Molecular comparison system and method thereof

Abstract

According to a molecular comparison system and a method thereof provided by the present disclosure, after a substance analysis result is obtained, on the basis of a molecular list and matching degree values as well as related questions and corresponding weights in a knowledge management system, a correlation database is used to determine and calculate by means of an AI LLM, so as to further determine the molecular species of a compound causing damage.

Description

Molecular alignment system and methods Technical Field

[0001] This invention relates generally to molecular alignment systems and methods, and more specifically, to systems and methods for analyzing potentially damaging molecules using knowledge management systems and large language models. Background Technology

[0002] Currently, wearable devices such as smart bracelets, due to their small size, cannot use traditional screws to secure their waterproof structures. Instead, they rely on adhesives to achieve a tight seal and waterproofing. However, once the adhesive cures, it may chemically degrade due to environmental chemicals, causing it to fail and compromising the device's seal and waterproofing. In such cases, testing is necessary to determine the specific chemical molecules causing the adhesive failure.

[0003] Many well-known methods for analyzing substances, such as FTIR (Fourier transform infrared) spectroscopy, can infer the types of molecules that may cause chemical damage through absorption spectra. However, FTIR can only provide general guidance on the functional groups of molecules and is insufficient to estimate the specific types of substances causing damage. Therefore, there is a need for systems and methods to identify the damaging molecules.

[0004] Summary of the Invention

[0005] In a first embodiment, this disclosure pertains to a molecular alignment method. The molecular alignment method includes obtaining multiple questions related to material destruction and corresponding weights from a knowledge management system based on multiple molecular categories in a molecular list. The molecular list contains multiple matching degree values ​​corresponding to the multiple molecular categories. The molecular alignment method also includes forming multiple AI question terms from the multiple molecular categories and multiple questions in the molecular list. The molecular alignment method also includes inputting the multiple AI question terms into at least one LLM (Limited Language Management System) and generating corresponding multiple AI answers. The molecular alignment method also includes calculating multiple weight scores for the multiple molecular categories based on the multiple AI answers, the multiple weights corresponding to the multiple questions, and the multiple matching degree values ​​corresponding to the multiple molecular categories. Based on the multiple weight scores, it determines which of the multiple molecular categories is the destructive molecule.

[0006] In a second embodiment, this disclosure pertains to a molecular alignment system. The molecular alignment system includes a substance analysis module for generating a molecular list based on the analyte. The molecular list includes multiple molecular types and multiple matching degree values ​​corresponding to these molecular types. The molecular alignment system also includes a knowledge management system, which includes multiple questions related to substance destruction and multiple weights corresponding to these questions. The molecular alignment system also includes an association database for obtaining multiple questions and their corresponding weights from the knowledge management system based on the multiple molecular types in the molecular list, and generating multiple AI question terms based on these multiple molecular types and questions. The molecular alignment system also includes at least one LLM (Limited Learning Manager) for receiving multiple AI question terms and generating corresponding multiple AI answers. The molecular alignment system also includes a weighted calculation module for calculating multiple weight scores for each of the multiple molecular types based on the multiple AI answers, the multiple weights corresponding to the multiple questions, and the multiple matching degree values ​​corresponding to the multiple molecular types. The molecular alignment system also includes a judgment module for determining which of the multiple molecular types is the destructive molecule based on the multiple weight scores.

[0007] To provide a better understanding of the above and other aspects of the present invention, specific embodiments are described below in conjunction with the accompanying drawings: Attached Figure Description

[0008] Figure 1 shows a schematic diagram of an example molecular alignment system according to certain aspects of this disclosure;

[0009] Figures 2A and 2B are schematic diagrams of an example list of molecules according to certain aspects of this disclosure;

[0010] Figure 3 is a flowchart of an example procedure for molecular alignment according to certain aspects of this disclosure.

[0011] Figure 100: System QS; Question 110: Material Analysis Module Q AI AI Question 120: Relationship Database (RA) AI AI Answer 130: Knowledge Management System RI: Related Information 140: RAG Architecture Rs: Comparison Results 141: LLM SP: Test Item 142: Retrieval Module WS: Weight 143: Private Domain Database WR: Weight Score 150: Weighted Calculation Module 300: Program 160: Judgment Module S310, S320, S330, S340, S350: Steps HQI: Matching Degree Value ML: Molecular List Detailed Implementation

[0012] Figure 1 illustrates a schematic diagram of an example molecular alignment system 100 according to certain aspects of this disclosure. System 100 may include a material analysis module 110, which can be used to generate a molecular list ML based on an analyte SP, such as an analyte SP sampled around a damaged substance (e.g., the substance causing the damage), for example, a molecular list generated through similarity comparison using an FTIR spectral database. This example uses the comparison to obtain the molecular types of a damaged polyurethane reactive (PUR) hot melt adhesive, but is not intended to be limiting. In this example, after obtaining the analyte around the damaged PUR adhesive, the material analysis module 110 can obtain a molecular list ML, such as the molecular list ML shown in Figure 2A. The molecular list ML includes multiple names representing different molecular types and multiple High Quality Indices (HQIs) corresponding to the multiple molecular types. In the molecular list ML of this example, possible molecular types (destructive substances) include L-Lysine (HQI 85.37%), L-Glutamic acid, magnesium salt, trihydrate (HQI 84.65%), Monaquest IA (HQI 84.55%), N-Hippuryl-His-Leuhydrate (HQI 84.31%), Zirconium carbonate, Hydrated (HQI 84.29%), etc.

[0013] System 100 may include a knowledge management system 130, which includes, for example, multiple questions QS related to material damage and multiple weights WS corresponding to the multiple questions, as shown in Table 1 below.

[0014] (Table 1)

[0015] In some implementations, the weights in the knowledge management system 130 can be represented by multiple values, each corresponding to different answers to a question. In this example, the binary value of the weight is represented by 1 or 2, for example, 1 indicates no relevance to the corresponding question, and 2 indicates relevance to the corresponding question. In other implementations, the weights can have three or more different values ​​to indicate the degree of relevance to the corresponding question. For example, the question "Does it exist in everyday life?" can have a weight between 0 and 4. A weight of 0 indicates no existence, a weight of 1 indicates a 25% probability of existence, a weight of 2 indicates a 50% probability of existence, a weight of 3 indicates a 75% probability of existence, and a weight of 4 indicates a 100% probability of existence.

[0016] In some implementations, multiple QS related to material damage in the knowledge management system 130 can be linked to glue failure analysis, and the multiple QS belong to multiple attributes, including environmental source, product source and pH value, as shown in Table 1.

[0017] System 100 may include an association database 120, which can be used to obtain multiple question words (QS) and corresponding weights (WS) from the knowledge management system 130 based on multiple molecular types in the molecular list ML, and to form multiple AI question words (Q) based on the multiple molecular types and multiple question words in the molecular list ML. AI In some implementations, AI question terms, including molecular species (for example, the top 5 molecular species with the highest HQI in the molecular list ML) and multiple questions, can be represented by a matrix table. System 100 may include at least one LLM 141. In this example, the input AI question term Q is... AI Taking the RAG (Retrieval-Augmented Generation) architecture 140, which includes LLM 141, as an example, but not as a limitation, the RAG architecture 140 may include a retrieval unit 142 for receiving AI question terms Q. AI And including a private database 143, used by the searcher 142 based on the AI ​​question word Q AI Search and provide questions related to AI-generated question words Q. AI The relevant information RI. Private domain database 143 may include chemical resistance manuals, such as Kuraray's Chemical Resistance Guide for Kuraray Resin, content from chemical journals such as the American Chemical Society and Elsevier, but is not limited to this, to provide AI-generated question keywords Q. AI The relevant information RI. Regarding the RAG architecture 140, it can help improve the accuracy of subsequent LLM responses, reducing the occurrence of AI illusions (responses that are too general and often irrelevant). After obtaining the relevant data RI, the AI ​​question word Q can be... AI Enter the relevant information RI into LLM 141 to obtain the AI ​​answer RA. AI In some implementations, the AI ​​question word Q can be used. AI Input multiple LLMs with relevant information RI, or only input the AI ​​question word Q. AI Input multiple LLMs to obtain AI answers from multiple different LLMs, thereby increasing the accuracy of the AI ​​answers.

[0018] System 100 may include a weighted calculation module 150, which is used to calculate the weighted calculation module 150 based on the obtained AI answer RA.AI The weights WS corresponding to multiple questions QS and the matching degree values ​​HQI corresponding to multiple molecular categories (included in the molecule list ML) are calculated to obtain multiple weight scores W corresponding to multiple molecular categories. R For example, the weighted calculation module 150 can generate different weight scores based on AI answers from different LLMs (LLM 1 and LLM 2), as shown in List 2 and Table 3 below.

[0019] (Table 2)

[0020] (Table 3)

[0021] The AI ​​answers generated by LMM can have any suitable format. In some implementations, the indicator can be a value corresponding to the weight of the question, as discussed above regarding the weight of the question. In the examples in Tables 2 and 3, the weighted calculation module 150 calculates a weight score based on the HQI and the weights corresponding to the AI ​​answers for each question. For example, in Table 2, the weight score of L-Lysine is the sum of the HQI multiplied by the weights corresponding to each AI answer (1365.92 = 85.37 x (2 + 2 + 2 + 2)).

[0022] System 100 may include a judgment module 160, which can determine which of multiple molecular types is the destructive molecule based on a weight score WR, and output a comparison result Rs. For example, the judgment module 160 can determine which molecule is the destructive molecule based on the weight scores in Table 2, or Table 3, or both Table 2 and Table 3. For example, in Table 3, L-Lysine has the highest weight score (1365.92), so the destructive molecule can be determined to be L-Lysine. In this example, the destructive molecule causing the failure of PUR adhesive is determined to be L-Lysine. The comparison result Rs can be output to a display device (not shown in the figure) to be displayed to the user, or stored as data in a media, such as a data storage device or non-volatile memory.

[0023] Next, the molecular alignment system 100 provided in this disclosure will be used as an example to analyze substances that cause damage to the conductive pads of a PCB (printed circuit board). In this example, after obtaining the analyte SP around the damaged conductive pad (causing conductive pad aberration), a molecular list ML can be obtained through the material analysis module 110. For example, a molecular list generated by similarity comparison using energy-dispersive X-ray spectroscopy (EDS) or gas chromatography-mass spectrometry (GCMS) in conjunction with an FTIR spectral database can be used. For example, after using a scanning electron microscope (SEM) to locate the possible location of the damaging substance, the analyte SP can be analyzed using EDS in conjunction with an FTIR spectral database to obtain the molecular list ML, as shown in Figure 2B. The EDS can also obtain the elemental composition percentage corresponding to the molecular list, similar to the aforementioned HQI. For example, in some embodiments, the elemental composition percentage of the EDS can be used alone instead of the HQI, or used together with the HQI to calculate the weighted score. In this example, the molecular list ML includes possible molecular types (destructive substances) such as 2-ETHYLHEXANAL (HQI 82.24%), Hexanal (HQI 81.65%), Cyclopentanecarboxaldehyde, 2-hexyl-,trans- (HQI 79.95), etc. Next, multiple questions (QS) related to substance destruction and corresponding weights (WS) from the knowledge management system 130 can be used, as shown in Table 1 below.

[0024] (Table 4)

[0025] Next, as explained above, the correlation database 120 can be used to obtain multiple question words (QS) and corresponding weights (WS) from the knowledge management system 130 based on multiple molecular types in the molecular list ML of Figure 2B (e.g., from Table 4), and form multiple AI question words (Q) based on the multiple molecular types in the molecular list ML and the multiple question words (QS). AI .

[0026] In this example, we also use the AI ​​question word Q as the input. AI Taking the RAG (Retrieval-Augmented Generation) architecture 140, which includes LLM 141, as an example, let's provide the relevant information RI for the AI ​​question term QAI. After obtaining the relevant data RI, the AI ​​question term QAI can be... AI Enter the relevant information RI into LLM 141 to obtain the AI ​​answer RA. AI .

[0027] Next, the AI ​​will ask the question Q. AI Input LLM weight calculation module 150, the weight calculation module 150 is used to calculate the AI ​​answer RA. AI The weights WS corresponding to multiple questions QS and the matching degree values ​​HQI (included in the molecule list ML) corresponding to multiple molecule types are calculated to obtain multiple weight scores W corresponding to multiple molecule types. R As shown in Table 5 below (only the three molecules with higher weight scores are listed).

[0028] (Table 5)

[0029] The judgment module 160 can determine, based on the weight scores in Table 5 above, that Abietic acid (312.2), which has the highest weight score, is the damaging molecule that causes poor conductive contacts.

[0030] In another example, the molecular alignment system 100 provided in this disclosure can analyze the damaging substance that causes the circuit pin breakage. Similarly, after obtaining the analyte SP around the broken circuit pin, a molecular list ML can be obtained through the material analysis module 110. For example, after using a scanning electron microscope (SEM) to find the possible location of the damaging substance, the analyte SP is analyzed using EDS in conjunction with an FTIR spectral database to obtain the molecular list ML and elemental composition percentages. The multiple questions QS related to the material damage and the multiple weights WS corresponding to the multiple questions in the knowledge management system 130 can be related to the elements contained in the pin itself (Fe, Co, Ni, etc.). For example, if it is an element contained in the pin itself, the weight WS is lower, and vice versa. Then, through a procedure similar to the aforementioned example (details omitted here), chlorine (Cl) with the highest weight fraction can be determined as the damaging molecule causing the circuit pin breakage.

[0031] Figure 3 is a flowchart of an example procedure 300 for molecular alignment according to specific aspects of this disclosure. In step 310, based on multiple molecular categories in a molecular list of the analyte obtained from a material analysis module (e.g., a molecular list ML obtained from the analyte SP in material analysis module 110 of Figure 1), multiple questions related to material damage (e.g., conductive contacts of PUR adhesive or PCB) and multiple weights corresponding to the multiple questions are obtained from a knowledge management system (e.g., knowledge management system 130 of Figure 1). The molecular list contains multiple matching degree values ​​(HQI) corresponding to the multiple molecular categories. In step 320, for example in the correlation database 120 of Figure 1, multiple AI question words (e.g., AI question word Q in Figure 1) are formed using the multiple molecular categories and multiple questions in the molecular list. AIIn step 330, multiple AI question terms are input into at least one LLM (e.g., LLM 141 in Figure 1), and multiple corresponding AI answers are generated (e.g., Table 2, Table 3, Table 5, or the AI ​​answer RA in Figure 1). AI In some implementations, as discussed above, AI answers can be generated by inputting AI question terms into a RAG architecture containing an LLM (e.g., RAG architecture 140 in Figure 1). In step 340, for example, the weighted calculation module 150 in Figure 1 can calculate multiple weight scores (e.g., weight scores W in Tables 2, 3, 5, or Figure 1) based on multiple AI answers, multiple weights corresponding to multiple questions, and multiple matching degree values ​​corresponding to multiple molecular categories. R In step S350, for example, the judgment module 160 of FIG1 can determine which of the multiple molecular types is the destructive molecule based on multiple weight scores. For example, by comparing the weight scores of each molecular type, the molecular type with the largest weight score is determined to be the destructive molecule. Program 300 can be implemented wholly or in part on any suitable computing device, such as system 100 or other computing devices. For example, in some cases, steps S320, S340 and S350 can be executed using system 100 to determine the destructive molecule. Steps S310 (obtaining multiple questions related to material destruction and multiple weights corresponding to the multiple questions) and S330 (inputting multiple AI query terms into the LLM) can be executed on system 100 or on other computing devices, such as other computing devices connected via a network.

[0032] The technology provided by various embodiments of this disclosure can effectively identify which molecules are destructive from a variety of molecular types. This can be used to improve products (e.g., wearable devices), replace the assumption of known molecules, and increase the accuracy of verification. For example, by identifying the specific destructive molecules causing PUR adhesive failure, it is possible to change the composition of the adhesive, develop new adhesives, or identify the specific destructive molecules causing poor conductive contacts on a PCB, thereby improving the durability and reliability of the product.

Claims

1. A method of molecular alignment, characterized by, include: Based on multiple molecular categories in a molecular list, a knowledge management system obtains multiple questions related to material destruction and multiple weights corresponding to these questions, wherein the molecular list contains multiple matching degree values ​​corresponding to these molecular categories. Multiple AI question terms are generated by combining the molecular types in the molecular list with the questions. Input these AI question terms into at least one LLM and generate multiple corresponding AI answers; Based on these AI answers, the weights corresponding to these questions, and the matching degree values ​​corresponding to these molecular types, multiple weight scores corresponding to these molecular types are calculated; and Based on these weighted scores, determine which of these molecular types is the disruptive molecule.

2. The molecular alignment method of claim 1, wherein, The weights corresponding to these problems have binary values ​​to indicate whether these molecular species are related to these problems.

3. The molecular alignment method of claim 1, wherein, The weights for these questions have three or more different values ​​to indicate the degree to which these molecular species are relevant to these questions.

4. The molecular alignment method of claim 1, wherein, Input these AI question terms into at least one LLM and generate corresponding AI answers, including: Input these AI query terms into a search engine; The search engine searches a private database for relevant information about the AI's query terms; and Input the relevant information and the AI ​​question words into at least one LLM, and generate the corresponding AI answers.

5. The molecular alignment method as described in claim 1, characterized in that, These issues related to material damage are relevant to adhesive failure analysis, and these issues fall under multiple attributes, including environmental origin, product origin, and pH value.

6. A molecular alignment system, characterized in that, include: A substance analysis module is used to generate a list of molecules based on an analyte. The list of molecules includes multiple molecule types and multiple matching degree values ​​for each molecule type. A knowledge management system, including multiple issues related to material destruction and multiple weights corresponding to these issues; A relational database is used to obtain the questions and their corresponding weights from the knowledge management system based on the types of molecules in the molecular list, and to generate multiple AI question terms based on the types of molecules in the molecular list and the questions. At least one LLM is required to receive these AI question words and generate multiple corresponding AI answers; A weighted calculation module is used to calculate multiple weight scores for each molecular category based on the AI ​​answers, the weights corresponding to the questions, and the matching degree values ​​for each molecular category; and A judgment module determines which of these molecular types is the destructive molecule based on these weight scores.

7. The molecular alignment system as described in claim 6, characterized in that, The weights of these questions in the knowledge management system have binary values ​​to indicate whether these molecular types are related to these questions.

8. The molecular alignment system as described in claim 6, characterized in that, The knowledge management system assigns three or more different weights to these questions to indicate the degree to which these molecular species are relevant to these questions.

9. The molecular alignment system as described in claim 6, characterized in that, Also includes: A search engine is used to receive these AI question terms; and A private database is used by the search engine to search based on these AI query terms and provides relevant information about these AI query terms. The relevant information and the AI ​​question words are input into at least one LLM, and corresponding AI answers are generated.

10. The molecular alignment system as described in claim 6, characterized in that, The knowledge management system links these issues related to material damage to the glue failure analysis, and these issues belong to multiple attributes, including environmental source, product source, and pH value.

Images