21-0089
21-0089
Page 1U.S. Department of Transportation Pipeline and Hazardous Materials Safety Administration 1200 New Jersey Avenue, SE Washington, DC 20590 December 10, 2021 Mr. Nick Talken Molecule Corp. Company Henkel 5110 Port Chicago Highway Concord, CA 94520 Reference No. 21-0089 Dear Mr. Talken: This letter is in response to your August 23, 2021, letter requesting clarification of the Hazardous Materials Regulations (HMR; 49 CFR Parts 171-180) applicable to hazardous materials classification. You state that your company has developed a software data platform which provides—for any new chemical formulation—the appropriate hazard classification and description. The output includes the UN Identification Number (UNID), hazardous materials shipping description, hazard class, and packing group (PG) as appropriate. The automatic assignment of hazard classification and description information is based on historical company data (e.g., in-house testing data) and you provided further detail on how this function is performed by the software. You seek feedback from this Office on your company’s software approach and technique for determination of hazard class and descriptions. In accordance with § 173.22, it is the shipper’s responsibility to properly classify and describe a hazardous material. There is no prohibition in the HMR against using a software approach as part of the process to perform the functions of classifying and describing a hazardous material; however, a software approach cannot act as a substitute for the performance of testing in determining a hazard classification, where it is required. It is our understanding that your software approach is an assistive tool and not a replacement for company testing in the absence of historical company test data that can be used as a comparison for a new formulation. Therefore, if your software approach—as described in your August 23, 2021, letter—provides the hazard classification and material description in compliance with the requirements of the HMR and international regulations for the transport of hazardous materials, then we see no issue with your company using this tool with regard to the shipper’s responsibility.#
Page 2Please note, this response letter is neither a validation nor a certification of your software approach. I hope this information is helpful. Please contact us if we can be of further assistance. Sincerely, Dirk Der Kinderen Chief, Standards Development Branch Standards and Rulemaking Division#
Page 3Baker From: Foster, Glenn (PHMSA) To: Date: Dodd, Alice (PHMSA); Hillman, Kenetha CTR (PHMSA) Subject: FW: Henkel Company Correspondence Monday, August 30, 2021 12:23:15 PM Attachments: EO - Kelley(Henkel) 30Aug21.pdf 21-0089 Alice and Kenetha, Please log this in as a request for a LOI and assign. Thanks, Glenn From: Kelley, Shane (PHMSA) <shane.kelley@dot.gov> Sent: Monday, August 30, 2021 12:20 PM To: Foster, Glenn (PHMSA) <Glenn.Foster@dot.gov> Subject: FW: Henkel Company Correspondence Can we log this as an interp? I’d like to respond that we do not review software as a function of our Office, but that automated methods of classification are acceptable provided that the resulting classification is accurate and conforms to the requirements of the HMR. Something like that. Can be short and sweet. From: Lisak, Frank (PHMSA) <frank.lisak@dot.gov> Sent: Monday, August 30, 2021 12:00 PM To: Kelley, Shane (PHMSA) <shane.kelley@dot.gov> Subject: Henkel Company Correspondence Hi, Shane: The attachment is for your review. I will keep the hardcopy in my locked cabinet until DOT resumes normal working hours. Frank#
Page 4Albert - Open Letter to Department of Transportation - United States of America Date: August 23rd, 2021 Director, Standards and Rulemaking Division Mr. Shane Kelley U.S. DOT/PHMSA (PHH-10) Washington, DC 20590 1200 New Jersey Avenue, SE East Building, 2nd Floor Dear Mr. Kelley, formulations and materials. Specifically, this data platform was built to capture structured data which can be used to accelerate innovation and Over the paşt 4 years, we have been developing a software data platform (called Albert) which facilitates our ability. to invent new chemical formulation and design of experiment, task management, reporting and visualization, and many more. However, the intent of this Open Letter is enhance collaboration between chemists or laboratories. Albert has many capabilities and functions, include chemical inventory management, to describe a specific feature of Albert that is related to the shipment of Hazardous Materials (HM) / Dangerous Goods (DG). for each relevant transportation regulation (49 CFR, IATA, IMDG). The output of this classification includes: UN Number, Hazard Materials Within the Albert platform, any chemist who has created a formulation can automatically receive the appropriate HM / DG Shipping Classification based on all historical data within Albert, where the newly created formulation is compared to historical formulations, to ensure compliance with Shipping Description, Hazard Class or Classes and Packing Group as appropriate. This automatic assignment of HM / DG shipping information is the shipping regulations. The remaining content within this letter will describe in detail how this comparison and automatic assignment is done. Historical Data: the number of formulas which have been created, per Transportation method: Within Albert, we have over 20 years worth of historical formulas that have been commercialized and shipped within the USA. Below you can see Transport Method Count of Unique Formulas 49. CFR 26449 IATA 29699 IMDG 30147 you can see the specific UN Numbers and Packing groups that are contained within this dataset. These Classifications were all performed For all of these historical formulations, a Risk Assessment and HM / DG Classification was performed by a certified expert within Henkel. Below manually by a certified Risk Assessment expert team, primarily located in Dusseldorf, Germany. *Classification Format: (UNnumber_PackingGroup)* TRANSPORT METHOD: CFR 100000 17852 10000 1461 1364 1000 872 729 654 Count of Unique Formulas 430 302 271 263 250 217 211 175 168 153 152 146 130 130 130 100 107 100 96 86 10 1 Not Regulated UN3082II UN1133_]1 UN1993_1 UN3264_| UN1814_1 UN285 I UN1824,1 UN3264_II UN2922 _II UN1139_1 UN1823 _I UN2735_II UN3262_]I UN2735_]1 UN1263_1 ÷ 1: → 118G he ture co Classification#
Page 5TRANSPORT METHOD: IATA 100000 19085 10000 3096 1409 1198 1000 749 510 477 Count of Unique Formulas 306 277 267 258 228 223 177 172 165 154 151 133 133 100 132 112 103 98 86 10 1 Not Regulated UN7264_1 4UN3334II U280 I UN1824_ UN3264 I UN1139.| UN2922_11 UN1823_]| UN2235_II UN3262_JI UN2735_]! UN1993_! UN1760_1 UN1719_ UN279_| Classification TRANSPORT METHOD: IMDG 100000 21147 10000 1999 1462 1000 763 707 635 Count of Unique Formulas 306 276 267 265 230 225 190 177 177 165 163 142 134 100 131 126 110 96 90 82 82 10 1 Not Regulated UN1133_" UN3264_|| UN1950 UN1814.1 UN2922_" UN3262,1 UN2735 " UN 1866_11 UN3266_I| Classification Automatic Generation of Classification for New Formulas: Algorithm Similarity: factors, including GHS classification, Flash Point, Reaction Exotherm, Product Type, State of Matter, Chemical Composition, and others. The It is very important to understand how the similarity is determined by the above mentioned Algorithm. This similarity is a combination of many purpose of the Algorithm is to determine the best combination of factors (mentioned above) which allow for the closest formula comparison. Validation of Algorithm:#
Page 6data. First, we split the historical data into 2 partitions, with 80% in the "training set" and 20% in the "test set". For each of the transportation In order to ensure this method of "lookup" is accurate for a wide variety of new potential formulas, we performed statistical testing on the historical DG classification to the actual classification from the "test set". The results of this test ensures complete accuracy across all historical and new methods, we took the "test set" and used algorithm with the "training set" to determine the HM / DG classification. We then compared this HM / data. We understand that compliance with these regulations is paramount at all times from the beginning to end of the classification and transport cycle. Open Questions: 1. Based on the information provided, assuming the above information is accurate, we would like to solicit any input from the Department on our approach. 2. Please provide any other feedback on our software driven approach and techniques. 3. Please provide your position on our methods stated above, given that it has been shown to be more accurate than manual classification. Thank you and we look forward to your response. Sincerely, Nick Talken Ken Kisner Head of Albert Data Platform Head of Digital Innovation nick.talken@henkel.com ken.kisner@henkel.com (925) 381-6796 Nick Talken (925) 381-1253 Ken D. Kisner#
This material provides agency context. It does not replace binding regulatory text, and its legal effect depends on the underlying authority and facts.