Background

The National Fire Protection Association (NFPA) recently undertook work to consolidate the multiple standards addressing the requirements for handling combustible dust. The consolidated document will be published as NFPA 660, likely being released in 2024. NFPA 660 is touted as an improvement over the existing set of standards because it will centralize the relevant requirements in a single document, and it will provide improved consistency across industries handling different types of dusts.

In the meantime, any organization that handles or processes combustible dust must comply with NFPA 652 (Standard on the Fundamentals of Combustible Dust) and the commodity standard that applies to the organization's industry. One example of a commodity standard is NFPA 61, which applies to organizations working with agricultural and food-based dusts (Standard For The Prevention Of Fires And Dust Explosions In Agricultural And Food Processing Facilities). The need to consult multiple standards can lead to confusion, especially when conflicts exist amongst them.

Analysis

We recently wrote some MATLAB code to compare the existing standards in order to provide a visual representation of their similarity. The code can run on any arbitrarily large portion of the standards, but for illustration, the figure below compares the NFPA 652 and NFPA 654 requirements for bucket elevators.

For each requirement of NFPA 652 (the rows of the grid), the code performs a textual comparison and identifies the NFPA 654 requirement with the best agreement. Keep in mind that even if all the options are bad, it will still come back with a "best".

The case with the best agreement is marked with a score, where smaller numbers indicate better agreement. For example, a zero score indicates perfect agreement between the corresponding NFPA 652 and 654 requirements. Darker colors also represent better agreement.

We considered adding a tooltip that would be populated with the corresponding NFPA 652 and 654 requirements so they could be easily compared, but were certain this would get us into copyright trouble.

A few observations are in order:

• The listings of zero values readily show which requirements are identical.
• NFPA 654 has some requirements for bucket elevators which are unique from NFPA 652. The unique requirements are visible in the columns which show no match. For example, NFPA 654 9.3.10.1.1 deals with explosion protection requirements.
• The correct way to interpret the numbers isn't apparent -- see the next section for more comments.

Interpretation of the Numbers

So, what do the numbers actually mean? What number constitutes significant similarity? The answer is that "it depends". The calculated numbers represent the "distance" between the text of each requirement being compared, which is the number of single-character edit operations that convert the text from one requirement into the other. These calculations were performed with a code published by Miguel Castro, and the single edit operations include deletions, insertions and/or replacements. It's clear when you consider the comparisons that the number of single edit operations may scale with the length of the requirement text. It may be useful to normalize the numbers by the text length of the requirement.

But, as an illustration, consider the comparison between NFPA 652 9.3.14.1 and NFPA 654 9.3.10.2, which are separated by a distance of 37. We're still waiting for copyright approval from NFPA to excerpt the text to show the discrepancy, but the reader will eventually find that there is no substantial difference between the two. NFPA 654 favors use of the term "dusttight", while NPFA 652 provides a functional description for what it means to be dusttight. A reasonable reading and interpretation of the two requirements indicates they are identical, but a liberal interpretation may disagree.

One thing that's evident from this comparison is that any number greater than zero is somewhat arbitrary. It's useful to know there's a perfect match, but one could imagine other replacements for "dusttight" in this example which would have increased or decreased the distance between the requirements.

Nevertheless, the analysis is useful for quickly visualizing similarities and differences between a set of standards. These visualizations may still be useful after the publication of NFPA 660, but for making comparisons between chapters instead of unique standards.