Risk is inherent in operating ships at sea. Sometimes it manifests itself in tragic errors that impact both ships and careers. On 12 February 2014, the USS Taylor (FFG-50) ran aground in the Black Sea. The commanding officer was relieved.1 On 21 June 2019, the littoral combat ship USS Billings (LCS-15) got under way from a berth in Montreal and allided with a merchant vessel.2 Again, the commanding officer was relieved. But, as with these and many similar incidents, determining the root cause and what could have been done prior to the accident to prevent it can be difficult. This is especially true in determining the role human factors played.
In fact, the Navy has learned from these and similar events and made adjustments based on the lessons, even instituting in the surface force a formal policy for reporting, analyzing, and sharing near misses.
But what if we could look at the precursors of a potential mishap and model the risk based not so much on the individuals involved, but instead on the planned evolution? What if we could posit that the average risk of a particular evolution is “X,” and a predictable level of stress and fatigue of “Y” increases the risk of a mishap by a factor of “Z”? This might actually be possible, applying a set of tools currently used in academia but with potential utility for the operational Navy.
IMPRINT
Improved Performance Research Integration Tool (IMPRINT), developed by the Army Research Laboratory, was first released in 1995, under the name Integrated MANPRINT Tools. IMPRINT uses cognitive task analysis to examine workflow models and has been used extensively in military applications, as well as by NASA, to model small-team performance. Once a team is built, its workflow can be modeled, the results analyzed, potential failure points identified, and mitigations applied. In 2010, for example, the Army wanted to reduce the crew size of the M-1 Abrams tank from four soldiers to two. An IMPRINT workflow and cognitive task analysis showed that doing so increased the risk of failure (the tank destroyed and crew killed) from 7 percent to 35 percent. This was deemed unacceptable, and the idea was scrapped.3
In the wake of its 2017 Comprehensive Review of Recent Surface Force Incidents, where the key events shared many similar factors (i.e., crossing a traffic separation scheme at night, with increased levels of stress and fatigue), the Navy has focused on small-team performance in general and while under increased stress. Cognitive task analysis could be another tool to model Navy small-team operations and maintenance procedures, identifying probable failure points and mitigations. The challenge is how to move this process into the “predictive” and allow analysts to plot potential failures related to human factors on a standard operational risk matrix to quantify the risk of a particular failure or mishap under given conditions—a “human factors risk analysis” (see figure 1).
IMPRINT and the LCS
I recently completed a research project using IMPRINT to determine a model for the optimized manning and scheduling tool for an LCS engineering watch team. The focus areas were maintenance tasking (planned and corrective), watch tasking, fatigue, and automation. At the outset, the crew was given a 25-question survey. The majority of respondents felt they had both stood watch and performed maintenance while fatigued (73 percent) and had more maintenance- and watch-related tasks than they could accomplish (67 percent). Surprisingly, the majority (65 percent) felt they could trust the increased automation of their ship and that it made their jobs easier than they had been on other ships. Of note, this survey was conducted prior to the 2018 implementation of circadian-sleep-cycle watch rotations and combining the LCS module and crew manning for a total of 70 personnel (versus the previous 40). Survey answers in 2021 could be markedly different.
The next step was building workflow models. We used planned maintenance requirement cards, engineering operational sequencing support procedures, and the current ship’s maintenance project to build models of a representative set of procedures that crew members are expected to perform in the course of a normal watch or maintenance period. These models were inserted into a four-hour watch period using a reasonable sequence and sampling, then subjected to a standard failure mode analysis.
The IMPRINT software program produced a series of data tables and graphs that showed a few interesting things:
- Based on workflow analysis, the program showed what procedures were most likely to result in mission failure, even down to the step that caused the failure.
- The fatigue and stress model showed areas in which fatigue and stress increased risk of mission failure and what procedures were most affected by operator fatigue.
- The automation model showed the reduced workload on operators because of automation, but made some assumptions as to the reliability of automated features that may bear closer scrutiny.
IMPRINT is very flexible, allowing the programmer to assign specific tasks to each operator to adjust parameters (called “taxons”) such as training, experience, stress, and fatigue level. For a post-facto case, such as the Billings and Taylor incidents, a similar process for a bridge watch team is a possibility. These models can be built on the basis of a notional crew, using normalized inputs for training, experience, and watch rotation (circadian or not), and then tailoring them for specific teams if individual human factors data, such as the examples listed above, are made available. But how to apply the model to the future?
Predictive Modeling for the Fleet
This process could help the Navy improve manpower and operational risk decisions, if the following steps are taken:
1. Combine the lessons from Naval Postgraduate School (NPS) studies on small-team performance with Navy Manning Analysis Command and Center for Naval Analyses studies, apply them to programs such as IMPRINT or something similar, and merge existing models into one Navy model.
2. Build a catalog of procedures for modeling. One significant contribution of past NPS research has been the accumulation of a small library of missions with repeatable processes the Navy could leverage.
3. Run models for new ship classes using lessons from the LCS. The newly awarded Constellation-class frigate could even use such analyses as part of the design process. By incorporating NPS research data on circadian-sleep-cycle watch rotations, it is possible that a four-section watch team basis is achievable.
4. Move the cognitive task analysis operations model from the classroom to the fleet. Adding data and feedback from crews will validate the model and ensure real value is realized. At this point, it could be used along with the operational risk management program to capture risks associated with manning gaps, fatigue, and overtasking.
5. Build reporting systems to use the risk matrix shown above and clearly articulate the risk in terms of equipment damage or human injury and death based on modeling. Require meaningful impact statements.
If the Navy wants to transform from its current “backward-looking” processes and move to the “forward-looking” domain of artificial intelligence, predictive maintenance, and gaming technology, using workflow modeling programs such as cognitive task flow modeling to predict failure points for small crews could save money, careers, and possibly lives. The Army and NASA have a head start in this area. The Comprehensive Review was tough on the Navy in terms of shortfalls in the human-systems engineering process. Workflow modeling presents an opportunity to improve.4 The Navy can no longer maintain the status quo of taking too much risk and then being surprised when doing so results in an accident, especially when sailors pay the price in terms of catastrophic equipment damage, serious injury, or even death.
1. Stephen Beardsley, “Taylor CO Relieved after Ship Runs Aground,” Stars and Stripes, 25 February 2014.
2. Sam LaGrone, “LCS Billings Commander Removed After Hitting Merchant Ship in Montreal,” USNI News, 29 June 2019.
3. Ming Mao, Fang Xie, Jian-Jun Hu, and Bo Hu, “Analysis of Workload of Tank Crew Under the Conditions of Informatization,” Defense Technology 10, no. 1 (March 2014): 17–21.
4. Renaldo N. Hollins and Kelly M. Leszczynski, “USN Manpower Determination Decision Making: A Case Study Using Imprint Pro to Validate the LCS Core Crew Manning Solution,” master’s thesis (Monterey, CA: Naval Postgraduate School, 2014)