Mitigation of Chemical Reporting Liabilities through Systematic Modernization of Chemical Hazard and Safety Data Management Systems

Abstract

In the United States alone, approximately 2 billion tons of hazardous material products are manufactured each year for both household and industrial applications and contribute to thousands of worker chemical exposures with as many as 50,000 deaths from prolonged exposure each year. The potential hazards and impacts of these chemicals for human health and the environment are primarily communicated to the public through Safety Data Sheets (SDSs) from the chemical vendors or distributors. These documents provide a standardized approach for how and what information is provided to product users to assist them with assessment of precautionary measures, hazard mitigation, emergency response or cleanup procedures, and environmental, health, and safety (EHS) management. Despite the criticality for hazard communication (HAZCOM) precision, legacy SDS management and industry business practices leave the overall ability to effectively manage chemicals vulnerable to significant liability through a lack of full constituent disclosure, injection of data quality errors through various handling of SDS information and manual data entry, and the lack of direct SDS-to-product association. Chemical spills and accidents often require individuals to look for the appropriate SDS on a local computer, online, or in workplace binders; each of which results in information returned that is often found to be outdated or incorrect. Workplace HAZCOM violations remain among the top citations during EHS inspections by regulatory agencies. More important, however, is the lack of precise association of SDS to hazardous products that can occur through chemical management lifecycles. Incorrect SDSs can yield significant liability, as subsequent environmental and occupational health analyses and reporting are based upon incorrect and, in some cases, entirely different chemical formulations. This paper focuses on the need for a paradigm shift in our chemical management systems and how a standardized management system and various recent technological advances can be incorporated into Environmental Management System operations to reduce or eliminate these liabilities. The following advancements can be used to enhance the lifecycle management of workplace chemicals, reduce potential exposure and spill risks, reduce workplace hazards, and increase the efficiency and accuracy of environmental reporting through a more streamlined systems approach. EHS system enhancement applications discussed in this paper include the following: the need for a centralized universal SDS repository with full chemical disclosure of all product constituents and a nationally adopted machine language SDS standard. The use of artificial intelligence/machine learning in environmental systems and how they can be used as a medium to transition toward an automated standard by reverse-engineering and partitioning SDS components into machine-encoded text that can be validated and uploaded to a centralized repository. Algorithmic and meta-algorithmic approaches to SDS requirement and data validation, hazard characteristic code calculations, and determination of potentially less hazardous substitutions. Application of Natural Language Processing methods for real-time updates from scientific journals, regulatory agencies, and other reputable sources to produce “living” SDSs capable of informing users of relevant regulatory updates, news, and research. Embedded SDSs or SDS links in product barcodes with QR code reader technology to retrieve precise SDSs for each product in emergency situations. Use of advanced QR codes embedding authentication layers, authenticity verification, and alerts of potential product or inventory problems or discrepancies. Benefits of radio frequency identification technology in providing accurate SDS associations while also minimizing manual tracking of hazardous material and hazardous waste containers and monitoring for expired shelf life, incompatible storage, temperature sensitivities, and other inventory concerns.

1. Introduction

1.1. Hazard Communication Breakdown

Over 2 billion tons of hazardous and toxic chemicals are manufactured in the United States each year.¹ According to the 2019 U.S. Environmental Protection Agency (EPA) Toxic Release Inventory (TRI), U.S. Federal Facilities alone managed a total of 30.7 billion tons of TRI-listed chemicals and production-related waste during 2019.² Of the 30.7 billion tons managed, 11%, or 3.38 billion tons, were released into the environment (proportions released to air, water, and land). In addition to environmental concerns, each year in the U.S., thousands of workers become sick from workplace chemical exposures with as many as 50,000 people dying each year from the adverse effects of long-term chemical exposure.³ One of the primary factors in both the documentation of toxic releases and in the proper assessment of workplace exposure hazards is the vendor-communicated Safety Data Sheet (SDS) information, which provides the chemical ingredients, composition information, physical and chemical attributes, and hazard information needed to derive such calculations (Figure 1).

Figure 1.

2019 Toxic Release Inventory data and locations.

The current hazard communication (HAZCOM) system allows for opportunities of missing and/or incomplete data and various errors through the processing and handling of data through the chemical lifecycle. These issues, in turn, lend themselves to other problems including our ability for precise SDS selection, fast retrieval of SDSs during emergencies, and accurate cradle-to-grave tracking of these hazardous containers. These problems include the following:

• missing SDS data from vendors

• non-GHS compliant SDSs

• SDS data errors

• manual data entry errors in downstream user databases

• costs associated with manual entry of data in downstream systems

• confusion associated with SDSs reproduced by suppliers

• lack of product-to-SDS association

• incorrect reporting due to tracking inefficiencies

1.2. Lack of SDS Accessibility, Precision, and Data Quality

Correct SDSs are often difficult to find due to numerous variations of similar products, variations among manufacturers/vendors/distributors, and hosting of SDSs spread across thousands of chemical manufacturer domains. Due to these inefficiencies, workplace users often receive violations regarding incorrect SDSs, outdated SDSs, and in some cases no SDSs for the chemicals they are using. Perhaps of greater concern is the liability these issues pose in the form of incorrect chemical calculations for environmental compliance and incorrect employee chemical exposure assessments⁴ resulting in inappropriate personal protective equipment. For example, chemical usage for a product with a unit of measure of a gallon (Gal) is calculated by⁵

1

An incorrect SDS has the potential to alter the ingredient list, chemical composition, and specific gravity (SG) (or density); yielding incorrect usage calculations for the total number of product containers associated with that SDS. Liabilities not only exist for management of the products themselves but also for treatment, cleanup, and disposal. SDS content can be used in support of tort claims in civil litigation suits and in determining where the fault lies depending on how the hazardous content was reported on the SDS and how the employer used and managed that information.⁶ In the case of Tolley and Tolley v ACF Industries, the SDS was used as evidence to support claims that the employer was aware of an isocyanate hazard.⁷ The term “SDS” resulted in 814 citations in the legal case repository CaseText.com, reflecting the importance of HAZCOM in regard to liabilities for manufacturers, employers, and chemical users. SDSs are also used for user-knowledge hazardous waste characterization, which could result in incorrect waste treatment or disposal with large-scale contamination concerns of landfills and other processing facilities.

In order to retrieve SDSs for chemical products used in the workplace, chemical users are largely responsible for finding the appropriate SDS that corresponds with each product used. SDSs are commonly housed on manufacturer websites and differ from site to site (including seasonal formulations) with further confusion often introduced as distributors can create their own SDSs, making product matches difficult. Lastly, not all chemicals in product formulations are required by the manufacturer on the SDS. Instead, only those that are deemed as hazardous chemicals by the manufacturer.⁸ As chemical research progresses and new chemicals are added by the EPA as emerging contaminants,⁹ liability exists in the now-hazardous chemicals that were previously omitted from older SDSs because they were not deemed hazardous at the time of SDS creation. As a study in 2002 showed, limitations to Material SDSs existed since Occupational Health and Safety Administration (OSHA) permitted chemical exclusions when the manufacturer deemed it as non-hazardous or protected as a trade secret.¹⁰ These limitations hold true today. Per- and polyfluoroalkyl substances (PFASs), for instance, were added under the EPA Toxic Substances Control Act¹¹ and TRI¹² in 2020. Prior SDS for products containing these chemicals largely omitted these making it now difficult to quantify PFAS inventory and usage in the workforce.

1.4. Lack of Immediate SDS Retrieval Abilities

U.S. OSHA listed failures in HAZCOM as the second highest most frequently cited standard.¹³ In total, 3624 HAZCOM enforcement citations occurred in the fiscal year 2019 totaling approximately $4,682,380 in proposed penalties.¹⁴ Unfortunately, many current procedures still include maintaining binders of printed SDSs or available copies downloaded on neighboring workstations. These methods are consistently found during inspections to include numerous records that are the incorrect SDS for specific products or are out-of-date, incomplete, and illegible. For environmental, health, and safety (EHS) systems that require the loading of SDS for exposure and environmental reporting calculations, non-EHS personnel are frequently used to find the correct SDS for the products they are using and often not provided sufficient training to determine whether SDSs are GHS compliant and the proper version for the product. OSHA recommends employers “designate a person(s) responsible for maintaining SDSs”; it does not reference training for the nuances of precise SDS selection and management.¹⁵

1.5. Loss in Inventory Accountability

Lastly, inventory accountability and management are difficult to maintain, as shown in a recent study of chemical storage at research laboratories.¹⁶ For hazardous material users, maintaining accurate inventory counts, SDS-inventory associations, proper material segregation, and usage logs can prove an arduous task provided that for many users these are side compliance tasks to their primary duties; often what the materials are being used for. While many of the proposed applications in this paper are not inherently novel, the applications of these in industry, individually and as part of a more robust, integrated system, remain uncommon and their use could help improve or resolve many of the more common EHS violations.

2. EHS Technology Optimization

2.1. Universal SDS Database and Repository

Benefits: Single accessible source of chemical data for all chemical users, elimination of separate duplicative systems across the industry, product comparisons for safer product selection, personnel and time savings from researching SDSs across thousands of vendor websites, data quality control, point source for systems to electronically access SDS data, and the singular source for precise SDS access methods.

One change that would provide immediate benefit on many of these issues is also the most difficult to implement and would require a paradigm shift in the regulatory compliance measures imposed on chemical vendors. This shift would be for a single universal SDS repository for all chemical users. Although some repositories such as SDS.com exist, existing SDS repositories are not all inclusive, can require membership payments, and have varying levels of quality assurance. A single national or universal SDS repository would allow (and require) a standardized SDS formatting directly from manufacturers into one centralized database that is accessible by any downstream user requiring SDS information (Figure 2). Standardized values and format would, in turn, produce a significant reduction in varying or missing information, a reduction in back-and-forth communication between data users and manufacturers, a significant time and monetary savings in accessing thousands of manufacturer websites in addition to reducing duplicative loading of SDS across numerous existing repositories. While SDSs largely remain on chemical vendor websites or various locations, the complexity for systematic improvement remains challenging. Pollutant reduction assessments, including greenhouse gases, could be performed much more efficiently by product comparison and allow for greater visibility of the usage of these products nationwide. A universal SDS system could be created using an existing SDS repository as a model to be expanded upon. For example, the Department of Defense (DoD) Hazardous Material Information Resource System¹⁷ already serves as an effective repository for hundreds of thousands of SDSs while ensuring quality control and includes an XML standard for streamlined digital transfer. A system such as this could be expanded upon to serve as the singular HAZCOM platform for the public.

Figure 2.

Centralized SDS repository.

For an effective so-named Universal SDS system, the system would need to address the many possible ways SDSs can be received and how these varying SDS submittals methods can be consolidated into a single management approach. Ideally, SDSs would be transmitted via a user interface directly to the centralized repository in a standardized XML or equivalent data transfer form. Additionally, for new SDSs creations, a manual creation option would allow vendors to efficiently create new GHS-compliant SDSs with immediate loading into the universal SDS repository. Finally, the last SDS feature would allow other systems using SDSs to communicate the information directly to the centralized universal system via interfaces following the same XML or equivalent data standard. Regardless of the SDS submittal method, standardized validation algorithms could be run to ensure SDSs are complete, GHS compliant, and contain accurate data values. Chemical management responsibility has quickly surged to the forefront of global needs and a universal SDS system would be a large step in our ability to reduce our environmental global footprint while increasing safety and occupational health standards.¹⁸ From climate change to chemical exposure reduction, data quality and chemical analysis precision are paramount in our ability to gather actionable data and use it for effective pollution and exposure reduction.

2.2. Artificial Intelligence/Machine Learning for SDS Processing

Benefits: Significant reduction in manual data entry needs, cost savings, provides transitionary method as industry moves toward the streamlined automated approach.

While a direct vendor-to-SDS repository automated chemical data transfer would be the ideal state, an intermediate solution would be needed to transition from our current methodologies to an automated approach. For the millions of SDSs that still reside in PDF form, one obstacle that many EHS systems have is the need to manually enter chemical information from SDSs into their respective systems for reporting calculations and tracking needs. The 2012 OSHA HAZCOM Standard provided an economic analysis for the cost of compliance, and SDS document management costs were estimated between $75,000 and $100,000 for small global companies and over $100,000 for larger global companies.^19,20 Optical Character Recognition (OCR) allows for these documents to be broken down into machine-encoded text using a variety of prominent OCR tools (e.g., Adobe Pro, Abbey, and Google Tesseract). One difficulty with SDSs is the vast differences in format from manufacturer to manufacturer. Although GHS and REACH regulations have provided some structure in required sections, how and where the data is relayed to readers varies widely. Artificial neural networks (ANNs) allow us to take OCR another step further and use machine learning applications to provide structure to unstructured data sets and essentially learn to “read” an SDS and parse desired fields as required. The benefits of incorporating this technology can yield significant time and monetary savings for organizations that require data entry teams to manually load thousands of these documents each year. Various segments of the SDS are essentially broken down and classified based upon trained models segregating specific SDS values into usable text yielding a level of confidence with each assignment. If the level of confidence is sufficient based upon measured thresholds, the values are recorded. A meta-algorithmic approach is then applied using a machine learning key-value pattern array and tessellation and recombination of the neural network and pattern array results allowing for further validation and reducing the probability of false positives.²¹

Similar to the use of neural networks for SDS text parsing and classification, neural networks can also be used for image recognition of GHS or other regulatory pictograms on an SDS. This can be used to ensure proper labeling and validation of the SDS itself. For instance, an ANN performs recognition on a GHS pictogram and classifies it as a corrosive pictogram. Validation can occur through the text recognition of the pH and determine whether the pH value accurately supports this classification. Analytics and validation queries can be applied to virtually all SDS fields to ensure correct formatting, values are within expected thresholds, GHS requirements have been met, flag banned or chemicals of concern, and to calculate additional necessary (but not always provided) data points.

For workplaces containing personnel with various languages, SDSs can be maintained on job sites in various languages but still need translation for EHS assessment and data entry. Once OCR has been performed on SDSs, an Application Programming Interface (API) can be used for a variety of automatic translation services. Translation can occur through general basic translations or through automated machine-learning translations that can meet domain-specific needs. For large industrial operations, this can yield considerable savings in translation services. The end results of the OCR AI parsed text are machine-encoded files formatted to a given standard for upload into a host SDS repository saving considerable time and manual data entry (Figure 3).

Figure 3.

SDS OCR and AI processing system.

2.3. SDS Data Analytics, Hazard Classification Calculation, and Meta-Algorithmic Validation

Benefits: Automated hazard calculation for storage compatibility, validation of GHS compliance and required fields, and improved SDS data extraction accuracy. Analytics provide data quality control and actionable data for business practice improvement opportunities.

Regardless of input type (i.e., XML, PDF, and manual input), analytical and validation layers can be applied to the SDS processing to ensure data standardization, document completeness, and GHS validation while also calculating hazard characteristic codes (HCCs) and performing meta-algorithmic validation (advanced hybridization of two or more algorithms) between computer vision-derived pictogram classification and HCC classification. The HCC is a code used by the United States DoD to classify materials by their primary hazard for proper segregation and storage of hazardous materials. These codes are calculated based on values extracted or derived from SDSs. The use of HCC assures uniformity in the identification and management of hazardous materials and will assist in the proper recognition and safe storage by compatibility.²² Once calculated using SDS values (whether XML or OCR-based machine learning), these codes can be used for both validation purposes and to enhance the advanced labeling methods described below [e.g., use of radio frequency identification (RFID) for storage proximity warnings for incompatible hazardous materials]. The HCC can be calculated using the OCR-derived fields (e.g., flashpoint, boiling point, pH, etc.) and can then be used to validate the computer vision-derived hazardous classification pictogram on the SDS (e.g., C1-acid, corrosive, and inorganic HCC matches the corrosive GHS pictogram). Once verified, the HCC has additional benefits such as being used in conjunction with RFID tags to ensure storage compatibility.

With a centralized SDS repository, product comparison queries can be run for potential chemical replacement initiatives. For product comparison, naturally, a common denominator must exist by which we can assess similar materials. One such way in the federal logistics system is the national stock number (NSN). Despite varying manufacturers, product names, and chemical characteristics, similarly used products with closely associated size ranges are often assigned the same NSN. This allows for queries of a chemical of concern on one SDS that can be used to trace back to all other product chemical detail associations for various products falling under the same NSN to determine if another product would effectively serve as an eco-friendlier substitute with a more benign chemical formulation. Additionally, resources such as the General Services Administration Green Procurement Compilation provides listings of potential alternatives for sustainable acquisition covered by mandatory and non-mandatory federal environmental programs (e.g., Bio-Preferred, Safer Choice, and Energy Star).²³

Transactional analytics can also be employed for chemical usage monitoring and accuracy assessments. While manual data entry is still largely used in the industry for chemical reporting systems, minor data entry errors can yield significant errors and liabilities in reporting. An incorrect SG or documentation of whether the SG is relative to air or water, for instance, can result in exponential errors in chemical usage. Simple threshold analysis can ensure these values are within expected thresholds likewise for SDS numerical values. The SDS preparation dates can also serve as a useful metric to analyze chemical usage accuracy. Since the GHS adoption by chemical manufacturers in 2015, all manufacturers have been required to abide by GHS standardization requirements ensuring that the oldest preparation dates for products created should be no older than 2015. Many manufacturers also commonly revise or release new SDSs for their products on a nearly annual basis. While some exceptions certainly exist, most deteriorative hazardous products commonly have an associated SDS within a year or two of manufacture. Trend analysis of older SDS associations to recently used materials can often be used to indicate potential EHS data reporting discrepancies.

2.4. “Living” SDSs Using NLP

Benefits: Proactive chemical analysis and alerting measures by providing immediate notification of applicable research, news, and regulatory updates.

NLP is a branch of artificial intelligence that combines computational linguistics with statistical, deep learning, and machine learning models. NLP allows computers to process text and voice recognition using a rule-based model of human language and understand the meaning, intent, and sediment of the language expressed. Regarding SDSs, this allows us to transform SDSs into a “living” document capable of receiving and interpreting feeds on changes to regulatory chemical listings, product recalls, product newsfeeds, and alternative products. Incorporation of NLP into a centralized SDS repository would offer significant enhancement and auxiliary services dramatically expanding on HAZCOM to chemical end users. The consistent monitoring and tracking of chemical updates in suitable research publications, EPA publications, and manufacturer recalls are largely infeasible given the tens if not hundreds of thousands of chemicals large organizations can potentially manage. NLP allows for precision chemical literature curation and can provide recommendation reviews and prioritized classification of these feeds for human review. SDS data points, such as Chemical Abstract Service (CAS) numbers, chemical names, product trade names, and part numbers, would be classified and then run through a named entity recognizer before external document analysis and recommendations would be researched. Given the numerous synonyms for chemicals, an additional semantic dictionary mapper would be needed to replace synonym references to the primary chemical name or CAS. References, in turn, would be run through document analysis and recognition process to identify potential items of interest but also perform filtration on irrelevant terminology and reduce inflectional forms of these key terms.²⁴ Vector space models would then be used for identification of SDS term frequency and inverse document frequency for term importance analysis and pairwise cosine similarity between each document.^25,26 If pairwise frequency thresholds are sufficient, the document can become a candidate for recommendation. Lastly, SDS data stewards would have a final review and relevant feedback could then be propagated back into the system for machine learning and relevance weighting and relevant NLP feedback tied to the chemical of interest for knowledge distribution.

2.5. Label Advancements

Benefits: Precise SDS-to-product association, increased tracking ability, product authenticity, and anti-counterfeit measures.

Improvements to precise SDS selection and information availability to the user can be made by moving from traditional legacy product barcodes to the following modern applications than can be used to quickly direct users to the product’s SDS.

2.5.1. Basic SDS Quick-Response (QR) Codes

In the advent of the COVID-19 global pandemic, the use of QR codes has seen a dramatic rise with many restaurants replacing hard copy menus with digitally accessible versions online using common cell phone camera applications. QR codes are available in two forms—static and dynamic, dependent on whether a user wants data essentially hardcoded into a barcode or using a URL that can be modified after printing.²⁷ Numerous QR code generators exist for chemical vendors or EHS personnel to create these codes at little to no additional cost. These codes could be printed on custom organizational barcodes or, ideally, by the vendors on the products themselves.

The ability to immediately retrieve the correct SDS for any hazardous material by any personnel with a cell phone helps alleviate the reliance on often unmaintained SDS binders and greatly expedites the time for personnel to access these SDSs in the event of an emergency.²⁸ Additionally, direct association of the SDS has the added benefit of remaining tied to the product outside of the workplace, should the product move from its original intended use location. Lastly, association of a product to the chemical SDS is mandatory for many environmental, health, and safety compliance systems that use the ingredients and hazards on the SDS to calculate usage for regulatory reporting, determine exposure for personnel, and employ appropriate safety measures. Direct QR code association from a vendor would greatly increase the speed and efficiency in loading SDS information into downstream user systems by eliminating the need for manual research on vendor sites.

QR code storage capacity depends on the size and version used. Modern QR codes (currently version 40) contain 31,329 squares encoding up to 3 KB of data translating to 7089 numeric characters or 4269 alphanumerics (additionally Kanji/Kana, Arabic, and other languages can be stored with varying capacities).²⁹ The average website URL contains approximately 40–50 alphanumeric characters so the potential clearly exists for schema development that could provide numerous relevant hazardous material data points to a user.

2.5.2. QR Code Application Data Expansion

In addition to retrieving SDSs from a cloud-based server, the additional benefits of the expansion of product-related data that can be retrieved can greatly enhance industrial operations. One example is the direct transmission of an SDS to a downstream user system. If an XML or equivalent file is hosted on the server, the machine-encoded file could be transmitted directly to subsequent systems (assuming an XML standard is met). If the file is in PDF, a combination of OCR and AI meta-algorithmics (described above) can be used to reverse engineer and parse and validate SDS fields into a usable format by the downstream system. This eliminates not only the need for the chemical user to scour the internet looking for the correct SDS but also immediately submits it to a database in the machine-encoded text, which would greatly expedite if not eliminate the need for SDS manual data entry for hazardous material tracking needs. The SDS is automatically routed for environmental, safety, and occupational health review (which could also be performed via mobile application).

Second, the QR code can be used to retrieve other product-related data to enhance inventory operations. Information such as unique SDS identifiers, product batch/lot information, container numbers, NSNs, container/unit/package information (e.g., 16 oz bottle), manufacturer, trade name, noun, manufacture date, and expiration date could all be retrieved by the same API connection. For the DoD, expiration dates could include not only the original product expiration but also return updated expiration and service life dates based upon lab result testing or user extension, whereas traditional barcodes are simply printed with the original dates and require database searches for these updates (Figure 4).

Figure 4.

Examples of various expanded QR codes capable of product authentication and SDS data retrieval via object library association.

Object storage services are used to provide supplemental material such as images (SDSs, technical data sheets, and specifications) and any other associated documents to the downstream user. QR codes have also advanced for uses in supply control measures and counterfeit detection.^30,31 Several anti-counterfeit designs have been created including QR codes containing secure graphics that can verify the authenticity of the product and provide detailed information on the location and devices used to scan the product. While this technology was originally developed for security needs, it can also be used to meet a host of EHS needs and requirements and can be used to forward SDSs to an AI processing system, which would read and load the SDS into the host repository while maintaining supply chain authentication (Figure 5).

Figure 5.

QR code SDS retrieval and AI processing methodology.

The proposed methodology can not only provide numerous benefits to chemical users but to the manufacturers as well. Negative implications are primarily cost-driven, so the cost–benefit analysis would depend on the potential workload savings and the monetary losses associated with counterfeit products. The many benefits to chemical vendors include the following:

• reduction of back-and-forth correspondence to chemical users on SDS locations and SDS data points

• marketing of new added benefits of streamlined SDSs and product information to users

• benefitting from brand protection while greatly reducing counterfeiting and offering validation of product authenticity

• increased customer engagement in their products

• significant reduction in counterfeit products; greater product security

2.6. Container Tracking Optimization

Benefits: Automated tracking ability, cost/time savings from manual tracking reduction, compatibility validation, and warning opportunities for inventory concerns.

2.6.1. RFID Container Tracking

Another possible labeling improvement that could either work in conjunction with QR code systems or as an alternative to traditional barcodes is the use of RFID tags on hazardous materials and hazardous waste, either by the vendor or by downstream user labeling upon receipt. Hazardous material and hazardous waste management typically require a database for large operations to accurately and consistently monitor where containers are and closely observe the amount of time containers exist at these sites and approach regulatory thresholds. Movement of containers from site to site within the database thus requires manual transactions that must occur for transfer documentation. The use of RFID technology can be used to eliminate the need to track hazardous material containers from supply points to end users or waste containers as they move from satellite accumulation points to centrally managed hazardous waste storage areas or treatment, storage, and disposal facilities. Rather than relying on follow-up manual transfers to occur, sensors at each location would automatically document the movement and record the appropriate transaction in the host database. RFID tags also eliminate the need for a line of sight between tags and readers allowing expedited inventory management and accountability.³⁰ Many inventories in the DoD are still performed either by manual count or scanning using legacy barcode scanners. RFID tags enable constant accurate accounting of inventory. Similarly, emergency response personnel can be equipped with technologies that not only provide what chemicals are stored in a response location based on receipt transactions but also what products and chemicals RFID tags are communicating that exist within the RFID boundaries (Figure 6).

Figure 6.

RFID hazardous material/waste tracking.

Microchips in RFID tags can be either read-only or read-write. The latter allows users to add data to the RFID tag, such as the waste profile number for the waste, the accumulation start date, and the site-specific regulatory start date. Likewise, for hazardous materials, SDS IDs, expiration dates, and other inventory data points can be loaded against the RFID tag. The benefits of incorporating this technology could yield significant time savings of personnel, monetary savings of reduced data entry and tracking, and increased EHS accountability of these containers. Additionally, dangers associated with improper hazard segregation could be mitigated by cross-analysis of incompatible storage items through RFID proximity. For instance, acids should not be stored with bases or flammable products should not be stored with oxidizers. RFID technologies could be used to trigger a warning system of possible incompatible storage. Similar warnings can be presented to inventory managers on a host of other potential inventory concerns such as expiration, shortages, max allotment exceedances, and temperature threshold concerns. Active RFID tags can store between 16 bytes and 128 KB, and passive UHF tags are capable of 32 KB at a frequency between 865 and 956 MHz allowing for longer ranges,³² both offering ample capacity for the limited EHS inventory fields desired.

3. Results and Discussion

While a national centralized SDS repository would likely require regulatory agency approval, some of the other proposed advancements have been implemented or are in the process of being implemented in various industrial sectors. NASA’s Dryden Flight Research Center developed a working solution using RFID to provide real-time information on usage, shipment, tracking, and storing of chemicals.³³ The system included abilities to notify personnel using a light system if containers were incorrectly removed or added, and to verify that users were properly authorized and trained on the chemicals they were using. NASA plans for a secondary phase to automatically check vehicles entering and leaving access points for RFID-tagged items and track climate-controlled chemicals.

QR code technology also continues to advance, allowing authentication and security protocols to be embedded in the code itself. Scantrust, a Switzerland-based firm, has developed such technology which has improved product validation, reduced counterfeiting, and improved supply chain management and customer engagement.³⁴ Scantrust has developed technology that expands on standard QR codes by adding secure serialization (unit-level detail with the delivery of targeted messages), authentication, verification, and alerts and messaging in a single QR format. QR codes such as these can be used not only to retrieve product SDSs but also to validate product authenticity to reduce counterfeit products and increase tracking abilities down to the specific users, locations, and models of phones used to scan products. Additionally, QR codes could work in conjunction with the proposed RFID solutions.

We (the authors) have also conducted studies on meta-algorithmic applications using AI for SDS processing. As part of ongoing research for the DoD, a prototype is currently in development, which has the potential to reduce over $3 million in manual data entry costs while greatly expediting the time for SDSs to get loaded.

Using the previously mentioned meta-algorithmic approach, we have improved upon the 77 and 71% accuracy rates for convolutional neural networks and key-value pattern arrays, respectively. The meta-algorithmic tessellation and recombination of the other algorithms improved SDS data point extraction to a 92% accuracy rate (Figure 7).²¹ Building upon this approach, Colorado State University was awarded research funding for proof-of-concept prototype development of a web-based application that can process thousands of records and retrieve greater than 90% accuracy of the desired fields.

Figure 7.

AI SDS meta-algorithmic processing results.

The objective will be for the AI prototype to develop into a fully operational system and serve as an intermediary tool as the DoD pursues efforts to require XML SDSs from its vendors for a more streamlined data transfer. The end system will be designed to accommodate PDF data parsing through the discussed approach as well as accept XML in the DoD SDS XML, through the manual creation of an SDS (for institutions without the resources to accommodate XML), and interface with systems procuring SDSs. Regardless of the submittal method, all SDSs will be validated with a standardized set of validation algorithms and will perform other SDS-specific calculations.

4. Conclusions

The discussed technologies all represent possibilities in employing 21st technologies to increase safety and minimize human and environmental exposures. To minimize greenhouse gas emissions and our chemical footprints, we first need an accurate account of the products we use and translate this into actionable data. Additionally, upgrades in EHS are needed for a more proactive approach to hazard identification and minimization. These process additions would effectively allow us instant HAZCOM in emergencies and more efficient, accurate, and robust HAZCOM allowing us access to the latest changes in regulatory tracking and chemical research. Proactivity would also extend to safety measures informing personnel of chemical temperature or incompatibility warnings before reaction and adding efficiency gains for chemical lifecycle tracking. While each of the discussed methods can be employed either as a combined system or independently in existing industrial management processes, all of which would benefit from a centralized and universal SDS repository providing an easily accessible and standardized data source while minimizing data quality issues garnered from data transfers. Applications like machine learning and the proposed meta-algorithmic approach provide a quicker and more efficient means by which to transition to fully automated HAZCOM. The number of hazardous communication failure findings remains staggering and associated hazard data quality loss and availability is insufficient and unacceptable for emergencies, both immediate and the longer-term climate change emergencies. We have the means to currently implement these suggestions and move the industry forward; it only requires action.

The authors declare no competing financial interest.