A method and system for screening greener and safer alternatives to chemical substances based on application characteristics

Abstract

This invention belongs to the field of chemical substance and environmental management technology, and discloses a method and system for use-oriented auxiliary screening of greener and safer alternatives to chemical substances. The method, geared towards the industrial manufacturing sector, incorporates core parameters regarding the impact of chemical substances on human health, ecological safety, and their environmental migration and transformation capabilities. It employs use-oriented, data-driven, and tiered screening techniques, including the following steps: screening chemical substances to be replaced and confirming their uses; extracting chemical substances with similar uses to establish a candidate library; incorporating multiple indicators across three dimensions, processing hazard data, and grading the hazard endpoints based on their level of concern; and using the analytic hierarchy process (AHP) to assist in screening greener and safer alternatives. This invention also provides a screening system for implementing this method. The alternative evaluation is highly scientific, consistent, and operable, and can meet the requirements of matching the function of the alternative with the use of the replaced substance and ensuring green safety from multiple aspects and dimensions.

Description

Technical Field

[0001] This invention belongs to the field of chemical substance and environmental management technology, specifically relating to a use-oriented auxiliary screening method and system for greener and safer alternatives to chemical substances. It addresses the problems of lack of green alternative screening methods, difficulty in integrating multi-source data, and lack of use-adaptation in traditional evaluation systems. It constructs a "use-oriented, data-driven, and graded screening" technical method and information system, which is particularly suitable for solving problems such as secondary research and development caused by improper alternative screening in the industrial sector. Background Technology

[0002] With the rapid development of industrial manufacturing in recent years, the types, quantities, and uses of chemical substances have increased dramatically. Toxic and hazardous chemicals used in industrial production are widely present in the air, water, soil, and food chain, entering the human body and the bodies of animals and plants through various pathways, posing a significant threat to human health and the ecological environment. In the 1970s, to comply with environmental protection regulations, US chemical companies reduced pollutant emissions through end-of-pipe emission reduction measures such as installing pollution control and waste treatment devices. However, with the increasingly stringent regulations of the US Environmental Protection Agency (USEPA)'s Clean Water Act and Clean Air Act, end-of-pipe emission reduction alone was no longer sufficient to meet regulatory requirements. Companies had to find less harmful chemical substances to replace toxic and hazardous chemicals in their products or processes. To reduce the harm of toxic and hazardous chemicals to human health and the ecological environment, the concept of "substitution" has been introduced worldwide, leading to various policies, control measures, and industry management initiatives for sustainable chemicals. The EU's REACH (Registration, Evaluation, Authorization, and Restriction of Chemicals) regulation, USEPA, and the Occupational Safety and Health Administration all include the substitution of toxic and hazardous chemicals as a core component of their chemical substance management policies.

[0003] The core of substitution efforts for toxic and hazardous chemicals is to select a safer alternative that does not affect the performance of the final product and / or process. However, "regrettable substitution" incidents frequently occur during the process of enterprises undertaking substitution, meaning that the alternative is found to be equally toxic or even more toxic. International alternative assessment methods originated in the early 20th century, and with the continuous improvement of understanding of the hazards of chemicals, the various indicators focused on in substitution assessments have also been continuously developed and improved. To reduce unfortunate substitutions and achieve informed replacements with safer chemicals, many governments and non-governmental organizations have successively formulated more than 20 substitution assessment frameworks or guidelines and developed numerous substitution assessment tools, such as the hazardous chemicals management methods and systems disclosed in CN118586862A, the deep learning method for designing the molecular structures of green alternatives to chemicals disclosed in CN120108556A, and the latest developments in the OECD Guidelines for Testing Alternative Methods for Local Toxicity of Chemicals (DOI: 10.20099 / j.issn.1000-4475.2022.0132). However, for chemical substitution in industrial manufacturing, there are still many technical problems in terms of functional adaptation, safety assessment, data integration, and system implementation, including: I. Difficulty in accurately matching multi-dimensional uses and functional adaptability 1. Insufficient granularity and dynamic adaptation in application classification. While CN118586862A categorizes chemicals into broad classes such as "laboratory reagents, process system chemicals, and auxiliary materials," it fails to further refine the "use-function" relationship (e.g., the functional differences of the same "process system chemical" in specific uses like "metal cleaning" and "coating film formation"). In actual industrial scenarios, the uses of chemicals often change dynamically with production processes and product formulations (e.g., the functional effectiveness of a solvent at different temperatures and pressures), but existing classification systems are mostly static and cannot respond to subtle adjustments in use in real time. CN120108556A focuses on molecular structure design and uses SMILES codes to predict functional properties (e.g., surface tension). TS Critical micelle concentration logCMC However, it only sets functional scenarios for "surfactants" and does not cover multi-purpose categories such as "flame retardants" and "plasticizers". Furthermore, the function prediction relies on fixed molecular descriptors and cannot be associated with "application scenario parameters" (such as the melting temperature and mechanical property requirements that plasticizers need to be matched in plastic processing), which leads to the mismatch problem of "functionality meets the requirements but application is not applicable" (such as a substitute having the required surface tension, but decomposing and failing in high-temperature processes).

[0004] 2. Lack of functional transferability assessment for cross-industry applications The local toxicity testing methods in the OECD guidelines (such as OECD 497 skin sensitization and OECD 498 phototoxicity) only target general safety endpoints such as "skin contact" and "eye irritation," without considering the specific functional requirements of different industry applications (such as "migration resistance" for food contact materials and "insulation" for electronic chemicals). Existing screening methods lack a linked assessment framework of "use-function-safety." For example, an alternative may meet the toxicity test standards, but it cannot be practically applied because it cannot meet the "low volatility" requirements of the electronics industry. II. Bottlenecks in the Integration and Standardization of Multi-Source Security Data Green and safe alternatives need to cover multiple safety indicators such as "health hazards, ecotoxicity, and environmental fate," but existing technologies face insurmountable obstacles in data integration and standardization. 2.1 Lack of prioritization and conflict resolution mechanisms for heterogeneous data The OECD guidelines mention that OECD methods should integrate "in vitro testing (such as DPRA), computer prediction (such as DEREKNEXUS), and human trial data," but they do not clearly define the priority rules for data from different sources. For example, when "in vitro skin sensitization testing (OECD 442C) shows a negative result, but the QSAR model predicts a positive result," there is a lack of a scientific method for weighting. In actual screening, health hazard data often come from different laboratories (e.g., cytotoxicity test results from different institutions can differ by more than 30%), and ecotoxicity data come from different species (fish, algae, etc.). EC 50 (Value contradiction) Existing systems (such as MSDS management in CN118586862A) can only store data and cannot resolve conflicts. The deep learning model in CN120108556A relies on a single dataset such as the "PFAS molecular structure list" and does not integrate local toxicity data (such as phototoxicity and eye irritation) from the OECD guidelines, resulting in incomplete safety assessment dimensions (e.g., only predicting...). logKow (Considering physical and chemical toxicity, but ignoring local hazards such as skin sensitization), there is a risk of "safety underestimation". 2.2 Difficulties in real-time access and adaptation of dynamic regulatory data CN118586862A is associated with a material code through its MSDS, but the safety information in the MSDS is outdated (e.g., if a substance is added as a REACH VHC, the MSDS does not promptly indicate this); furthermore, it is not aligned with international regulatory updates (e.g., the risk assessment system for CN118586862A was not simultaneously incorporated into the OECD 494 eye irritation test method updated in 2021). The existing system lacks an automatic linkage mechanism between the "regulatory database and screening model," which may result in screening results that do not meet the latest regulatory requirements (e.g., a substitute material is classified as "green" under the old standard, but should be classified as "red risk" under the new standard). III. Difficulties in Coordinating "Function-Safety-Cost" in Multi-Objective Optimization Green and safe alternatives need to meet the requirements of "functional compliance, low toxicity and controllable cost" at the same time, but existing technologies have significant shortcomings in multi-objective collaborative optimization. IV. Engineering Obstacles to the Implementation of Intelligent Screening Systems Application-oriented screening needs to be implemented through a complete system of "data input - model calculation - result output - decision support", but existing technologies face many challenges in engineering implementation.

[0005] In summary, existing technical methods have not yet considered green parameters such as human health, ecological safety, and environmental migration and transformation of alternatives. They cannot meet the core requirements of screening greener and safer alternatives to chemical substances that are combined with application-oriented approaches (functional adaptation, safety assessment, data integration, and system implementation). Moreover, most technical methods are only at the framework stage and do not provide operable solutions for data collection, processing, and evaluation. The scientific rigor, uniformity, and operability of alternative evaluation still need to be strengthened in order to achieve a complete match between the function of the alternative and the use of the replaced substance from multiple aspects. Summary of the Invention

[0006] This invention addresses the aforementioned challenges, including the difficulty in accurately matching multi-dimensional application-functionality, the bottleneck in integrating and standardizing multi-source safety data, the difficulty in synergizing "function-safety-cost" in multi-objective optimization, and the obstacles to the engineering implementation of intelligent screening systems. It provides an application-oriented method and system for screening safer alternatives to chemical substances. Targeting the industrial manufacturing sector, it constructs a "application-oriented, data-driven, and tiered screening" technical approach. Based on three layers of collaborative data—"application-substance-evaluation results"—it forms an application-oriented method for screening safer alternatives and integrates it into an automated screening system to assist in screening greener and safer alternatives. This invention solves the aforementioned technical problems from multiple dimensions, including technical adaptability, data integration, evaluation system, and level of intelligence.

[0007] The technical solution adopted by this invention to solve its technical problem is: A use-oriented method for screening greener and safer alternatives to chemical substances is proposed. This method, geared towards the industrial manufacturing sector, incorporates core parameters related to human health, ecological safety, and environmental migration and transformation. It employs a use-oriented, data-driven, and tiered screening approach, primarily including: screening chemical substances to be replaced based on environmental management lists and confirming their uses; extracting chemical substances with similar uses to establish a candidate library; incorporating multiple indicators across three dimensions—health hazards, ecotoxicity, and environmental fate—and utilizing a four-layer optimization method based on heterogeneous data, combined with algorithms based on geometric mean, majority vote, and the most sensitive principle to process hazard data, and grading the hazard endpoints based on their level of concern; and constructing a four-level tiered classification using the analytic hierarchy process (AHP), and after calculation and evaluation, assisting in the screening of greener and safer alternatives.

[0008] A use-oriented chemical substance greener and safer alternative auxiliary screening system for implementing the aforementioned greener and safer alternative auxiliary screening method includes a network server, management terminal, user access terminal and external database connected and communicating via the Internet; The network server has a built-in program to assist in screening greener and safer alternatives by executing steps S1-S4. The program is based on a three-layer collaborative data system of "use-substance-evaluation results" to assist in screening greener and safer alternatives, and includes: an input unit, a search and verification unit, an alternative evaluation unit, and an output unit. The external databases include: a database of substituted chemical substances and their uses, and a database of laws and regulations; The database of replaced chemical substances and their uses includes: Structured database, functional application database FUse, chemical substance data for the industrial chain; domestic chemical substance control data, chemical substance data for international conventions; The legal and regulatory database includes: official directories and industry standard data, and an international regulatory database. The management terminal is used to manage the auxiliary screening program for greener and safer alternatives; The user access terminal is used to provide users with an interactive access interface. Users input the CAS number or purpose of the chemical substance being replaced, and the system automatically outputs auxiliary search results for greener and safer alternatives. Beneficial effects

[0009] The beneficial effects of the present invention include at least the following: 1. This invention addresses several bottlenecks in existing technologies for chemical substance substitution in industrial manufacturing, proposing a novel technical approach and systematic tools. Through the synergy of multiple algorithmic models and data optimization rules, it constructs a "use-oriented, data-driven, and hierarchical screening" technical method. Based on three layers of collaborative data—"use-substance-evaluation results"—it forms a use-oriented method for screening safer alternatives and integrates it into an automated screening system to assist in screening greener and safer alternatives. This invention solves problems such as the difficulty of accurately matching multi-dimensional use-functional adaptability, the bottleneck of integrating and standardizing multi-source safety data, the difficulty of synergizing "function-safety-cost" in multi-objective optimization, and the engineering implementation obstacles of intelligent screening systems. It provides an efficient and convenient use-oriented method and system for screening safer alternatives to chemical substances, which can be widely used in the industrial manufacturing sector.

[0010] 2. This invention incorporates green parameters such as human health, ecological safety, and environmental migration and transformation of alternatives, which can meet the core needs of screening greener and safer alternatives to chemical substances based on application orientation (functional adaptation, safety assessment, data integration, and system implementation). Furthermore, this technical method is implementable and provides an operable solution for data collection, processing, and evaluation methods, thereby simultaneously enhancing the scientific rigor, uniformity, and operability of alternative evaluation. It can achieve the requirement of complete matching between the function of the alternative and the use of the replaced substance from multiple aspects.

[0011] 3. This invention addresses the challenges of multi-source data integration and standardization by adopting a priority sequence of data sources ("GHS classification > REACH registration > experimental data > predicted data") from multiple dimensions, including technological adaptability, data integration, evaluation system, and level of intelligence. It achieves the scientific transformation of qualitative and quantitative data through algorithms such as geometric mean, majority vote, and the most sensitive principle. Furthermore, it utilizes a hierarchical analysis method (AHP) to establish three-dimensional indicators (health hazards, ecotoxicity, and environmental fate) and a four-level classification assessment, conducting evaluation and analysis at the molecular level and outputting four-level assessment results: Level I (green), Level II (yellow), Level III (orange), and Level IV (red). This fills the gaps in existing alternative evaluation methods and provides a replicable methodological framework for finding safer alternatives to chemical substances globally.

[0012] 4. This invention addresses the difficulty of accurately matching multi-dimensional uses and functions, improves the granularity and dynamic adaptability of use classification, and incorporates cross-industry use function transferability assessment, increasing the success rate of industry-specific parameter adaptation from 40% to over 70%. The function matching efficiency is more than 10 times higher than manual screening (approximately 5 minutes / compound), reducing the probability of mistakenly selecting substitutes containing toxic groups to below 1%.

[0013] 5. This invention strengthens the synergy of "function-safety-cost" in multi-objective optimization. Based on a three-layer synergy of "use-substance-evaluation results" at the molecular level, it constructs a safer and greener alternative screening system for multiple industrial sectors. This system breaks through the traditional alternative screening model, integrating two core technologies to form system-level innovation: ① Developing a dual-mode input system that supports CAS number retrieval of known substances and use scenario retrieval; ② Innovating a multi-dimensional output architecture that outputs hazard endpoint concern level classification results and evaluation results, comprehensively assisting in the screening of safer alternatives to chemical substances, achieving the following technical effects: Application-oriented: Substitutes must meet the functional requirements of the original substance (such as desulfurization, flame retardancy, and exhaust gas treatment) to avoid substitutions with missing functions; Data-driven: All methods combine multi-source data (such as chemical structure, biological activity, and environmental behavior) to screen for alternatives (such as QSAR models, Tox21 data, and QSAR predictions). Prioritization: All alternatives are prioritized based on hazard assessment (such as toxicity classification) and exposure assessment (such as usage scenarios) (e.g., "environmentally friendly priority" for fluorine-free desulfurizers and "safety priority" for flame-retardant PA6). Technological assistance: Using computational tools (including machine learning and cross-reference models) to improve screening efficiency aligns with the development direction of "intelligent screening systems".

[0014] 6. The safer and greener alternative screening system provided by this invention is based on a B / S architecture, integrates multiple evaluation algorithms and intelligent screening modules, supports dual-mode input of CAS retrieval and application screening, and can realize automated and intelligent assisted screening of safer alternatives to chemical substances with application orientation. Furthermore, its use of machine learning to automatically learn the complex relationships between indicators from molecular data can effectively solve the problem of traditional analytic hierarchy process (AHP) relying on expert experience for weights, greatly improve the level of intelligence and the objectivity of screening results, and solve the obstacles to the engineering implementation of intelligent screening systems. Attached Figure Description

[0015] Figure 1 A schematic diagram of the composition and structure of the auxiliary screening system for greener and safer alternatives to chemical substances for the purpose-oriented application of this invention. Figure 2 A schematic flowchart illustrating the auxiliary screening method for application-oriented greener and safer alternatives to chemical substances in embodiments of the present invention; Figure 3 A schematic diagram of the retrieval process for an application-oriented, greener, and safer alternative chemical substance screening system for embodiments of the present invention; Figure 4This is a schematic diagram of the endpoint query results for the carcinogenic effect of bisphenol A in REACH according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the ozone-depleting effect endpoint query results of bisphenol A in REACH according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the hazard endpoint concern classification results of bisphenol A in an embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the evaluation results of some solvent-based substances in the embodiments of the present invention; Figure 8 This is a schematic diagram of the CAS query interface of the chemical substance safety substitution assessment system for the coating industry in an embodiment of the present invention; Figure 9 This is a schematic diagram of the CAS query interface of the optimal green and safe alternative assessment system for chemical substances in the electronics industry in an embodiment of the present invention; Figure 10 This is a schematic diagram of the CAS query interface of the optimal green and safe alternative assessment system for chemical substances in the textile printing and dyeing industry in an embodiment of the present invention. Detailed Implementation

[0016] The following is in conjunction with the appendix Figure 1-10 The technical solution of the present invention will be described in detail through several embodiments.

[0017] Basic Implementation See Figures 1 to 3 This embodiment provides a method for screening greener and safer alternatives to chemical substances based on their intended use. This method is geared towards the industrial manufacturing sector and incorporates core parameters regarding the impact of chemical substances on human health, ecological safety, and their environmental migration and transformation capabilities. It employs a use-oriented, data-driven, and tiered screening approach, primarily including: screening chemical substances to be replaced based on an environmental management inventory and confirming their uses; extracting chemical substances with similar uses to establish a candidate library; incorporating multiple indicators across three dimensions—health hazards, ecotoxicity, and environmental fate—and using a four-layer optimization method based on heterogeneous data, combined with algorithms based on geometric mean, majority vote, and the most sensitive principle to process hazard data, and tiering the hazard endpoints based on their level of concern; constructing a four-level tiered classification using the analytic hierarchy process (AHP); and calculating and evaluating the results to assist in screening greener and safer alternatives. The method includes the following steps: S1: Screening for alternative chemical substances and their uses Based on the existing environmental management inventory of chemical substances, chemical substances that need to be replaced are screened, and the uses of the replaced substances are confirmed; all the chemical substances mentioned are organic or inorganic substances with CAS numbers, specifically including: S1-1: Screening for alternative chemical substances By comparing existing environmental management lists of chemical substances with toxicity information for initial screening, and then prioritizing the replacement of highly toxic substances with a strong need for substitution, the chemical substances to be replaced are selected, including the following steps: S1-1-1 First, collect a list of chemical substances involved in the entire industrial chain. All chemical substances must be organic or inorganic substances with CAS numbers. S1-1-2. Comparison with domestic chemical substance control lists: By comparing with lists such as the "List of Key Controlled New Pollutants (2023 Edition)," "List of Priority Controlled Chemicals (First Batch)," and "List of Priority Controlled Chemicals (Second Batch)," chemical substances in the lists are considered to be of the highest priority and need to be replaced. S1-1-3. Comparison with international convention lists of chemical substances: By comparing with conventions such as the Stockholm Convention on Persistent Organic Pollutants and the Rotterdam Convention on Prior Informed Consent Procedures for Certain Hazardous Chemicals and Pesticides in International Trade, chemical substances within the scope of the conventions are considered as secondary priority substances that need to be replaced. S1-1-4. Comparison with foreign chemical substance control lists: By comparing with the EU's list of chemical substances of very high concern (SVHC), the EU's list of restricted authorized chemical substances under the REACH regulation, and the US TSCA list of PBT chemical substances, chemical substances on the list are considered to be the third priority that need to be replaced. S1-1-5. Screening for highly toxic chemicals: Based on existing data, screen for chemicals with PBT or vPvB properties, or those with carcinogenic, mutagenic, or reproductive toxicity (CMR) properties. These chemicals are considered the last priority to be replaced. S1-2: Confirm the use of the chemical substance being replaced S1-2-1: Referencing official directories and industry standards By combining industry application scenarios and accessing official directories, industry standards, and international regulatory databases, we can obtain direct reference uses for the substances being replaced, including the "Recommended List of Substances for Ozone Depletion in China," REACH regulations and the ECHACHEM platform, EPA databases, etc., to determine the main uses of the substances being replaced. S1-2-2: Calling a structured database (such as ECHA CHEM, CPDat) Information from multiple databases (such as ECHA CHEM and CPDat) can be accessed to obtain conflicting data on the various uses of the substituted substance. For example, if a substance is listed as a solvent in the REACH registration, but the FUse database predicts that it may be used as an antioxidant, further verification through literature or experiments is required. S1-2-3: Call the functional database (FUSE) to resolve data conflicts. The Quantitative Structure-Function Relationship (QSUR) model built on the FUse database resolves conflicting information on multiple uses of substituted substances. The FUse database integrates functional use data of more than 14,000 chemicals, uses a unified classification (such as surfactants and fragrances) and a random forest classification algorithm to predict whether chemicals have a specified function (such as flame retardants), and combines high-throughput screening data to assess their safety. S1-2-4: Addressing Data Uncertainty Finally, by combining experimental data with supply chain data, data uncertainties are addressed, ensuring the accuracy and comprehensiveness of application confirmation. S2: Establish a list of alternative alternatives Extract substances from existing chemical substances that have the same uses as the chemical substance to be replaced, and extract their functional molecular structures and molecular groups, removing substances already included in the environmental management list of existing chemical substances and their main molecular groups; the alternative substitutes are organic or inorganic substances with CAS numbers, specifically including: S2-1: Establish a list of alternative substitutes Based on the molecular structure, main molecular groups and physicochemical properties of chemicals, and with a focus on their intended use, we search and compile chemical substances with the same uses as the chemical substances being replaced in existing databases, extract their functional molecular structures and molecular groups, remove substances and their main molecular groups that have been included in the existing environmental management list of chemical substances, and establish a list of alternative substances. Alternatively, machine learning models can be used to make predictions, predict the intended use, analyze molecular descriptors using deep learning neural networks, and combine this with lifecycle impact data from the Ecoinvent database to screen for alternatives with the same functionality. S2-2: Priority sorting For each alternative, its functional molecular structure and molecular groups are extracted, and the safety of its functional molecular structure and molecular groups is evaluated. The alternatives are sorted according to the priority sequence of "GHS classification > REACH registration > experimental data > predicted data" to establish a list of alternatives with data source adoption priority. Functional molecular structures and functional groups were extracted from alternative molecules. A hierarchical analysis (AHP) was first employed to construct a four-level classification method (Level I green to Level IV red). Core parameters related to human health, ecological safety, and environmental migration and transformation were used as the criteria layer factors for the AHP. Based on the relative importance of each factor, a comprehensive score was calculated for each alternative to assess its suitability and determine its classification. After assessment, safety was ranked, and then data sources were prioritized according to the data source order: "GHS classification > REACH registration > experimental data > predicted data." Finally, a priority list of alternative candidates was established. The weight adjustment of the AHP was implemented using a Long Short-Term Memory (LSTM) model. The input data for the LSTM model consisted of environmental policy change texts from historical alternative cases and an expert weight allocation matrix. The output was dynamic weight parameters for the current assessment task. The criteria layer weights of the AHP were dynamically adjusted using the LSTM model, and the training data included at least 5000 historical alternative assessment records and expert weight allocation results. To enhance the objectivity of data fusion, this embodiment further employs the deep learning-enabled hierarchical analysis method. By constructing an LSTM network to analyze the expert weight allocation pattern in historical alternative cases, the weights of the AHP criterion layer are dynamically optimized (e.g., the ecotoxicity weight is adjusted from 0.3 to 0.4). The Shapley value algorithm is introduced to explain the model decision and generate a visual report (e.g., "contribution of bioaccumulation to the current Class IV classification").

[0018] The safety assessment steps for the molecular structures of alternative alternatives are as follows: First, the SMILES codes of the alternative alternatives are parsed to extract a set of molecular descriptors, which include LogP, TPSA, and the number of hydrogen bond donors / acceptors; the molecular descriptors are input into a pre-trained QSAR prediction model, which outputs predicted values ​​for carcinogenicity, mutagenicity, and aquatic toxicity; and the chemical bond breaking energy is obtained by combining quantum chemical calculations. ΔE ,like ΔE Substances with a binding energy of <250 kJ / mol are labeled as readily degradable. The binding energy with biological enzymes is obtained through molecular docking. If the absolute value of the binding energy is >7 kcal / mol, the bioaccumulation level is upgraded. In the assessment of fate in molecular dynamics simulation environments, the following are included: (1) Degradation pathway simulation: The degradation half-life of molecules in water was simulated using GROMACS software, and the breaking energy of key chemical bonds was evaluated by quantum chemical calculations (DFT). ΔE ), classifying persistence levels ( ΔE < 200 kJ / mol → Easily degradable.

[0019] (2) Biomembrane permeability analysis: The binding energy of compounds to cytochrome P450 enzymes was predicted by molecular docking (AutoDock), and the metabolic rate (Kcat) was assessed to adjust the bioaccumulation classification; (3) Analysis of molecular group activity and functional adaptability Functional group library construction: Establish the mapping relationship between flame retardant groups (-PO(OCH3)2), plasticizing groups (-COOCH2CH3), etc. and their uses, and identify the matching degree of core functional groups of alternatives through convolutional neural networks (CNN).

[0020] Side effect group screening: embeds a blacklist of known toxic groups (such as nitrobenzene ring, polybrominated diphenyl ether), and automatically marks high-risk compounds (directly classified as level IV) that match the candidate library.

[0021] Hazard prediction models based on molecular descriptors and QSAR models include: Key molecular descriptor extraction: Define molecular parameters directly related to health hazards and ecotoxicity, including: Topological polar surface area (TPSA) → predicts transmembrane permeability; Octyl alcohol-water partition coefficient (LogP) → bioaccumulation; Number of hydrogen bond donors / acceptors → Reactivity is correlated with toxicity; Molecular weight (MW) and volume → environmental mobility; Dynamic QSAR model construction: Based on Random Forest or GNN models, the nonlinear relationship between molecular structure and toxicity endpoints in training datasets (such as ToxCast, ECOTOX) is used to output predicted values ​​for carcinogenicity, mutagenicity, etc. (using model formulas: Toxicity = f ( LogP , TPSA , H - bond )).

[0022] For example, for azelaic acid esters (CAS 103-24-2), the CO bond breaking energy calculated by DFT is 180 kJ / mol, which is determined to be easily degradable (low persistence); its binding energy with CYP3A4 is -8.2 kcal / mol, and the predicted metabolic half-life is <12h, which downgrades the bioaccumulation level by one level.

[0023] To address the issue of migration prediction errors exceeding 30% caused by the lack of integration of dynamic environmental parameters in traditional QSAR models, this embodiment combines molecular descriptors with QSAR models, which can reduce the chronic toxicity prediction error from ±30% to ±12%. This embodiment uses quantum chemical bond energy thresholds to classify degradation levels, fully considering the quantitative influence of molecular structure on degradation pathways, and differentiates the differences in the breaking energies of CO and CC bonds. This overcomes the shortcomings of using fixed weights to assess the environmental risk of alternatives, which leads to a persistence assessment error of more than 40%. By calculating chemical bond energy thresholds through quantum chemistry, it achieves accurate classification of degradation capabilities, and the assessment results are consistent with OECD testing standards by more than 90%. S2-3: Safety-related exclusions Chemicals already included in the existing environmental management list of chemical substances are removed, and the remaining substances are identified as alternatives in order of priority.

[0024] S3: Hazard endpoints are classified according to their level of concern. This study incorporates multiple indicators across three dimensions: health hazards, ecotoxicity, and environmental fate. It collects hazard data for alternative alternatives corresponding to these indicators, establishes a four-layer optimization method for heterogeneous data, and employs a data standardization process built using algorithms that combine geometric mean, majority vote, and the most sensitive principle. Based on hazard prediction using molecular descriptors and QSAR models, it analyzes and processes the hazard data for alternative alternatives and categorizes the hazard endpoints by concern level, obtaining qualitative or quantitative data for these concern levels. Specifically, this includes: S3-1 incorporates 15 indicators across three dimensions—health hazard, ecotoxicity, and environmental fate—as endpoint analysis parameters for hazard classification. Specifically, these include: 10 human health hazard endpoint parameters: carcinogenicity, mutagenicity, reproductive and developmental toxicity, acute mammalian toxicity, specific target organ toxicity (single exposure), skin irritation, eye irritation, endocrine disruption activity, specific target organ toxicity (repeated exposure), and sensitization; 2 ecotoxicity endpoint parameters: acute aquatic toxicity and chronic aquatic toxicity; and 3 environmental fate endpoint parameters: persistence, bioaccumulation, and migration. S3-2. Collect hazard data corresponding to 15 hazard classification endpoint indicators for alternative alternatives. Collect four types of data from multiple sources: GHS classification data, REACH registration data, OECD GLP certification experimental data, and QSAR model prediction data. Perform four-level optimization on the multi-source data. Then, according to the priority principle of GHS classification data > REACH registration data > OECD GLP certification experimental data > QSAR model prediction data, select high-priority data for each endpoint. Analyze and process the hazard data of alternative alternatives using a data standardization processing program constructed by fusion algorithms based on geometric mean, majority vote, and the most sensitive principle, to obtain a usable dataset with high data quality. The four-layer optimization of heterogeneous data includes: First layer: Data reliability screening: assess the credibility of data sources, prioritize data from authoritative institutions and databases, and eliminate data with low reliability.

[0025] The second layer: Data relevance screening: Based on the intended use, data in scenarios related to the intended use of the substitute material are screened to ensure that the data is consistent with the actual application.

[0026] The third layer: Data integrity screening: Prioritize records that contain multiple key parameters and have high data integrity, and supplement or use data with missing key parameters with caution.

[0027] Fourth layer: Data consistency screening: Compare data from different sources, analyze the reasons for significant differences, and use algorithms such as majority vote and most sensitive principle to determine the final data to be used; S3-3: Hazard Prediction Based on Molecular Descriptors and QSAR Models Based on the quantitative structure-activity relationship (QSAR) model, the SMILES codes of alternative substitutes are input, the molecular descriptors LogP, TPSA and the number of hydrogen bond donors are extracted, and the quantitative predicted values ​​of carcinogenicity and mutagenicity are output and incorporated into the hazard data standardization process. Using the data processed in step S3-2, combined with the quantitative predictions from the QSAR model, the composition of the hazard data is divided into five categories. Based on these categories, the data is processed as follows, and the hazard indicators for each alternative are calculated: S3-3-1, Scenario 1: If both quantitative and qualitative data are available for the hazard endpoint, quantitative data should be prioritized. S3-3-2, Case 2: For GHS classification and REACH classification data, the most sensitive classification result under the latest GHS revision is preferred; S3-3-3, Case 3: For quantitative experimental data for OECD GLP certification, first take the geometric mean of the data for the same species and under the same experimental conditions, and then take the most sensitive result for the data for different species and under different experimental conditions to calculate the hazard index data. S3-3-4, Case 4: For qualitative experimental data for OECD GLP certification, first take the majority vote for data of the same species and under the same experimental conditions, and then take the majority vote for data of different species and under different experimental conditions to calculate the hazard index data. S3-3-5, Case 5: For quantitative / qualitative data predicted by the QSAR model, before processing according to the methods in Case 2 / 3, first select the prediction result with the highest model confidence. S3-4. Based on the hazard index data calculated in step S3-3, the concern level of hazard endpoints is classified to obtain quantitative data for classification. Among them, the concern level of hazard endpoints is classified as follows: Based on the calculated hazard data and referring to the hazard endpoint concern level classification standard, the six hazard endpoints of carcinogenicity, mutagenicity, reproductive and developmental toxicity, endocrine disruption activity, specific target organ toxicity - repeated exposure, and sensitization are divided into three levels: "high concern", "medium concern", and "low concern". The nine hazard endpoints of mammalian acute toxicity, specific target organ toxicity - single exposure, skin irritation, eye irritation, acute aquatic toxicity, chronic aquatic toxicity, persistence, bioaccumulation, and migration are divided into four levels: "very high concern", "high concern", "medium concern", and "low concern". S4: Evaluation and Classification of Alternatives A analytic hierarchy process (AHP) was employed to construct a four-level classification method for alternatives (Level I: Green, Level IV: Red). Core parameters related to human health, ecological safety, and environmental migration and transformation were used as the criteria layer factors in the AHP. Based on the relative importance of each factor, a comprehensive score was calculated for each alternative to assess its suitability and determine its classification. Then, the classification results were adjusted based on regulatory consistency verification, ultimately selecting the greener and safest alternatives. The specific steps include: S4-1. Using the Analytic Hierarchy Process (AHP), a four-level classification method (Level I green to Level IV red) is constructed. The core parameters of human health, ecological security and environmental migration and transformation are used as the criteria layer factors of the AHP. Based on the relative importance of each factor, the comprehensive score of each alternative is calculated, the alternative is evaluated, and the classification of each alternative is determined. The four-level classification method is based on the combination of different hazard endpoint concern classification results calculated in steps S3-4, and assesses alternatives into four levels: Level I (green), Level II (yellow), Level III (orange), and Level IV (red). Specifically: If an alternative substance meets any of the following criteria in the Chemical Substances List: PBT, vPvB, PMT, vPvM, or CMR, it will be assessed as Level IV. Among them, PBT refers to persistent, bioaccumulative, and toxic chemicals; vPvB refers to very persistent and very bioaccumulative chemicals; PMT refers to persistent, mobile, and toxic chemicals; vPvM refers to very persistent and very mobile chemicals; and CMR refers to chemicals that are carcinogenic, mutagenic, or reprotoxic.

[0028] If a substitute hazard endpoint concern combination meets any of the criteria of PT, BT, MT, PB, PM or any human health or ecological endpoint concern level of the highest level ("very high concern" or "high concern"), it will be assessed as Level III. The combinations PT, BT, MT, PB, and PM are used to identify chemicals that do not possess the complete PBT / PMT characteristics but still have significant hazard potential. Among them, PT refers to persistence (P) + toxicity (T), BT refers to bioaccumulation (B) + toxicity (T), MT refers to mobility (M) + toxicity (T), PB refers to persistence (P) + bioaccumulation (B), and PM refers to persistence (P) + mobility (M) chemicals. An alternative can be assessed as Level I only if all hazard endpoints are classified as "Low Concern". If an alternative is not classified as Level I, III, or IV, it is assessed as Level II. S4-2. Based on the results of regulatory consistency verification, adjust and optimize the grading results to ensure that the substitutes must meet the functional requirements of the original substances, avoid functional deficiencies, and ultimately select greener and safer substitutes. S5. Construct auxiliary screening systems for greener and safer alternatives to chemical substances (such as...) Figure 1 As shown), this system is used to execute steps S1-S4. Based on the user-input CAS number or usage information of the chemical substance, the system queries and automatically outputs the results of screening for selectable greener and safer alternatives. For chemical substances for which no screening results are yet available, hazard endpoint calculation data is input until a greener and safer alternative (such as...) can be screened. Figure 3 (As shown).

[0029] A use-oriented chemical substance greener and safer alternative auxiliary screening system for implementing the greener and safer alternative auxiliary screening method includes a network server, management terminal, user access terminal and external database connected and communicating via the Internet; The network server has a built-in program to assist in screening greener and safer alternatives by executing steps S1-S4. The program is based on a three-layer collaborative data system of "use-substance-evaluation results" to assist in screening greener and safer alternatives, and includes: an input unit, a search and verification unit, an alternative evaluation unit, and an output unit. The external databases include: a database of substituted chemical substances and their uses, and a database of laws and regulations; The database of replaced chemical substances and their uses includes: Structured database, functional application database FUse, chemical substance data for the industrial chain; domestic chemical substance control data, chemical substance data for international conventions; The legal and regulatory database includes: official directories and industry standard data, and an international regulatory database. The management terminal is used to manage the auxiliary screening program for greener and safer alternatives; The user access terminal is used to provide users with an interactive access interface. Users input the CAS number or purpose of the chemical substance being replaced, and the system automatically outputs auxiliary search results for greener and safer alternatives.

[0030] In the process of assisting in the screening of greener and safer alternatives The input unit includes: an input module and a data standardization processing module; The retrieval and verification unit includes: a database module for a list of alternative alternatives, a module for mapping substances by industry application, a database module for information on the toxicity of substances; a priority sorting module, a molecular group identification module, and a safety elimination module; among them, the molecular group identification module calls a pre-trained molecular group identification model to score the functional group matching degree of alternative alternatives and eliminates substances with a functional matching degree of <80% or containing toxic groups. The alternative evaluation unit includes: a hazard classification endpoint analysis parameter module, a heterogeneous data four-layer optimization module, a hazard quantitative prediction module, a hazard endpoint attention classification module, a hierarchical analysis module, an evaluation algorithm model library, and a regulatory consistency verification module. The hazard quantitative prediction module has a built-in hazard prediction model based on molecular descriptors and QSAR, and the hierarchical analysis module has a built-in LSTM model. The evaluation algorithm model library has a built-in geometric mean, majority vote, and most sensitive principle fusion algorithm model, and a molecular structure safety evaluation model. The output units include: an evaluation results database and a visualization module.

[0031] In this embodiment, the application-oriented approach refers to requiring that the substitutes meet the functional requirements of the substituted material in the original industrial application scenario during the screening of substitutes, including but not limited to physicochemical performance parameters, process compatibility indicators, etc. The aforementioned data-driven approach refers to an auxiliary screening system based on multi-source data, which is composed of standardized data source integration and processing, multi-dimensional feature calculation, and dynamic weight allocation, to assist in the screening of greener and safer alternatives.

[0032] The specific physiological indicators for the "health hazard" dimension, including the criteria for determining carcinogenicity, mutagenicity, etc., shall refer to the standards specified in the OECD Test Guidelines 451-453 series.

[0033] The following combination Figures 4 to 10 The following, along with several more specific embodiments, will be described in detail.

[0034] Example 1 See Figures 4-6 The application-oriented method and system for assisting in the screening of greener and safer alternatives to chemical substances provided in this embodiment are based on the basic embodiment and applied in a specific way. Taking the scenario of bisphenol A (80-05-7) in the coating industry, where there is a demand for alternatives to highly hazardous solvents but no alternative formulations are currently available, as an example, it is basically the same as the basic embodiment. The difference is that the method for assisting in the screening of greener and safer alternatives to bisphenol A (80-05-7) in the coating industry further includes the following steps: S1: Screening for alternative chemical substances and their uses Based on the existing environmental management list of chemical substances in the coatings industry, chemical substances that need to be replaced are screened, a coatings industry use-substance mapping library and a substance toxicity information database are constructed for retrieval and verification units, and then the uses of the replaced substances are confirmed.

[0035] Constructing a coating industry application-material mapping library: Information on chemical substances in the coating industry was collected from seven sources, including databases and public reports, coating industry and enterprise information, and literature retrieval. After merging the chemical substance data from the seven sources, chemical substances without CAS numbers were first removed, and 9,923 data entries for 2,530 chemical substances were retained. Information on the uses of chemical substances in the coating industry was collected from publicly available formulation databases of coating companies, the US ChemExpo database, the Chemview database, and the Comptox database. This information was processed with reference to the OECD's standardized functional uses. A total of 1661 chemical substances were assigned 56 functional uses, while information on the functional uses of 869 chemical substances was missing. This information was supplemented primarily using the ChemBK and ChemicalBook websites, which provide information on the functional uses of chemical substances in all usage scenarios. Therefore, functional uses clearly relevant to the coating industry were selected. For example, the main use of tricarboxymethyl ether (25723-16-4) is described as "used as a rust inhibitor, defoamer, and thickener in metal processing, painting, and other industrial processes." Uses relevant to the coating industry, namely defoamer and thickener, were selected as supplementary functional uses of tricarboxymethyl ether and added to the list, thus constructing a coating industry use-substance mapping library. A database of toxic substance information was constructed: based on the chemical substances in the coating industry's application-substance mapping library, 17 hazard endpoint indicators of concern to the coating industry were selected in 4 categories, including 10 health hazards (carcinogenicity, mutagenicity, reproductive / developmental toxicity, endocrine disruption effects, acute mammalian toxicity, specific target organ toxicity - single exposure, specific target organ toxicity - repeated exposure, sensitization, skin irritation, and eye irritation), 2 ecotoxicities (acute aquatic toxicity and chronic aquatic toxicity), and 3 environmental fates (persistence, bioaccumulation, and migration). All hazard data were collected for the assessment of 2,530 chemical substances in the coating industry.

[0036] S2: Establish a list of alternative alternatives Same as the basic embodiment; S3: Hazard endpoints are classified according to their level of concern. Taking bisphenol A (80-05-7) as an example, the hazard information of chemical substances in the coating industry was collected from four sources: GHS classification information, REACH source information, experimental data, and predicted data. The method for collecting chemical hazard endpoint data and optimizing multi-source data is as follows: (1) Confirm the bisphenol A labeling information after integration through 3.3.2 chemical substance information, as shown in Table 1.

[0037] Table 1

[0038] (2) Based on the chemical substance identification information, GHS classification information of bisphenol A was collected from GHS databases in China, Japan, the European Union, Australia and New Zealand (see Table 2, search results of bisphenol A in GHS databases of various countries). For endpoints with existing classifications and unique / consistent classification results, the classification results were directly adopted. For example, the mutagenicity was classified as no classification and the eye irritation was classified as category 1. For endpoints with multiple sources having GHS classifications and inconsistent classification results, the most sensitive hazard endpoint was selected. For example, the reproductive toxicity was classified as category 1B. Finally, the GHS classification results of 10 hazard endpoints of bisphenol A were obtained. The carcinogenicity and ozone depletion effect endpoints with no data were marked as DG, and REACH source information was collected.

[0039] Table 2

[0040] Search the EU ECHA CHEM database for REACH source information on the carcinogenicity of bisphenol A (see the REACH endpoint search results for bisphenol A carcinogenicity). Figure 4 The recorded result was "no classification," therefore, the supplementary data on the carcinogenicity of bisphenol A was also "no classification." The REACH source information on ozone depletion effects was then queried (the query results for the ozone depletion effect endpoints of bisphenol A in REACH can be found in...). Figure 5 The result was "not evaluated," therefore, the ozone depletion effect endpoint retained the "DG" result.

[0041] For endpoints lacking GHS and REACH source information, including six hazard endpoints—endocrine disruption, persistence, bioaccumulation, migration, ozone depletion, and greenhouse effect—experimental data for bisphenol A were collected. The results are shown in Table 3 (Experimental Data Collection Results for Bisphenol A). All "DG" endpoints were supplemented through experimental data collection, thus eliminating the need for predictive models to fill data gaps. The data were recorded, and the next step of hazard endpoint concern grading was carried out.

[0042] Table 3

[0043] The hazard endpoint concern classification results for bisphenol A are as follows: Figure 6 As shown.

[0044] S4: Substance-Use Based Alternative Assessment and Grading Screening The molecular evaluation and alternative screening of bisphenol A (BPA, CAS 80-05-7) includes the following steps: (1) Molecular descriptor extraction and QSAR prediction (hazard prediction): Target substance: Bisphenol A (SMILES: OC(Cc1ccc(O)cc1)c2ccc(O)cc2) Tool: RDKit for extracting descriptors LogP = 3.8 (High lipid solubility, risk of bioaccumulation) TPSA = 40.5 Ų (moderate polarity, easily crosses the blood-brain barrier) Hydrogen bond donor number = 2 (the two hydroxyl groups provide endocrine-disrupting activity). QSAR model prediction: Model: Graph Neural Network (GNN) based on the ToxCast and Endocrine Disruptome datasets. Input: Molecular diagram structure (atomic type, bond type) Output: Estrogen receptor α binding activity: IC50 = 5.2 nM (significant, >10 nM indicates low risk) Acute toxicity to aquatic organisms (LC50, fish): 4.7 mg / L → Highly toxic (threshold <10 mg / L) (2) Molecular dynamics simulation (environmental fate assessment): Simulation environment: Aqueous solution system (TIP3P water molecules, 298K, 1 atm) Tools: GROMACS 2022.3, CHARMM36 force field Degradation pathway analysis: Key step: Cleavage of the CO bond between the benzene rings of bisphenol A (rate-limiting step of degradation) Quantum chemical verification: Software: ORCA 5.0 (DFT / B3LYP / def2-TZVP basis set) Calculation results: CO bond breaking energy ΔE = 298 kJ / mol → High persistence ( ΔE >280 kJ / mol threshold) Degradation half-life prediction: based on ΔE Based on the Arrhenius equation, t1 / 2 (25℃) = 150 days → meets PBT (persistent, bioaccumulative, toxic) substance standards. (3) Analysis of molecular group activity and functional adaptation: Functional group identification: Core functional group: Bisphenol hydroxyl group (provides cross-linking of polymer chains) Detection of toxic groups: The para-hydroxyl group on the benzene ring (known for its endocrine-disrupting activity) → triggers a high-risk labeling process. Isopropyl bridging structure (metabolic stability → bioaccumulation) Alternative screening: Candidate substance: Bisphenol S (BPS, CAS 80-09-1) Functional compatibility analysis: Sulfonic acid group substitution for hydroxyl group → CNN model evaluation showed a function fit of 87% (threshold ≥ 85%). Toxicity screening: No phenolic hydroxyl group → estrogen receptor IC50 = 58 nM (low risk) Environmental trend comparison: BPS has a CS bond breaking energy ΔE = 210 kJ / mol → it is easily degraded (half-life t1 / 2 = 30 days). Technical effectiveness verification: Predictive accuracy: The error between BPA estrogen activity prediction and in vitro experiments (HEK293 cell reporter gene assay) is <8%; The calculated degradation half-life differs from the OECD 301D standard test result (t1 / 2 = 160 days) by 6.25%. The advantages of the alternative are: BPS has a functional matching rate of 87% and a bioaccumulation rate of LogP=1.2 (68% lower than BPA); The risk of endocrine disruption has been downgraded from Category IV (no alternative) to Category II (recommended alternative). The aforementioned QSAR model adopts a transfer learning architecture. Based on the pre-training on the ToxCast dataset, it is adapted to specific industry toxicology data through fine-tuning, and the model's R² value is improved from 0.62 to 0.81 (validated with the REACH database).

[0045] For the various solvents commonly used in the coatings industry, a search revealed 396 substances with solvent uses in the application-substance mapping library. Based on the diversified data already processed in the second step, an alternative assessment was conducted for these 396 substances.

[0046] The results of the evaluation of some solvent uses among 396 solvents are as follows: Figure 7 As shown, isopropyl palmitate (CAS: 142-91-6) and oleyl alcohol (CAS: 143-28-2) were selected as safer alternatives for use.

[0047] S5. Constructing an auxiliary screening system for greener and safer alternatives to chemical substances. Integrating a material mapping library covering 51 categories of coating applications and 2530 chemical substances, a chemical hazard data and classification results library, and a chemical substance sales price library, a coating industry substitute screening system is built. Figure 1The system supports dual searches based on CAS number and chemical substance usage, adapting to various scenarios of product component substitution in the coatings industry. Furthermore, because chemical substance classification results are closely related to international chemical management policies, the system connects to regulatory API interfaces to update regulatory dynamics in various countries in real time. For example, once a chemical substance has been assessed for inclusion in the EU SVHC list, the system updates the classification results for that chemical substance within 24 hours.

[0048] The application scenarios and usage procedures of the greener and safer alternative auxiliary screening method and system provided in this invention mainly include: The auxiliary screening system provided in this embodiment offers a CAS number input search, suitable for companies that already have target alternatives and need to confirm the component classification results in alternative formulations to select safer alternatives. Companies can directly search the evaluation results within the system to select alternatives. For usage input search, it is suitable for companies that have alternative needs but do not yet have target alternatives. It can search the database for existing chemical substances with the same purpose in the domestic and international coatings industry. The system automatically combines the classification results and costs to provide a list of recommended alternatives, making it easier for companies to select safer chemical substances for subsequent laboratory research. For example, users (company R&D personnel) can directly input the use of "solvent" to find existing chemical substances used as solvents in the coatings industry, their domestic purchase costs, and evaluation results. They can select safer alternatives for product development without having to conduct their own evaluation.

[0049] The alternative screening method and results provided in this embodiment improve the depth of molecular-level evaluation, the objectivity of data fusion, and the level of intelligence. They can also be used for cross-industry application adaptation based on transfer learning. Building upon a model trained with coating industry data (such as solvent applications), data on lubricants from the electronics industry (such as viscosity and dielectric constant parameters) are introduced for transfer learning to quickly generate alternative classification rules for new industries. When a user selects an electronics industry lubricant application, the system calls a pre-trained coating industry solvent model, combines it with a transfer learning algorithm to supplement lubrication performance parameters (viscosity > 50 mPa·s), and recalculates the alternative classification.

[0050] This invention, through the synergy of screening data, methods, and systems, solves the engineering implementation problem of intelligent screening systems. This application-oriented screening method achieves full-process coverage through an auxiliary screening system encompassing "data input - model calculation - result output - decision support," primarily addressing the compatibility issues of multi-module integration in existing technologies. While CN118586862A's system includes MSDS management, risk assessment, and approval processes, and CN120108556A includes molecular generation, property prediction, and Pareto optimization, and the OECD guidelines include toxicity testing methods, these three lack a unified technical interface, preventing direct access and resulting in data silos and low screening efficiency. In this invention, the multi-source data has undergone standardization, enabling direct communication between units, significantly improving data utilization and screening efficiency. This invention also addresses the issues of uncertainty quantification and lack of risk warning. The graph neural network attribute prediction model in CN120108556A provides predicted values ​​(e.g., logCMC = −0.86), but it does not quantify the prediction uncertainty (e.g., the 95% confidence interval is [−1.0, −0.72]), leading to uncontrollable decision-making risks. For example, if the predicted toxicity value of a certain substitute is at the "safe-hazard" boundary (e.g., logKow = 3.0, close to the bioaccumulation threshold), the existing system cannot prompt "the uncertainty of this result is 20%, and supplementary experimental verification is required"; the early warning mechanism in CN118586862A only targets approval timeouts and does not trigger early warnings for data uncertainty, which can easily lead to misjudgments. This embodiment solves this problem through quantitative prediction and quantitative analysis of multiple models.

[0051] Example 2 See Figure 8 The CAS query interface of the chemical substance greener and safer alternative auxiliary screening system for the coating industry is shown in this embodiment. The application-oriented chemical substance greener and safer alternative auxiliary screening method and system provided in this embodiment is also a specific application of the basic embodiment. It is basically the same as Embodiment 1. The difference is that this embodiment takes the scenario where a coating industry enterprise has already carried out laboratory research and development and passed a new formulation for highly hazardous solvents (such as toluene), and needs to select a safer ingredient in it as an example. The greener and safer alternative auxiliary screening method further includes the following steps: S1: Screening for alternative chemical substances and their uses (1) Constructing a coating application-material mapping library During the R&D process, companies identify a small number of chemical substances that meet the performance requirements of coatings and construct a targeted application-material mapping library. Taking solvents as an example, the selectable substances are shown in Table 5 (Application-Material Mapping Library in Specific Coating R&D Formulations): Table 5

[0052] The toxicity information database for substances covers 17 hazard endpoints across 4 categories in the coatings industry, including 10 health hazards (carcinogenicity, mutagenicity, reproductive / developmental toxicity, endocrine disruption effects, acute mammalian toxicity, specific target organ toxicity - single exposure, specific target organ toxicity - repeated exposure, sensitization, skin irritation, and eye irritation), 2 ecotoxicities (acute aquatic toxicity and chronic aquatic toxicity), and 3 environmental fates (persistence, bioaccumulation, and migration). All endpoint data for these substances are collected.

[0053] S2: Establish a list of alternative alternatives The process of selecting and establishing a list of alternatives from multiple data sources is basically the same as the steps in Example 1. S3: Hazard endpoints are classified according to their level of concern. The steps are basically the same as those in Example 1; S4: Evaluation and Classification of Alternatives The substance-use substitution assessment was conducted by first evaluating the data after data optimization in step S2. The assessment results are shown in Table 6 (Use-Substance Mapping Library in Specific Coating R&D Formulations): Table 6

[0054] Since the chemicals in Table 6 are those that have been confirmed by companies to be usable as solvents, dipropylene glycol butyl ether can be directly recommended as a safer alternative to highly hazardous solvents without going through a model validation phase.

[0055] More specifically, taking toluene (CAS 108-88-3), an aromatic hydrocarbon solvent, as an example, we evaluated its chemical substitutes by classifying concerns based on 17 hazard endpoints.

[0056] When evaluating a monocyclic aromatic hydrocarbon, the system first calculates the π-π stacking energy (E_π) of the benzene ring. If E_π > 10 kJ / mol, it is marked as having potential neurotoxicity risk. Volatility (vapor pressure P_v) and ozone formation potential (OFP) are then simulated using molecular dynamics. If P_v > 3 kPa or OFP > 0.5 g O 3 / g triggers an air pollution label; recommended alternatives are cyclohexane or ester solvents, and benzene ring structures should also be avoided.

[0057] Molecular group activity and functional adaptation analysis Functional group identification: Core functional group: benzene ring (provides solvent polarity and solubility); Side effect group: Methyl substitution (increases volatility and ozone generation potential).

[0058] Alternative screening: Candidate substances: cyclohexane (CAS 110-82-7), butyl acetate (CAS 123-86-4); The functional matching degree analysis is shown in Table 7: Table 7

[0059] Toxicity screening results: Cyclohexane: Neurotoxicity LOAEL = 200 mg / kg / day (safe), but LogP = 3.1 (risk of bioaccumulation); Butyl acetate: QSAR predicted carcinogenicity probability = 0.05 (safe), biodegradation half-life t1 / 2 = 10 days (easily degradable).

[0060] (4) Comprehensive classification and recommendation results Tolueneogen assessment results: Health hazard classification: Level III (neurotoxicity + potential carcinogenicity); Environmental hazard classification: Level III (high air pollution + persistent degradation).

[0061] Alternative optimization suggestions: Preferred alternative: Butyl acetate → 92% functional match, environmental classification I (low volatility + easy degradation); Industrial compatibility: Butyl acetate coatings have a 15% longer drying time than toluene, but VOC emissions are reduced by 70%.

[0062] Comparison of technical effects: The error between the QSAR model's predicted LOAEL and the experimental value is less than 5%; The vapor pressure obtained from molecular dynamics simulations is in perfect agreement with the experimental values. Butyl acetate is recommended as the best alternative, with a 7% improvement in solvent performance but a 10% increase in overall cost.

[0063] (5) Actual experimental verification Through actual laboratory research and small-scale studies, it was found that butyl acetate (123-86-4) has a strong broad-spectrum solubility; its solubility parameter δ=8.5, which is suitable for resin components in coatings; its boiling point is 126.1℃, and its evaporation rate is moderate, which can ensure both dissolution effect and control drying time, making it suitable for coating spraying; its viscosity is 0.734 mPa·s (20℃), which has good fluidity and is easy to mix with other solvents, making it a substitute for toluene solvent.

[0064] This invention further solves the problem of the difficulty in coordinating "function-safety-cost" in multi-objective optimization. Since green and safe alternatives must simultaneously meet the requirements of "functionality meeting standards, low safety and low toxicity, and controllable cost," existing technologies have several shortcomings in multi-objective collaborative optimization, including: (1) the difficulty in balancing the inverse conflict between safety and function. CN120108556A, based on Pareto optimization theory, achieves "functionality (… TS , logCMC )-harm( logKow The study considers the synergistic effect of functional enhancement on surface activity (functional optimization) and toxicity, but does not take into account the nonlinear relationship that may lead to increased safety risks: for example, a certain PFAS substitute improves surface activity (functional optimization) by increasing the length of the fluorine chain, but at the same time leads to bioaccumulation ( logKow As security deteriorates, existing Pareto boundary identification methods can only screen for "non-dominated solutions" and cannot provide optimal security decision recommendations with acceptable functional loss (such as allowing...). TS When the fluctuation is 5%, logKow (1) The minimum value can be reduced to how much? The local toxicity test of the OECD guidelines can only determine whether there is toxicity. It cannot quantify the trade-off between the degree of toxicity and the contribution of function (e.g., the stronger the antibacterial function of a preservative, the higher the risk of skin sensitization. Existing methods cannot determine the optimal balance point of "function-toxicity"). (2) The lack of economic cost dimension leads to insufficient feasibility. None of the above existing technologies are included in the cost assessment of alternative production: for example, an alternative is designed through deep learning (CN120108556A) and meets the toxicity test (OECD guidelines). However, due to the complexity of the synthesis process (requiring rare catalysts), the cost is 10 times that of the original substance, which makes it impossible for the industry to actually adopt it. Existing screening systems only focus on technical feasibility and ignore economic feasibility, which makes it impossible for a large number of laboratory-optimal alternatives to be implemented as industrial-optimal alternatives. However, the embodiments of the present invention achieve the "function-safety-cost" synergy of green and safe alternatives through the combination of graded screening, actual laboratory research and development, and pilot-scale research.

[0065] Example 3 See Figure 9 The application-oriented method and system for assisting in the screening of greener and safer alternatives to chemical substances provided in this embodiment is also a specific application of the basic embodiment. It is basically the same as Embodiment 1 and Embodiment 2. The difference is that, taking the scenario where an electronics manufacturing company already has a new laboratory formulation for a highly hazardous lubricant and needs to select a safer formulation from it as an example (the chemical to be replaced is 3-amino-4-octanol), the application-oriented method for assisting in the screening of greener and safer alternatives to chemical substances further includes the following steps: S1: Screening for alternative chemical substances and their uses When constructing the coating application-substance mapping library, for the new formulations already determined in the electronics manufacturing industry (Table 8: Lubricant Formulation Ingredient Table for Electronics Manufacturing Industry), a database of toxic chemical information for the formulation chemicals was constructed. The database contains 17 hazard endpoint indicators in 4 categories, including 10 health hazards (carcinogenicity, mutagenicity, reproductive / developmental toxicity, endocrine disruption effects, acute mammalian toxicity, specific target organ toxicity - single exposure, specific target organ toxicity - repeated exposure, sensitization, skin irritation, and eye irritation), 2 ecotoxicities (acute aquatic toxicity and chronic aquatic toxicity), and 3 environmental fates (persistence, bioaccumulation, and migration). All endpoint data for the above substances were collected.

[0066] Table 8

[0067] S2: Establish a list of alternative alternatives The priority sorting of multi-source data is consistent with the steps in Example 1 or Example 2. S3: Hazard endpoints are classified according to their level of concern. The steps are consistent with those in Example 1 or Example 2; S4: Evaluation and Classification of Alternatives (The substance-use substitution assessment is conducted based on the data optimization carried out in step S2. The assessment results are shown in Table 9 (Evaluation Results of Lubricant Formulation Components in the Electronics Manufacturing Industry).)

[0068] Table 9

[0069] Since the chemicals in Table 9 are those that have been confirmed by the companies to be usable as lubricants, we can directly recommend isopropyl oleate or glyceryl tristearate as safer alternatives to high-hazard lubricants without going through a model validation stage.

[0070] Example 4 See Figure 10 The application-oriented method and system for screening greener and safer alternatives to chemical substances provided in this embodiment is also a specific application of the basic embodiment. It is basically the same as that in embodiments 1 to 3. The difference is that it takes the textile printing and dyeing industry's demand for alternatives to highly hazardous components, taking diisooctylamine (CAS106-20-7) as an example, and uses the scenario of finding safer alternatives from the system as an example for explanation.

[0071] When evaluating a long-chain alkylamine, the molecular descriptor needs to additionally calculate the pKa value of the amine group. If the pKa > 10, it is marked as a strong alkaline corrosion risk. The pathway for amine oxidation to nitrosamine is simulated using molecular dynamics; if the nitrosamine formation energy barrier is present... ΔG If the concentration is less than 80 kJ / mol, it triggers a high risk of carcinogenicity; recommended alternatives are amide or ether amine compounds to avoid the activity of free amine groups.

[0072] The aforementioned method for screening greener and safer alternatives to application-oriented chemicals further includes the following steps: molecular descriptor analysis, QSAR prediction, environmental fate simulation, and functional group adaptation analysis. 1. Molecular descriptor extraction and QSAR prediction Target substance: Diisooctylamine (SMILES: CC(C)CC(C)N(C)CC(C)C) Descriptor extraction (ChemAxon computation): LogP = 5.2 (high lipid solubility, risk of bioaccumulation); TPSA = 26.3 Ų (low polarity, easily permeable to biomembranes); pKa = 10.8 (strongly alkaline, risk of skin corrosion). Hydrogen bond donor number = 1 (free amine activity).

[0073] QSAR model prediction: Model selection: Gradient Boosting Tree (GBDT) model based on the TOXNET and ECOTOX datasets; Input parameters: LogP, pKa, molecular weight (MW=215.4), topological polar surface area (TPSA); Prediction results: The carcinogenic probability of nitrosamines is 0.72 (threshold > 0.5 → high risk). Acute toxicity to fish (LC50) = 2.3 mg / L → extremely toxic (threshold <1 mg / L is a level I risk).

[0074] 2. Molecular dynamics simulations and quantum chemical verification Simulation of the nitrosamine formation pathway: System construction: Diisooctylamine and NO2 free radicals in simulated body fluid (pH=7.4, 310K). Force field parameters: OPLS-AA force field (amine reactivity correction); Simulation software: NAMD 3.0 (simulation time 50 ns).

[0075] Key results: The energy barrier for nitrosamine formation is ΔG‡=75 kJ / mol → high risk (ΔG‡<80 kJ / mol threshold). The reaction rate constant k = 3.8 × 10⁻ 4 s⁻¹ → Half-life t1 / 2 = 5 hours (rapid generation).

[0076] Quantum chemical verification (Gaussian 16, DFT / B3LYP / 6-311++G**): The transition state energy of amino oxidation: E_trans = 180 kJ / mol → consistent with the kinetic simulation.

[0077] 3. Analysis of molecular group activity and functional adaptation Functional group identification: Core functional group: long-chain alkylamine group (surfactant emulsifying function); Side effects: free amine groups (oxidative carcinogenicity), high LogP value (bioaccumulation).

[0078] Alternative screening: Candidate substances: Isooctylamide (CAS 1070-83-3) and polyetheramine (Jeffamine M-600, CAS 9046-10-0). Functional matching degree analysis results are shown in Table 10: Table 10

[0079] Chemical stability verification: Isooctamide: Amide bond hydrolysis energy ΔE = 210 kJ / mol → Easily degraded (ΔE < 250 kJ / mol); Polyetheramine: Ether bond oxidation energy ΔE=320 kJ / mol → High stability (suitable for industrial applications).

[0080] 4. Overall Grading and Recommendation Results Diisooctylamine pro-assessment results: Health hazard classification: Level IV (highly corrosive + highly carcinogenic); Environmental hazard classification: Level IV (highly toxic + difficult to degrade).

[0081] Alternative optimization suggestions: Preferred alternative: Polyetheramine (Jeffamine M-600) → 93% functional match, toxicity reduced to Grade I; Alternative option: Isooctamide → Better biodegradability (half-life t1 / 2 = 10 days).

[0082] 5. Verification of technical effects and data comparison; results are shown in Table 11. Table 11

[0083] This embodiment establishes a joint kinetic simulation and quantum chemical verification model for the oxidation of amines to nitrosamines, overcoming the limitation of traditional QSAR relying solely on descriptors. It also provides a mass-producible, low-toxicity alternative (polyetheramine is already in industrial production) for high-risk amines in the lubricant and surfactant industries. The QSAR model in this embodiment predicts an LC50 value with an error of less than 5% compared to the experimental value (2.1 mg / L); the simulated nitrosamine formation pathway is consistent with the results of in vitro metabolic experiments (HPLC-MS detection); and the recommended polyetheramine alternative offers only a 5% loss in surface activity performance and a 20% reduction in production costs.

[0084] Example 5 See Figure 8 The application-oriented method and system for assisting in the screening of greener and safer alternatives to chemical substances provided in this embodiment is also a specific application of the basic embodiment. It is basically the same as that in embodiments 1 to 4. The difference is that the scenario of finding safer alternatives from this auxiliary screening system is used as an example to illustrate the need for alternatives to highly hazardous components in the textile printing and dyeing industry, taking the ester compound dimethyl succinate as an example.

[0085] The application-oriented method for assisting in the screening of greener and safer alternatives to chemical substances provided in this embodiment further includes the following steps: S5. Construct an auxiliary screening system for greener and safer alternatives to chemical substances, which is used to execute steps S1-S4 through an interactive interface and automatically output the screening results.

[0086] This system integrates a usage-material mapping library covering 51 categories and 2530 chemical substances, a chemical hazard data and classification result library, and a chemical substance sales price library to build an industry-wide chemical substance safety substitution classification and screening system. Taking dimethyl succinate as an example, by entering "dimethyl succinate" in the search bar, the system can be searched using the auxiliary screening system. This system supports dual searches based on CAS number and chemical substance usage to adapt to different scenarios of product component substitution in the industry. Furthermore, since chemical substance classification results are closely related to international chemical management policies, the system connects to multiple external regulatory database API interfaces to update the regulatory dynamics of various countries in real time. For example, once a chemical substance has been assessed for inclusion in the EU SVHC list, the auxiliary screening system provided in this embodiment updates the classification results of the relevant chemical substance within 24 hours.

[0087] In the alternative selection technology and results using dimethyl succinate (CAS 629-14-1), an ester compound, the CAS number input search method of this system is applicable to enterprises that already have target alternatives and need to confirm the component classification results in the alternative formulations to select safer alternatives. Enterprises can directly input chemical substance information in the system's search bar to search, and the system will automatically output the evaluation results for enterprises to select alternatives.

[0088] It should be noted that when the chemical substance being evaluated is a saturated fatty acid ester, the quantum chemical fracture energy... ΔE The threshold is 200 kJ / mol. ΔE< It is classified as a readily degradable substance at 200 kJ / mol; the hydrolysis rate constant K_hydrolysis > 1 × 10⁻³ h⁻¹ of the ester group (RO-CO-O-R') needs to be verified by molecular dynamics simulation; the recommended alternatives are bio-based unsaturated esters or glycerol esters.

[0089] The molecular evaluation and alternative systematic screening calculation process for dimethyl succinate is as follows: (1) Molecular descriptor extraction and QSAR prediction Target substance: Dimethyl succinate (SMILES: COC(=O)CCC(=O)OC) Descriptor Extraction (RDKit Chemical Calculations): LogP=0.8 (low lipid solubility, low risk of bioaccumulation); TPSA = 60.2 Ų (high polarity, strong ability to migrate in the environment); The number of hydrogen bond acceptors is 4 (the oxygen atom in the ester group provides hydrolytic activity).

[0090] QSAR model prediction: Model selection: XGBoost model based on ECOTOX aquatic toxicity dataset (SHAP feature importance ranking); Input parameters: LogP, TPSA, molecular volume (V=147.5 ų); Prediction results: Acute toxicity of aquatic organisms (EC50, algae) = 120 mg / L → Low toxicity (threshold > 100 mg / L is safe); Probability of skin irritation = 0.15 (threshold < 0.3 is safe).

[0091] (2) Molecular dynamics simulation environment fate assessment Degradation pathway simulation: System construction: 20 dimethyl succinate molecules were dissolved in 5000 TIP3P water molecules (298K, 1 atm). Force field parameters: CHARMM36 corrected ester bond hydrolysis parameters; Simulation software: GROMACS 2022.3 (simulation time 100 ns).

[0092] Key results: Ester bond hydrolysis rate constant: K_hydrolysis = 2.5 × 10⁻³ h⁻¹ (pH = 7, 25℃); Degradation half-life: t1 / 2 = 280 h (approximately 11.7 days).

[0093] Quantum chemical verification (ORCA 5.0): The ester group (CO) cleavage energy ΔE = 185 kJ / mol → easily degraded (ΔE < 200 kJ / mol threshold).

[0094] (3) Analysis of molecular group activity and functional adaptation Functional group identification: Core functional group: diester group (-O-CO-O-) → solvent and plasticizing function; Side effect groups: No known high-risk groups (screened via ToxTree).

[0095] Alternative screening: Candidate substances: diethyl succinate (DES, CAS 123-25-1) and tributyl citrate (TBC, CAS 77-94-1). Functional matching degree analysis results are shown in Table 12: Table 12

[0096] Toxicity screening results: Diethyl succinate: LogP=1.2, EC50 (algae)=95 mg / L → upgraded to level III (threshold <100 mg / L); Tributyl citrate: LogP=4.5, triggering a high accumulation marker (LogP>3.5 threshold).

[0097] (4) Overall grading and recommendation results Evaluation results of dimethyl succinate: Environmental hazard classification: Level II (low toxicity, easily degradable); Health hazard classification: Level I (non-irritating).

[0098] (5) Suggestions for optimizing alternatives: Recommended option: Diethyl succinate (DES) → Environmental classification remains at Level II (faster degradation rate), but health hazard level rises to Level III; Preferred alternative: Bio-based dimethyl succinate (plant fermentation source) → Avoids the carbon footprint of petroleum-based feedstocks.

[0099] (6) Verification of technical effects and data comparison, the results are shown in Table 13: Table 13

[0100] In this embodiment, the error between the QSAR model's predicted toxicity and the experimental value is <10%; the error between the molecular dynamics simulation half-life (11.7 days) and the OECD 301B measured value (12 days) is only 2.5%; bio-based dimethyl succinate is recommended, as it has the best overall performance (100% functional matching, no change in chemical structure). This embodiment targets ester compounds, using a combined determination of hydrolysis rate and fracture energy to provide a bio-based alternative pathway, balancing environmental friendliness and industrial feasibility (avoiding performance loss), quantifying the correlation between functional matching (CNN model) and degradation kinetic parameters (GROMACS), and forming a complete technical solution.

[0101] Example 6 The application-oriented method and system for screening greener and safer alternatives to chemical substances provided in this embodiment is also a specific application of the basic embodiment. It is basically the same as that in embodiments 1 to 5. The difference is that the plasticizer dibutyl phthalate (DBP, CAS 84-74-2) is selected as the substance to be replaced. The scenario of using a combination of molecular descriptors and dynamic weights to find safer alternatives from the system is used as an example for illustration.

[0102] The method for assisting in the screening of greener and safer alternatives to application-oriented chemicals further includes the following steps: Molecular descriptor extraction: LogP=4.5 (high accumulation), TPSA=37.3 Ų (medium permeability); QSAR prediction: Carcinogenicity probability is 0.72 (threshold > 0.6 → High concern); Dynamic weight adjustment: The LSTM model increases the ecotoxicity weight to 0.5 based on recent environmental policies, causing DBP to be upgraded from level III to level IV; Alternative selection: The system recommends citrate ester (LogP=2.1, carcinogenic probability 0.08) as a Category I alternative.

[0103] Example 7 The application-oriented method and system for screening greener and safer alternatives to chemical substances provided in this embodiment is also a specific application of the basic embodiment. It is basically the same as that in embodiments 1 to 5. The difference is that the scenario of screening phthalate alternatives by combining molecular descriptors and dynamic weights to find safer alternatives from the system is used as an example for illustration.

[0104] The method for assisting in the screening of greener and safer alternatives to application-oriented chemicals further includes the following steps: (1) Target substance: diisononyl phthalate (DINP, CAS 28553-12-0) Tools: RDKit Cheminformatics Toolkit Output parameters: LogP = 9.2, TPSA = 52.1 Ų, number of hydrogen bond acceptors = 4 (2) QSAR model prediction: Model architecture: Random forest model based on the Tox21 dataset (1000 decision trees) Input data: LogP, TPSA, molecular weight (MW=418.6) Prediction result: Carcinogenicity probability P=0.84 (threshold>0.6→high risk) (3) Verification by quantum chemical calculations: Software: Gaussian 16 (DFT / B3LYP / 6-311+G** base set) Calculation results: CO bond breaking energy ΔE = 210 kJ / mol → classified as easily degradable (half-life < 60 days) (4) Molecular docking analysis: Target: CYP450 3A4 enzyme (PDB ID: 5W8F) Tools: AutoDock Vina Results: Binding energy = -9.3 kcal / mol → Predicted metabolic half-life T1 / 2 = 8h (5) Analysis of molecular group activity and functional adaptation Functional group identification: Core functional group: Phthalate diester group (plasticizing function, compatibility with PVC); Side effect groups: benzene ring structure (π-π stacking leads to endocrine disruption), long-chain alkyl (bioaccumulation).

[0105] Alternative screening: (6) Candidate substances: acetylated tributyl citrate (ATBC, CAS 77-90-7), epoxidized soybean oil (ESO, CAS8013-07-8). (7) Toxicity screening results: ATBC: Estrogen activity = 0.12 (safe), Daphnia NOEC = 1.2 mg / L (low toxicity); ESO: No endocrine-disrupting activity, but the toxicity of epoxy group metabolites needs to be verified (automatically marked with a yellow warning).

[0106] (8) Overall grading and recommendation results Original DINP assessment results: Health hazard classification: Level IV (high endocrine disruption + highly toxic); Environmental hazard classification: Level IV (recalcitrant + bioaccumulation).

[0107] Alternative optimization suggestions: Preferred alternatives: ATBC → Environmental Class II (intermediate durability), Health Class I (safe); Industrial validation: ATBC's PVC plasticizing efficiency is 95% of DINP's, and its volatility is reduced by 40% (TGA thermogravimetric analysis).

[0108] Example 8 The application-oriented chemical substance greener and safer alternative screening method and system provided in this embodiment is also a specific application of the basic embodiment. Taking the screening of alternatives to brominated flame retardants as an example, it provides an industry adaptability case. It is basically the same as that of Embodiments 1 to 7, except that alternatives to brominated flame retardants are screened from the system.

[0109] The method for assisting in the screening of greener and safer alternatives to application-oriented chemicals further includes the following steps: (1) Functional group matching: Target group: Brominated aromatic ring (core of flame retardant function) Alternative candidate: Aluminum diethylphosphinate (DEPAL, CAS 22560-50-5) Group analysis: CNN identified phosphooxy groups (-PO-O-) with a 92% matching rate. (2) Screening for toxic groups: Blacklist database includes 217 types of structures such as nitrobenzene and polybrominated diphenyl ethers. Screening results: DEPAL has no blacklisted groups and passed the initial screening. (3) Transfer learning adapted to the electronics industry: Basic model: Based on a dataset of 2000 flame retardants from the building materials industry. Migration parameters: dielectric constant requirements (ε>3.5), thermal stability (Td>300℃) Output results: DEPAL dielectric constant ε=4.2, thermal decomposition temperature Td=320℃ → Meets the Class II standard of the electronics industry. (4) Technical effects: Functional matching efficiency is 6 times higher than manual screening (20 minutes / compound); The success rate of adapting to industry-specific parameters has increased from 42% to 78%. The probability of mistakenly choosing a substitute containing a toxic group has been reduced to below 1%.

[0110] The embodiments of the present invention are all application-oriented methods and systems for assisting in the screening of greener and safer alternatives to chemical substances for the industrial manufacturing sector. They focus on addressing the problems of lack of green alternative screening methods, multi-source data integration, difficulty in prioritization, and lack of application-adaptation in traditional evaluation systems. They construct a "application-oriented, data-driven, and hierarchical screening" technical method to assist in the screening of the best safe alternatives to chemical substances for application-oriented purposes. The method supports dual-mode input of CAS search and application screening, and realizes intelligent and automated assistance in the screening of safer alternatives to chemical substances for application-oriented purposes.

[0111] It should be noted that in other embodiments of the present invention, other different solutions obtained by making specific selections within the scope of the steps, systems, models, algorithms, thresholds, parameters, substances, etc., described in the present invention can all achieve the technical effects described in the present invention, so the present invention will not list them one by one.

[0112] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any person skilled in the art can make many possible variations and modifications to the technical solution of the present invention using the methods and techniques disclosed above, or modify it into equivalent embodiments with equivalent changes, without departing from the scope of the present invention. All equivalent changes made to the components, proportions, and processes of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for assisting in the screening of application-oriented, greener, and safer alternatives to chemical substances, characterized in that, It targets the industrial manufacturing sector, incorporating core parameters regarding the impact of chemical substances on human health, ecological safety, and their environmental migration and transformation capabilities. It employs a use-oriented, data-driven, and tiered screening approach, including: screening chemical substances to be replaced based on environmental management lists and confirming their uses; extracting chemical substances with similar uses to establish a candidate library; incorporating multiple indicators across three dimensions—health hazards, ecotoxicity, and environmental fate—and using a four-layer optimization method based on heterogeneous data, combined with algorithms based on geometric mean, majority vote, and the most sensitive principle to process hazard data, and tiering the hazard endpoints based on their level of concern; and constructing a four-level tiered classification using the analytic hierarchy process (AHP), and after calculation and evaluation, assisting in the screening of greener and safer alternatives. The process includes the following steps: S1: Screening for alternative chemical substances and their uses Based on the existing environmental management inventory of chemical substances, chemical substances that need to be replaced were selected, and the uses of the replaced substances were confirmed; all the chemical substances mentioned have CAS numbers; S2: Establish a list of alternative alternatives Extract substances from existing chemical substances that have the same uses as the chemical substance to be replaced, and extract their functional molecular structures and molecular groups, eliminating substances already included in the environmental management list of existing chemical substances and their main molecular groups; all of the proposed alternatives have CAS numbers; S3: Hazard endpoints are classified according to their level of concern. This study incorporates multiple indicators across three dimensions: health hazards, ecotoxicity, and environmental fate. It collects hazard data of alternatives corresponding to these indicators, establishes a four-layer optimization method for heterogeneous data, and constructs a data standardization processing procedure using a fusion algorithm based on geometric mean, majority vote, and the most sensitive principle. Based on hazard prediction using molecular descriptors and QSAR models, it analyzes and processes hazard data of alternatives and classifies the hazard endpoints by concern level, obtaining qualitative or quantitative data on concern level classification. S4: Evaluation and Classification of Alternatives Using the Analytic Hierarchy Process (AHP), a four-level classification method for alternatives was constructed, ranging from Level I (green) to Level IV (red). Core parameters of human health, ecological safety, and environmental migration and transformation were used as criteria layer factors in the AHP. Based on the relative importance of each factor, the comprehensive score of each alternative was calculated to evaluate the alternatives and determine the classification of each alternative. Then, the classification results were adjusted based on the regulatory consistency verification effect, and finally, the greener and safer alternatives were selected.

2. The method for assisting in the screening of greener and safer alternatives to application-oriented chemical substances according to claim 1, characterized in that, It includes the following steps: S5. Construct an auxiliary screening system for greener and safer alternatives to chemical substances, which is used to execute steps S1-S4. The system queries based on the CAS number or usage information of the chemical substance input by the user. After running, it automatically outputs the screening results of selectable greener and safer alternatives. For chemical substances for which there are no screening results, the system inputs the hazard endpoint calculation data until greener and safer alternatives can be screened.

3. The method for assisting in the screening of greener and safer alternatives to application-oriented chemical substances according to claim 1, characterized in that, Step S1 specifically includes: S1-1: Screening for alternative chemical substances By comparing existing environmental management lists of chemical substances with toxicity information for initial screening, and then prioritizing the replacement of highly toxic substances with a strong need for substitution, the chemical substances to be replaced are selected, including the following steps: S1-1-1 First, collect a list of chemical substances with CAS numbers that are involved in the entire industrial chain; S1-1-2. Compare with the domestic chemical substance control list: By comparing the list, chemical substances in the list are considered the highest priority that need to be replaced; S1-1-3. Comparison with international convention chemical substance lists: By comparing with conventions, chemical substances within the scope of convention control are considered as secondary priority substances that need to be replaced. S1-1-4. Comparison with foreign chemical substance control lists: By comparing with the list of chemical substances of extremely high concern (SVHC), the list of restricted authorized chemical substances under the REACH regulation, and the list of PBT chemical substances under TSCA, chemical substances in the list are considered to be the third priority that needs to be replaced. S1-1-5. Screening for highly toxic chemicals: Based on existing data, screen for chemicals with PBT or vPvB properties, or those with carcinogenic, mutagenic, or reproductive toxicity. These chemicals are considered the last priority to be replaced. S1-2: Confirm the use of the chemical substance being replaced S1-2-1 Referencing Official Directory and Industry Standards By combining industry application scenarios and accessing official directories, industry standards, and international regulatory databases, we can obtain direct reference uses for the substances being replaced and determine their main uses. S1-2-2 Calling the Structured Database Access information from multiple databases to obtain conflicting data on the various uses of the substituted substance; S1-2-3 Call Function Purpose: Database FUSE resolves data conflicts The quantitative structure-function relationship (QSUR) model, constructed using the FUse database, resolves conflicting information regarding the multiple uses of the substitute substance; the random forest classification algorithm is used to predict whether a chemical possesses a specified function, and its safety is assessed by combining high-throughput screening data. S1-2-4 Addressing Data Uncertainty Finally, by combining experimental data with supply chain data, we can address data uncertainties and ensure the accuracy and comprehensiveness of the application.

4. The method for assisting in the screening of greener and safer alternatives to application-oriented chemical substances according to claim 1, characterized in that, Step S2 specifically includes: S2-1 Establish a list of alternative alternatives Based on the molecular structure, main molecular groups, and physicochemical properties of chemicals, and guided by their intended use, we search and compile chemical substances with the same uses as the chemical substances to be replaced in existing databases. Then, we extract substances with the same uses as the chemical substances to be replaced from the existing chemical substances and extract their functional molecular structures and molecular groups. We remove substances and their main molecular groups that have been included in the environmental management list of existing chemical substances and establish a list of alternative substitutes. Alternatively, machine learning models can be used for prediction, and their applications can be predicted. This involves analyzing molecular descriptors using deep learning neural networks, combined with... Ecoinvent The lifecycle of a database affects the data, allowing for the selection of alternatives with the same functionality; S2-2 Priority Sort For each alternative, its functional molecular structure and molecular groups are extracted, the safety of its functional molecular structure and molecular groups is evaluated, and the data sources are sorted according to priority sequence to establish a priority list of alternatives. S2-3 Safety Rejection Chemicals already included in the existing environmental management list of chemical substances are removed, and the remaining substances are identified as alternatives in order of priority.

5. The method for assisting in the screening of greener and safer alternatives to chemical substances according to claim 4, characterized in that, Step S2-2 further includes the following steps: Functional molecular structures and molecular groups are extracted from alternative molecules. First, the hierarchical analysis method is used to construct a four-level classification method from Level I (green) to Level IV (red). The core parameters of human health, ecological safety and environmental migration and transformation are used as the criteria layer factors of the hierarchical analysis method. Based on the relative importance of each factor, the comprehensive score of each alternative is calculated, the alternative is evaluated, the classification of each alternative is determined, and the safety is ranked after evaluation. Finally, a priority list of alternative alternatives is established. The weight adjustment of the analytic hierarchy process is achieved through a long short-term memory network (LSTM) model. The input data of the LSTM model are environmental policy change texts from historical alternative cases and expert weight assignment matrices, and the output is the dynamic weight parameters of the current assessment task.

6. The method for assisting in the screening of greener and safer alternatives to application-oriented chemical substances according to claim 1, characterized in that, Step S3 specifically includes: S3-1 incorporates 15 indicators across three dimensions—health hazard, ecotoxicity, and environmental fate—as endpoint analysis parameters for hazard classification. Specifically, these include: 10 human health hazard endpoint parameters: carcinogenicity, mutagenicity, reproductive and developmental toxicity, acute mammalian toxicity, specific target organ toxicity (single exposure), skin irritation, eye irritation, endocrine disruption activity, specific target organ toxicity (repeated exposure), and sensitization; 2 ecotoxicity endpoint parameters: acute aquatic toxicity and chronic aquatic toxicity; and 3 environmental fate endpoint parameters: persistence, bioaccumulation, and migration. S3-2 collects hazard data corresponding to 15 hazard classification endpoint indicators for alternative alternatives. Data is collected from multiple sources, including GHS classification data, REACH registration data, OECD GLP certification experimental data, and QSAR model prediction data. A four-layer optimization process is performed on the multi-source data. Then, prioritizing data for each endpoint according to the principle of GHS classification data > REACH registration data > OECD GLP certification experimental data > QSAR model prediction data, high-priority data is selected. A data standardization processing procedure, constructed using geometric mean, majority vote, and most sensitive principles, is employed to analyze and process the hazard data of alternative alternatives, resulting in a high-quality, usable dataset. The four-layer optimization of heterogeneous data includes: First layer: Data reliability screening: assess the credibility of data sources, prioritize data from authoritative institutions and databases, and eliminate data with low reliability; The second layer: Data relevance screening: Based on the intended use, data is screened in scenarios related to the intended use of the substitute material to ensure that the data is consistent with the actual application. The third layer: Data integrity screening: Prioritize records that contain multiple key parameters and have high data integrity, and supplement or use data with missing key parameters with caution; Fourth layer: Data consistency screening: Compare data from different sources, analyze the reasons for significant differences, and use a majority vote and most sensitive principle algorithm to determine the final data to be used; S3-3 Hazard Prediction Based on Molecular Descriptors and QSAR Models Based on the quantitative structure-activity relationship (QSAR) model, the SMILES codes of alternative substitutes are input, the molecular descriptors LogP, TPSA and the number of hydrogen bond donors are extracted, and the quantitative predicted values ​​of carcinogenicity and mutagenicity are output and incorporated into the hazard data standardization process. Using the data processed in step S3-2, combined with the quantitative predictions from the QSAR model, the composition of the hazard data is divided into five categories. Based on these categories, the data is processed as follows, and the hazard indicators for each alternative are calculated: S3-3-1 Scenario 1: If both quantitative and qualitative data are available for the hazard endpoint, quantitative data should be selected first. S3-3-2 Case 2: For GHS classification and REACH classification data, the most sensitive classification result under the latest GHS revision is preferred; S3-3-3 For quantitative experimental data for OECD GLP certification, the geometric mean is first taken for data of the same species and under the same experimental conditions, and then the most sensitive result is taken for data of different species and under different experimental conditions to calculate the hazard index data. S3-3-4 Case 4: For qualitative experimental data for OECD GLP certification, first take the majority vote for data of the same species and under the same experimental conditions, and then take the majority vote for data of different species and under different experimental conditions to calculate the hazard index data. S3-3-5 Case 5: For quantitative / qualitative data predicted by the QSAR model, before processing according to the methods in Case 2 / 3, first select the prediction result with the highest model confidence. Based on the hazard index data calculated in step S3-3, S3-4 classifies the hazard endpoints by level of concern, obtaining quantitative data for the classification. Specifically, the hazard endpoint concern level classification is as follows: Based on the calculated hazard data and referring to the hazard endpoint concern level classification standard, the six hazard endpoints—carcinogenicity, mutagenicity, reproductive and developmental toxicity, endocrine disruption activity, specific target organ toxicity—repeated exposure, and sensitization—are classified into three levels: "high concern," "moderate concern," and "low concern." The nine hazard endpoints—acute mammalian toxicity, specific target organ toxicity—single exposure, skin irritation, eye irritation, acute aquatic toxicity, chronic aquatic toxicity, persistence, bioaccumulation, and migration—are classified into four levels: "very high concern," "high concern," "moderate concern," and "low concern." 7. The method for assisting in the screening of greener and safer alternatives to application-oriented chemical substances according to claim 6, characterized in that, Step S4 specifically includes the following steps: S4-1 employs the Analytic Hierarchy Process (AHP) to construct a four-level classification method from Level I (green) to Level IV (red). The core parameters of human health, ecological safety, and environmental migration and transformation are used as the criteria layer factors of the AHP. Based on the relative importance of each factor, the comprehensive score of each alternative is calculated, the alternative is evaluated, and the classification of each alternative is determined. The four-level grading method, based on the combination of different hazard endpoint concern grading results calculated in steps S3-4, assesses alternatives into four levels: Level I (Green), Level II (Yellow), Level III (Orange), and Level IV (Red). Specifically: If an alternative substance meets any of the following criteria in the Chemical Substances List: PBT, vPvB, PMT, vPvM, or CMR, it will be assessed as Level IV. If a substitute hazard endpoint concern combination meets any of the criteria of PT, BT, MT, PB, PM or any human health or ecological endpoint concern level of the highest level "very high concern" or "high concern", it will be assessed as Level III. If all hazard endpoints of an alternative are classified as "low concern", the alternative can be assessed as Level I. If an alternative is not classified as Level I, III, or IV, it is assessed as Level II. Based on the compliance verification results, S4-2 adjusts and optimizes the grading results to ensure that the substitutes meet the functional requirements of the original substances, avoid functional deficiencies, and ultimately select greener and safer substitutes.

8. An application-oriented auxiliary screening system for greener and safer alternatives to chemical substances, characterized in that, The auxiliary screening method for greener and safer alternatives to chemical substances for implementing any one of claims 1 to 7 includes a network server, a management terminal, a user access terminal, and an external database that are connected and communicate via the Internet. The network server has a built-in program for assisting in the screening of greener and safer alternatives, which executes steps S1-S4. This program is based on a three-layer collaborative data system of "use-substance-evaluation results" to assist in the screening of greener and safer alternatives. It includes: an input unit, a search and verification unit, an alternative evaluation unit, and an output unit. The external databases include: a database of substituted chemical substances and their uses, and a database of laws and regulations; The database of replaced chemical substances and their uses includes: Structured database, functional application database FUse, chemical substance data for the industrial chain; domestic chemical substance control data, chemical substance data for international conventions; The legal and regulatory database includes: official directories and industry standard data, and an international regulatory database. The management terminal is used to manage the auxiliary screening program for greener and safer alternatives; The user access terminal is used to provide users with an interactive access interface. Users input the CAS number or purpose of the chemical substance being replaced, and the system automatically outputs auxiliary search results for greener and safer alternatives.

9. The application-oriented chemical substance greener and safer alternative screening system according to claim 8, characterized in that, In the process of assisting in the screening of greener and safer alternatives: The input unit includes: an input module and a data standardization processing module; The retrieval and verification unit includes: a database module for a list of alternative alternatives, a module for mapping substances by industry application, a database module for information on the toxicity of substances; a priority sorting module, a molecular group identification module, and a safety elimination module; among them, the molecular group identification module calls a pre-trained molecular group identification model to score the functional group matching degree of alternative alternatives and eliminates substances with a functional matching degree of <80% or containing toxic groups. The alternative evaluation unit includes: a hazard classification endpoint analysis parameter module, a heterogeneous data four-layer optimization module, a hazard quantitative prediction module, a hazard endpoint attention classification module, a hierarchical analysis module, an evaluation algorithm model library, and a regulatory consistency verification module. The hazard quantitative prediction module has a built-in hazard prediction model based on molecular descriptors and QSAR, and the hierarchical analysis module has a built-in LSTM model. The evaluation algorithm model library has a built-in geometric mean, majority vote, and most sensitive principle fusion algorithm model, and a molecular structure safety evaluation model. The output units include: an evaluation results database and a visualization module.

Images