Shipbuilding Research Program.

MAY 2002

JOURNAL OF SHIP PRODUCTION

115

Vous aimerez peut-être aussi