Process Stability and Control

Accuracy in all processes is critical in shipbuilding. It has been said that:

“The successful application of accuracy control techniques to shipbuilding is quite fundamental to achieving high levels of productivity.”

(Dr R Vaughan, Productivity in shipbuilding, Trans NECIES Volume 100 1983-84)

We are all aware that perfection is not possible, especially when it comes to applying coatings on board ships. This is a result of many factors that create an inherent variability in the process, making it difficult to control. (C Jefferson, R Kattan. Accuracy Control Training Manuals, British Shipbuilders 1987)

The variability of any process comprises two elements:

  • Assignable causes
  • Random variations

Examples of these for coatings application work may be:

Assignable Causes

  • Use of the wrong or a worn tip
  • Using the wrong pressure
  • The wrong stand-off distance
  • The addition of a cosmetic coat to the scheme
  • The size of the atomized particles
  • The workers' physical capabilities
  • Wind gust and temperature changes
  • Available air pressure

Random Variations
While assignable causes can be addressed and managed, the inherent variability of the process can only be improved by changing process technologies and practices. So, as many assignable causes as possible must be identified and managed.

Control Chart Method

The assessment of process stability/performance is often made using control charts. There are a variety of control charts.

The key element of the control chart is the setting of upper and lower tolerance limits and a mean value (normally the specified or target value).

The ideal situation is where specification limits or tolerances are set outside the capability limits. If this is the case, the process can be said to be capable of undertaking the work to tolerance.

If, however, the specification limits are inside the upper and lower capability limits, then the process would not be capable of carrying out all work to within the required tolerance.

Obviously, the further inside the capability limits that the specification limits can be set, the greater the probability of being able to meet the specification.

It is normal in most processes for the tolerance limits to be set around a mean value, e.g. a typical value may be 150 µ ± 50 µm.

The Impact of Minimum/Maximum DFT Values

The practical impact of the minimum and maximum values needs to be understood. First of all, a number of samples of coating inspections have been taken to get a general understanding of what is practically achievable.

Control of DFT is dependent on many factors such as:

  • Worker skill
  • Equipment
  • Access considerations
  • Complexity of the structure to be coated

Performance is likely best on a flat surface (Coating of Ships: the Design Challenge by Broderick, Kattan, Wright RINA Conference 2010), while more complex surfaces will tend to increase the range. The size of the area to be coated will also have an influence.

Data presented by Francis 2013 indicated that for small areas, the set of readings on a nominal DFT of 85 µm for a single coat of inorganic zinc silicate would give a range of 240 µm, with a minimum of about 20 µm and maximum of about 260 µm.

U.S. Navy Underwater Hull Retrofit Projects
A larger area work, carried out on the underwater hulls of U.S. Navy carriers (Whitiker, Wimmer and Bohlander, as reported by Francis R A), gave the following results:

USS Nimitz

Specified scheme DFT

680 µm


720 µm

Standard deviation

220 µm


Approximately 1000 µm

Ratio of standard deviation to the mean


Process capability to 3 s

60–1380 µm

USS Lincoln

Specified scheme DFT

680 µm


850 µm

Standard deviation

250 µm


Approximately 1000 µm

Ratio of standard deviation to the mean


Process capability to 3 s

100–1600 µm

An assessment of the U.S. Navy figures cannot be made without knowing the specification for the system that was applied (e.g. how many coats of paint). Also, these coating projects were in a refit scenario, rather than a new-build scenario.

Outer Hull New Build Commercial Ship
Figures from Safinah case studies for the outer hull of a new build commercial ship for the same ratio gave a mean of 0.18 for the ratio of standard deviation to the mean (also known as the coefficient of variation):

Specified DFT

610 µm

Average DFT

990 µm

Standard Deviation

170 µm

Process capability to 3 s

480–1500 µm

Thus, even on the relatively uncomplicated area of the underwater hull, there is considerable range of achieved quality of the application, with the ratio of standard deviation to the mean ranging from a relatively good 0.11 to a relatively poor 0.44 for the U.S. Navy.

In simple terms, the closer the standard deviation value is to the mean value (the higher the ratio), the greater the spread of the curve. Hence, the more likely you will get over or under application, because the process is not well controlled for a number of reasons such as:

  • Weather/wind conditions
  • Worker skill
  • Equipment capability/maintenance
  • Roughness of the surface to be coated

This would indicate that a process that isn’t well controlled results in excessive over-application of coatings, which can penalize a yard in a number of ways:

  • Increased cost of paint and thinners/cleaners
  • Increased application time
  • Increased curing/drying time
  • Increased emissions
  • Increased waste
  • Delay to build schedule
  • Increase in utilization of facilities

Cargo Holds / Safinah Data
In more complex areas, Safinah research has shown these results for cargo holds gave a mean of 0.19 for the ratio of standard deviation to the mean:

Specified DFT

250 µm

Average DFT

649 µm

Standard Deviation

133 µm

Process capability to 3 s

250–1048 µm

Thus, while the process for cargo holds does show a greater variability (higher standard deviation) than that for the outside shell, the ratio of mean to standard deviation is about the same (0.18 to 0.19).

The reason for this is relatively simple. The outer hull scheme typically comprises four or more coats of paint as compared to two coats in the cargo hold. The variability in the DFT of each coat is additive; thus the more coats of paint applied, the greater the variability that will need to be contained.

Thus the more steps in a process (i.e. the more coats of paint in the scheme), the greater the variability that should be expected, irrespective of the complexity of the surface to be coated.

Ballast Tanks / Safinah Data
The following results for ballast tanks, which are also generally two-coat schemes over more complex areas, should therefore offer a better comparison to the cargo holds.

These figures give a mean of 0.26 for the ratio of standard deviation to the mean:

Specified DFT

320 µm

Average DFT

602 µm

Standard Deviation

162 µm

Process capability to 3 s

116–1088 µm

Thus for ballast tanks, despite having only two coats of paint, both the standard deviation and the ratio are considerably higher as a result.

This would imply that design complexity has a much greater influence on the variability of the coating process than the number of coats.

This would also imply that to maximize the probability of a good coating application, both the design complexity and the number of coats should be minimized. It would also imply that simplification of design would offer the greater benefits.

Other Variables
In practice, the problem is aggravated further because not all the coating work in one location will be carried out by the same team. In fact, there may be more than one team working on each area and the skill/ability and equipment, as well as local conditions, may vary considerably.

Of course, the figures are also likely to change for different ship sizes, with smaller vessels providing more complex/tighter structures.

The authors suggest that perhaps the Compensated Gross Tonnage Coefficients (OECD Council working party on shipbuilding Compensated Gross Ton (CGT) system) could be considered for use to establish the complexity of different ship types and sizes.

Impact on a Coating Scheme

Consider a specification of 2 x 160 µm as required by the IMO PSPC and as shown on most paint supplier data sheets. In this case, the value on the Technical Data Sheet (TDS) is not the “nominal value”, which the authors have interpreted as a target value or the mean/average.

Maximum & Minimum DFT
Good practice from paint company guidelines would mean that the maximum DFT applied should be x2 the specified DFT for each coat and for the total scheme. These would give a maximum scheme of 2 x 320 µm.

However, applying the 90:10 rule or the 80:20 rule would give minimum values of:

  • 90:10 rule: 2 x 144 µm or 288 µm total
  • 80:20 rule: 2 x 128 µm or a 256 µm total

The standard deviation for water ballast tank application has been derived at 162 µm. Thus, if the minimum acceptable value is 288 µm as per the IMO PSPC, then three standard deviations would provide a mean of 774 µm [given by: 288 + 3(162) µm] and the maximum value that could be expected would be 1260 µm [given by 774 + 3(162) µm].

The mean that is likely to be achieved will itself exceed the recommended guideline of most paint suppliers, which is set at x2 the specified DFT (640 µm in this case) and also surpasses the x3 value in ISO 12904.

To achieve the required specification:

  • Minimum 288 µm
  • Maximum 640 µm

Here then, the standard deviation would have to be 58.7 µm or about 36% of that being achieved in the field based on the Safinah data.

The problem of a high achieved DFT is compounded even further if, during the inspection, areas of low DFT are identified (e.g. an area of 250 µm). If it is then touched up by airless spray, it will not be bought up to 288 µm or 320 µm, but more likely by an additional 160 µm to 410 µm, thus compounding the over-application problem.

If the coating is applied by brush, an additional 80 µm could be added. Therefore, any application of “build” coats to achieve the minimum DFT is likely to increase the mean DFT and push the scheme further out of the recommended guidelines provided.

Practical Distribution

In practice, however, the applied DFT data does not result in a normal distribution but a skewed distribution.

An actual set of data from a ballast tank is presented below. This WBT coating was specified according to IMO PSPC and thus should have a nominal DFT of 320 µm. What the analysis of the data revealed was:

Total number of readings


Minimum DFT

272 µm

Maximum DFT

1326 µm


1100 µm


611 µm

Standard deviation

Not relevant


564.5 µm

Breakdown of Readings in Bands of Microns:

The Impact of Minimum & Maximum DFT Values on Coating Performance

Given that the recommended practice would give a maximum of 640 µm, then 193 (34%) of the readings taken exceeded the maximum, while very few readings were below the minimum—this despite the mean and the mode being below the 640 µm maximum.

Thus the actual distribution of the DFT readings will be greater than the expected specified values. In particular, the actual distribution will tend to skew toward higher DFT values, and this is aggravated by the use of a minimum DFT rule.

As soon as a minimum rule is introduced, then the mean DFT achieved will end up being considerably higher than the specified DFT. This, combined with a complex space, results in the mean DFT being close to or greater than the x2 DFT maximum provided in paint company guidelines.

This article was co-written with John Fletcher.

With more than 45 years’ experience in the corrosion, protective coatings and electronic inspection technology fields, John Fletcher serves as technical support manager at Elcometer Ltd. in Manchester, England. He is the current president of the Institute of Corrosion (iCorr), and chairman of ASTM International Committee D01 on Paint and Related Coatings, Materials and Applications. As a top international expert in paint testing and inspection methods, Fletcher also leads Subcommittee D01.23 on Physical Properties of Applied Paint Film.

More in the "The Problems with Meeting Marine Paint Specifications" series:

Working to Overcome Poor Quality Coating Practices in International Shipbuilding
Ambiguities in Determining if 'Specified DFT' has been Achieved
Increasing Coating Performance through Application Process Control