Automatic dispatching method for HVAC terminal equipment maintenance operations based on NLP

Abstract

The present invention discloses a method for automatically dispatching repair and maintenance work for HVAC terminal equipment based on NLP technology, which belongs to the field of intelligent mechanical and electrical operation and maintenance technology for civil buildings. The method for automatically dispatching repair and maintenance work for HVAC terminal equipment based on NLP technology of the present invention extracts a work order data set for repair and maintenance of HVAC terminal equipment from a source system, and performs Chinese word segmentation of the maintenance text content according to the configuration settings of the Chinese word segmenter. In order to solve the problem that the manual work order allocation mode cannot achieve full coverage and respond to user repair information due to working time constraints, which affects the timeliness of the handling of HVAC terminal equipment, an automatic labeling unit for the repair and maintenance fault classification of HVAC terminal equipment is completed through training and debugging, and an automatic fault identification device for the repair and maintenance work order of HVAC terminal equipment is produced. On this basis, a HVAC terminal equipment fault processing decision table is constructed to realize the automatic allocation of HVAC terminal equipment maintenance work orders through machine learning.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent electromechanical operation and maintenance of civil buildings, and specifically to a method for automatically dispatching repair and maintenance operations for HVAC terminal equipment based on NLP. Background Art

[0002] With the gradual popularization of the concept of the entire life cycle of buildings and its in-depth penetration into the operation and maintenance stage of building facility management, more and more building facility operation and maintenance projects are in the transformation stage of digital operation and maintenance. In these projects adopting digital building operation and maintenance, building facility management systems are generally used for equipment repair and maintenance management within the building to realize digital management of equipment maintenance operations within the building. Traditional HVAC terminal equipment repair and maintenance operations are performed by customer service personnel after receiving user reports or system alarms for HVAC equipment. Based on experience, they assign repair and maintenance work orders to the corresponding HVAC operation teams for execution. This method of dispatching work relies on the experience and judgment of customer service personnel. There is a situation where the speed of manual dispatching lags behind the requirements of the site, resulting in problems such as delayed maintenance services for HVAC terminal equipment. According to statistics, the timely processing rate of HVAC terminal equipment repair work orders using manual dispatching is only 85%. The terminal node equipment is for end users, which will lead to a decline in user service satisfaction.

[0003] 1) During the actual dispatch process, HVAC repair and maintenance work orders may have problem descriptions that differ from the specific HVAC terminal equipment maintenance professional classification, resulting in a "misassignment" probability of approximately 15%. Because "misassignment" can cause work orders to be assigned to the wrong maintenance object, the correct response time for the work order will be longer, resulting in a poor customer service experience.

[0004] 2) Existing work order assignment methods rely on manual labor. Therefore, the speed and accuracy of work order assignments depend on human experience and judgment. When the work order volume is high, human error can lead to work orders being assigned to the wrong work team, impacting the timeliness of work order execution. Data shows that the probability of "misassignment" due to human error is approximately 3%-5%.

[0005] 3) The manual work order allocation model cannot provide 24 / 7 full coverage response to user repair information due to working hours constraints, which affects the timeliness of HVAC terminal equipment handling. Therefore, it does not meet existing needs. Therefore, an automatic dispatching method for HVAC terminal equipment repair and maintenance work based on NLP is proposed. Summary of the Invention

[0006] The purpose of the present invention is to provide an automatic dispatching method for HVAC terminal equipment repair and maintenance work based on NLP. Through machine learning of the Chinese text semantics of HVAC terminal equipment maintenance work order descriptions, the HVAC terminal equipment fault classifier prediction model network and parameters are built and configured. Through training and debugging, the HVAC terminal equipment repair and maintenance fault classification automatic labeling unit is completed, and an automatic fault identification device for HVAC terminal equipment repair and maintenance work orders is produced. On this basis, a HVAC terminal equipment fault processing decision table is constructed to realize the automatic allocation of HVAC terminal equipment maintenance work orders through machine learning, which can solve the problems in the existing technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP, comprising the following steps:

[0008] Step 1: Extract a work order dataset for HVAC terminal equipment repair and maintenance from the source system. Perform Chinese word segmentation on the maintenance text based on the Chinese word segmenter configuration settings. To ensure accurate word segmentation, a stop word library should be pre-configured to exclude meaningless word segmentation results.

[0009] Step 2: extracting the target keywords from the original work order text according to the device attributes of the maintenance work order, which may include: at least one of the equipment location keywords and the equipment fault problem type keywords;

[0010] The equipment location keyword may include: at least one of the name and number information of the building, floor, space, and room corresponding to each building maintenance work order in the building maintenance work order set;

[0011] The equipment failure keywords may include: at least one of the name and number of the HVAC terminal equipment, the HVAC terminal node and the HVAC terminal equipment component;

[0012] The problem type keywords may include: at least one of descriptions of common problems in building maintenance and common words;

[0013] Step 3: Build the fault classifier network, configure the pre-training model parameters and the training and test dataset paths, and then start the pre-training process of the fault classifier. After the pre-training is completed, evaluate the effectiveness of the fault classifier and iteratively adjust the model pre-training process based on the evaluation results.

[0014] Step 4: When configuring the automatic judgment of the fault process node, the automatic dispatching device should create a decision table judgment mechanism for the unit and determine the number of HVAC terminal equipment fault judgment rules. A decision table with n conditions has 2n rules.

[0015] List all condition piles and action piles for HVAC terminal equipment fault judgment;

[0016] Fill in the conditions for judging the failure of HVAC terminal equipment;

[0017] Fill in the action items for HVAC terminal equipment fault processing to obtain the initial decision table;

[0018] Simplify the fault decision table and merge similar rules;

[0019] Using the ID3 algorithm as a benchmark, the calculation formula for the device decision information gain is as follows:

[0020] Information entropy formula for HVAC terminal equipment failure decision-making:

[0021]

[0022] Calculate the air volume, cooling effect, heating effect, and other four formulas respectively:

[0023]

[0024] Information gain formula:

[0025]

[0026] Finally, select the attribute with the highest information gain for splitting;

[0027] Step 5: Set up a dynamic workload assessment unit for dispatching HVAC terminal equipment repair and maintenance work orders. This processing unit's dispatch constraints are based on three factors: job queue length, job quality, and job duration. The unit selection logic assigns maintenance work orders to the candidate operator with the shortest current job queue length, the highest job quality, and the shortest historical job duration.

[0028] Step 6: In the linkage mechanism between the automatic dispatching device and the building facilities management system, set up a maintenance service process rule library based on the fault type according to the standard mapping process;

[0029] Step 7: In the linkage setting between the automatic dispatching device and the building facilities management system, verify the applicability of the work order triggering process according to the standard mapping process for the fault type.

[0030] Preferably, in step one, the extracted maintenance work order data set is divided into a training set and a test set in a ratio of 8:2, and the training set and the test set are labeled with data categories.

[0031] Preferably, in the step 1, the automatic dispatching device for HVAC terminal repair and maintenance work includes a Chinese corpus for HVAC terminal equipment repair and maintenance, a stop word library, a fault classifier pre-training model, a fault category decision judgment table, a service process library, a dispatching linkage mechanism, a work order execution mechanism, a work order rule library, and a repair and maintenance evaluation unit;

[0032] The Chinese corpus for HVAC terminal equipment repair and maintenance consists of a repair and maintenance vocabulary and word frequency. This corpus is created by segmenting HVAC terminal equipment maintenance work orders into Chinese words, and then using general maintenance vocabulary and HVAC terminal equipment-specific vocabulary to create the vocabulary.

[0033] Stop word library: It consists of a list of stop words that can exclude unnecessary or meaningless words;

[0034] Fault classifier pre-training model: This model consists of a classifier network model, a pre-training dataset, and a test dataset. Based on the results of HVAC terminal equipment fault classification data annotation, the HVAC terminal equipment fault classifier is pre-trained to generate a prediction model output for semantic feature recognition of HVAC terminal equipment fault repair report text.

[0035] Fault category decision table: This table consists of fault code, fault description, fault type, and work team. Based on the decision table mechanism, the judgment logic for assigning work teams to HVAC terminal equipment is constructed, and service priorities are set.

[0036] Service process library: It consists of fault code, service process code, and service process priority. It matches work order service processes based on the service priority identified by the fault category decision table and the fault type identified by the fault classifier.

[0037] Dispatch linkage mechanism: It consists of service process code, work order process code, and trigger instructions, and performs work order allocation actions based on the dispatch rules set in the service process library;

[0038] Work order execution mechanism: It consists of a work order workflow engine and a work order message notification module. It triggers the work order workflow engine actions based on the dispatch linkage mechanism message mechanism.

[0039] Work order rule base: composed of service process code and maintenance team code, generates SMS notifications to maintenance personnel based on the work order rule base;

[0040] Repair and maintenance evaluation unit: It consists of a work order processing unit, a work order evaluation unit, and an evaluation standard library. It conducts statistical analysis based on work order processing records to evaluate the execution of repair and maintenance.

[0041] Preferably, the work order processing unit includes: a processor for executing multiple instructions; a memory for storing multiple instructions; wherein the multiple instructions are stored by the memory and loaded and executed by the processor to perform the above-mentioned HVAC terminal equipment maintenance work order analysis method.

[0042] Preferably, the device should be configured with at least two linkage actuators for triggering the generation of maintenance work orders. Each linkage actuator can execute independently and in parallel to meet the order processing requirements under different workload conditions.

[0043] Preferably, in step three, after the pre-training model adaptation is completed, the adapted inference model is packaged into the automatic dispatching device for deployment and call.

[0044] Preferably, in step 4, the HVAC terminal equipment fault decision tree algorithm process includes the following steps:

[0045] S1: parameter initialization;

[0046] S2: Create a fault node;

[0047] S3: If the fault type codes all belong to the same decision class, the fault node is regarded as a leaf node;

[0048] S4: If the attribute set of the decision table is empty, the fault node is treated as a leaf node;

[0049] S5: Select attribute a with the largest conditional entropy drop in the attribute set of the fault decision table, and set the node label to attribute a;

[0050] S6: Classify the fault type according to attribute a;

[0051] S7: A branch with the condition of attribute a grows out from the fault node;

[0052] S8: Let s be the sample set of attribute a in the fault type set;

[0053] S9: If s is empty, add a leaf node

[0054] S10: Otherwise, add a return node with a fault type set of s and an attribute set of the attribute set of the proposed attribute α, and return to S2;

[0055] S11: Output the final HVAC terminal equipment failure decision tree.

[0056] Preferably, in step 6, the device fault response level is divided into three levels:

[0057] Level 1: Extremely urgent, maintenance work requires a response within 15 minutes;

[0058] Level 2: Emergency, maintenance work needs to be responded to within 1 hour;

[0059] Level 3: Generally, maintenance work needs to be responded to within 4 hours.

[0060] Preferably, in step six, the maintenance service process rule base is set to include fault decision code, fault response level, expected maintenance time, fault escalation time, fault handling team, and fault handling operator.

[0061] Preferably, in step seven, the response mechanism of the linkage mechanism is tested for device robustness and pressure strength under a certain operating load to ensure that the linkage mechanism can work normally within the designed operating load.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] 1. The present invention uses NLP Chinese semantic recognition and analysis technology. Therefore, it can achieve automatic classification of HVAC terminal equipment faults and transmission of dispatch instructions by adding an interface mechanism linkage device without modifying the existing building facility management system work order processing device and its ancillary supporting execution mechanism. The advantages of the automatic dispatching device based on NLP technology compared with the traditional dispatching method are as follows:

[0064] 1) It can significantly reduce the order dispatching workload of customer service staff by more than 90%;

[0065] 2) Reduce the error rate of HVAC terminal equipment maintenance dispatch orders by approximately 10%;

[0066] 3) Improve the repair and maintenance performance of HVAC terminal equipment by approximately 30%;

[0067] 4) Significantly shorten the troubleshooting time of building HVAC terminal equipment by about 50%. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a data flow chart of the automatic work dispatching method of the present invention;

[0069] Figure 2 This is a schematic diagram of the configuration of the automatic work dispatching device of the present invention;

[0070] Figure 3 This is the HVAC terminal equipment fault decision tree described in the present invention;

[0071] Figure 4 This is a flow chart of the algorithm for generating a fault diagnosis decision tree for HVAC terminal equipment according to the present invention. DETAILED DESCRIPTION

[0072] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0073] See also Figure 1-4The present invention provides an embodiment of a method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP, comprising the following steps:

[0074] Step 1: Extract a work order dataset for HVAC terminal equipment repair and maintenance from the source system. Perform Chinese word segmentation on the maintenance text based on the Chinese word segmenter configuration settings. To ensure accurate word segmentation, a stop word library should be pre-configured to exclude meaningless word segmentation results.

[0075] Step 2: extracting target keywords from the original work order text according to the device attributes of the maintenance work order, which may include: at least one of the equipment location keyword and the equipment fault problem type keyword;

[0076] The equipment location keyword may include: at least one of the name and number information of the building, floor, space, and room corresponding to each building maintenance work order in the building maintenance work order set;

[0077] The equipment failure keywords may include: at least one of the name and number of the HVAC terminal equipment, the HVAC terminal node, and the HVAC terminal equipment component;

[0078] Problem type keywords may include: at least one of descriptions of common problems and common words in building maintenance;

[0079] Step 3: Build the fault classifier network, configure the pre-training model parameters and the training and test dataset paths, and then start the pre-training process of the fault classifier. After the pre-training is completed, evaluate the effectiveness of the fault classifier and iteratively adjust the model pre-training process based on the evaluation results.

[0080] Model startup parameter configuration

[0081] parameter Default value Optimization value loss squared_hinge hinge fit_intercept True False intercept_scaling 1 0.33 class_weight None balanced verbose 0 1 max_iter 1000 3000

[0082] Step 4: When configuring the automatic judgment of the fault process node, the automatic dispatching device should create a decision table judgment mechanism for the unit and determine the number of HVAC terminal equipment fault judgment rules. A decision table with n conditions has 2n rules.

[0083] List all condition piles and action piles for HVAC terminal equipment fault judgment;

[0084] Fill in the conditions for judging the failure of HVAC terminal equipment;

[0085] Fill in the action items for HVAC terminal equipment fault processing to obtain the initial decision table;

[0086] Simplify the fault decision table and merge similar rules;

[0087] Using the ID3 algorithm as a benchmark, the calculation formula for the device decision information gain is as follows:

[0088] Information entropy formula for HVAC terminal equipment failure decision-making:

[0089]

[0090] Calculate the air volume, cooling effect, heating effect, and other four formulas respectively:

[0091]

[0092] Information gain formula:

[0093]

[0094] Finally, select the attribute with the highest information gain for splitting;

[0095] Step 5: Set up a dynamic workload assessment unit for dispatching HVAC terminal equipment repair and maintenance work orders. This processing unit's dispatch constraints are based on three factors: job queue length, job quality, and job duration. The unit selection logic assigns maintenance work orders to the candidate operator with the shortest current job queue length, the highest job quality, and the shortest historical job duration.

[0096] The work order queue minimization function f1(x) is as follows: f1(x) = MaxQ

[0097] (1)

[0098] Formula 1: M is the number of work orders in the operator's work order set, and W is the i-th work order in the dispatched set;

[0099] The function f2(x) that maximizes the sum of work order completion satisfaction is as follows: f2(x) = MaxS

[0100] (2)

[0101] Formula 2: M is the number of historical work orders of the candidate operator, W is the i-th work order of the historical work orders, and is the historical completion satisfaction of the operator's historical work orders W;

[0102] The function f3(x) that minimizes the sum of the work order completion time is as follows: f3(x) = MinT

[0103] (3)

[0104] Formula 3: The historical average completion time of the operator to complete the work order W;

[0105] A set of work orders to be dispatched, where the constraints include the simultaneous satisfaction of work status (queue length), historical work quality, and historical completion time constraints.

[0106] Step 6: In the linkage mechanism between the automatic dispatching device and the building facilities management system, set up a maintenance service process rule library based on the fault type according to the standard mapping process;

[0107] Step 7: In the linkage setting between the automatic dispatching device and the building facilities management system, verify the applicability of the work order triggering process according to the standard mapping process for the fault type.

[0108] In step 1, the extracted maintenance work order data set is split into training set and test set in a ratio of 8:2, and the data categories of the training set and test set are labeled.

[0109] In step 1, the automatic dispatching device for HVAC terminal repair and maintenance work includes a Chinese corpus of HVAC terminal equipment repair and maintenance, a stop word library, a fault classifier pre-training model, a fault category decision judgment table, a service process library, a dispatching linkage mechanism, a work order execution mechanism, a work order rule library, and a repair and maintenance evaluation unit;

[0110] The Chinese corpus for HVAC terminal equipment repair and maintenance consists of a repair and maintenance vocabulary and word frequency. This is based on Chinese word segmentation of HVAC terminal equipment maintenance work orders, and then a vocabulary is set based on general maintenance vocabulary and HVAC terminal equipment-specific vocabulary. Automatic classification and recognition of HVAC terminal equipment repair and maintenance requires Chinese semantic recognition based on NLP technology. Therefore, it is necessary to first build a Chinese corpus for HVAC terminal equipment maintenance. To eliminate meaningless maintenance vocabulary that affects corpus judgment, a stop word library should be set to improve the analysis and recognition quality of the corpus.

[0111] Stop word library: It consists of a list of stop words that can exclude unnecessary or meaningless words;

[0112] Fault classifier pre-training model: This model consists of a classifier network model, a pre-training dataset, and a test dataset. Based on the results of HVAC terminal equipment fault classification data annotation, the HVAC terminal equipment fault classifier is pre-trained to generate a prediction model output for semantic feature recognition of HVAC terminal equipment fault repair report text.

[0113] Fault category decision table: This table consists of fault code, fault description, fault type, and work team. Based on the decision table mechanism, the judgment logic for assigning work teams to HVAC terminal equipment is constructed, and service priorities are set.

[0114] Service process library: It consists of fault code, service process code, and service process priority. It matches the work order service process based on the service priority identified by the fault category decision judgment table and the fault type identified by the fault classifier. After the pre-training of the HVAC terminal equipment repair and maintenance fault classifier is completed, a rule library for HVAC terminal equipment maintenance fault classification and HVAC terminal equipment maintenance process is constructed, which should include the mapping relationship between HVAC terminal equipment fault classification and maintenance process, terminal equipment maintenance classification service level setting, terminal equipment maintenance working hour threshold, terminal equipment maintenance event escalation rules, and it is stipulated that after the configuration and debugging of the automatic dispatching device for HVAC terminal equipment repair and maintenance operations is completed, the service process and specifications set in the device should also comply with the requirements of the building HVAC system part of the national civil building mechanical and electrical repair and maintenance specifications;

[0115] HVAC terminal equipment fault judgment decision table

[0116] Code Problem Description Electrical problems Mechanical problems Environmental issues decision making 01-1 Small air volume Y N N Motor damage 01-2 Small air volume N Y N The fan in the wall cabinet and coil unit does not rotate 01-3 Small air volume N N Y The return air outlet of the wall cabinet and coil is blocked by dust 02-1 Poor heating effect Y N N The motor itself is damaged 02-2 Poor heating effect N Y N The fan in the wall cabinet and coil unit does not rotate 02-3 Poor heating effect N N Y The return air outlet of the wall cabinet and coil is blocked by dust 03-1 Poor cooling effect Y N N The motor itself is damaged 03-2 Poor heating effect N Y N The fan in the wall cabinet and coil unit does not rotate 03-3 Poor heating effect N N Y The return air outlet of the wall cabinet and coil is blocked by dust 04-1 No heating Y N N The electric two-way valve is damaged, replace the electric two-way valve 04-2 No heating N N Y The filter is clogged with impurities and dirt; remove the filter and clean it with water 05-1 No refrigeration Y N N The electric two-way valve is damaged, replace the electric two-way valve 05-2 No refrigeration N N Y The filter is clogged with impurities and dirt; remove the filter and clean it with water

[0117] HVAC terminal equipment fault work order processing rule configuration table

[0118]

[0119] Dispatching linkage mechanism: It consists of service process code, work order process code, and trigger instructions. It performs work order allocation according to the dispatching rules set in the service process library. The mapping relationship between the HVAC terminal equipment fault classification and the fault service process of the automatic dispatching device is a configurable decision tree judgment processing unit, which enables the dispatching device to configure the user-side decision table device mechanism according to different business scenarios. The device will use the decision dichotomy method to find the most matching HVAC terminal equipment service process node that meets the fault business scenario based on the configuration result, and activate the building facilities management system work order service workflow engine execution mechanism action under the node. The automatic dispatching device calculates the optimization algorithm of the work load of the work team and the judgment logic for evaluating the complexity of the work task. Therefore, the automatic dispatching device can dynamically evaluate the current workload of each work executor, effectively assign the maintenance work order to the most suitable work executor, and avoid the occurrence of workload imbalance among HVAC terminal equipment maintenance work executors;

[0120] Work order execution mechanism: It consists of a work order workflow engine and a work order message notification module. It triggers the work order workflow engine actions based on the dispatch linkage mechanism message mechanism.

[0121] Work order rule base: composed of service process code and maintenance team code, generates SMS notifications to maintenance personnel based on the work order rule base;

[0122] Repair and maintenance evaluation unit: It consists of a work order processing unit, a work order evaluation unit, and an evaluation standard library. It conducts statistical analysis based on work order processing records to evaluate the execution of repair and maintenance.

[0123] The work order processing unit includes: a processor for executing multiple instructions; a memory for storing multiple instructions; wherein the multiple instructions are stored in the memory and loaded and executed by the processor to implement the above HVAC terminal equipment maintenance work order analysis method.

[0124] The device should be equipped with at least two linkage actuators to trigger the generation of maintenance work orders. Each linkage actuator can execute independently and in parallel to meet the dispatch processing requirements under different workload conditions.

[0125] In step three, after the pre-training model is adapted, the adapted inference model is packaged into the automatic dispatching device for deployment and call.

[0126] In step 4, the HVAC terminal equipment fault decision tree algorithm process includes the following steps:

[0127] S1: parameter initialization;

[0128] S2: Create a fault node;

[0129] S3: If the fault type codes all belong to the same decision class, the fault node is regarded as a leaf node;

[0130] S4: If the attribute set of the decision table is empty, the fault node is treated as a leaf node;

[0131] S5: Select attribute a with the largest conditional entropy drop in the attribute set of the fault decision table, and set the node label to attribute a;

[0132] S6: Classify the fault type according to attribute a;

[0133] S7: A branch with the condition of attribute a grows out from the fault node;

[0134] S8: Let s be the sample set of attribute a in the fault type set;

[0135] S9: If s is empty, add a leaf node

[0136] S10: Otherwise, add a return node with a fault type set of s and an attribute set of the attribute set of the proposed attribute α, and return to S2;

[0137] S11: Output the final HVAC terminal equipment failure decision tree.

[0138] In step 6, the device fault response level is divided into three levels:

[0139] Level 1: Extremely urgent, maintenance work requires a response within 15 minutes;

[0140] Level 2: Emergency, maintenance work needs to be responded to within 1 hour;

[0141] Level 3: Generally, maintenance work needs to be responded to within 4 hours.

[0142] In step six, the maintenance service process rule base is set to include fault decision code, fault response level, estimated maintenance time, fault escalation time, fault handling team, and fault handling operator.

[0143] In step seven, the response mechanism of the linkage mechanism is tested for device robustness and pressure strength under a certain operating load to ensure that the linkage mechanism can work normally within the designed operating load.

[0144] Working principle: Establish a Chinese corpus of HVAC terminal equipment failures in a certain tertiary hospital, and exclude useless and meaningless words, perform data annotation on the maintenance work orders of HVAC terminal equipment failures in a certain tertiary hospital, and put the annotated data set into the fault automatic classifier for pre-training, generate the HVAC terminal equipment failure classifier of the Construction Engineering Hospital, configure the HVAC terminal equipment failure type decision table of a certain tertiary hospital, and configure the work order service type according to the fault type and maintenance professional type, configure the linkage trigger device between the work order service type and the work order execution mechanism, after the above configuration, the device of the present invention can automatically assign HVAC terminal equipment failure work orders according to the user's fault reporting information. Through the device of the present invention, a certain tertiary hospital shortened the repair and maintenance operation time of HVAC terminal equipment by 50%, significantly reduced the workload of customer service personnel in dispatching orders by 95%, and reduced the dispatch error rate by 20%. At the same time, it improved the handling efficiency of operation and maintenance personnel by 15%, and the overall efficiency of its maintenance personnel increased by 8%.

[0145] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0146] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Automatic dispatching method for HVAC terminal equipment maintenance work based on NLP, including: The steps include: Step 1: Extract the work order dataset for HVAC terminal equipment repair and maintenance from the source system and perform Chinese word segmentation on the maintenance text content based on the Chinese word segmentation settings; Step 2: Extract target keywords from the original work order text based on the device attributes of the maintenance work order, where the target keywords include at least one of: a device location keyword and a device fault problem type keyword; The equipment location keyword includes: at least one of the name and number information of the building, floor, space, and room corresponding to each building maintenance work order in the building maintenance work order set; The equipment failure keywords include: at least one of the name and number of the HVAC terminal equipment, the HVAC terminal node and the HVAC terminal equipment component; The problem type keywords include: at least one of common problem descriptions and common words in building maintenance; Step 3: Build the fault classifier network, configure the pre-training model parameters and the training and test dataset paths, and then start the pre-training process of the fault classifier. After the pre-training is completed, evaluate the effectiveness of the fault classifier and iteratively adjust the model pre-training process based on the evaluation results. Step 4: When configuring the automatic judgment of the fault process node, the automatic dispatching device should create a decision table judgment mechanism for the unit and determine the number of HVAC terminal equipment fault judgment rules. A decision table with n conditions has 2n rules. List all condition piles and action piles for HVAC terminal equipment fault judgment; Fill in the conditions for judging the failure of HVAC terminal equipment; Fill in the action items for HVAC terminal equipment fault processing to obtain the initial decision table; Simplify the fault decision table and merge similar rules; Using the ID3 algorithm as a benchmark, the calculation formula for the device decision information gain is as follows: Information entropy formula for HVAC terminal equipment failure decision-making: Calculate the air volume, cooling effect, heating effect, and other four formulas respectively: Information gain formula: Finally, select the attribute with the highest information gain for splitting; In step 4, the HVAC terminal equipment fault decision tree algorithm process includes the following steps: S1: parameter initialization; S2: Create a fault node; S3: If the fault type codes all belong to the same decision class, the fault node is regarded as a leaf node; S4: If the attribute set of the decision table is empty, the fault node is treated as a leaf node; S5: Select attribute a with the largest conditional entropy drop in the attribute set of the fault decision table, and set the node label to attribute a; S6: Classify the fault type according to attribute a; S7: A branch with the condition of attribute a grows out from the fault node; S8: Let s be the sample set of attribute a in the fault type set; S9: If s is empty, add a leaf node; S10: Otherwise, add a return node with a fault type set of s and an attribute set of the attribute set of the proposed attribute α, and return to S2; S11: Output the final HVAC terminal equipment fault decision tree; Step 5: Set up a dynamic workload assessment unit for dispatching HVAC terminal equipment repair and maintenance work orders. The dispatch constraints for this processing unit are the job queue length, job quality, and job duration. The unit selection logic is to assign maintenance work orders to the candidate operator with the shortest current job queue length, the highest job quality, and the shortest historical job duration. Step 6: In the linkage mechanism between the automatic dispatching device and the building facilities management system, set up a maintenance service process rule library based on the fault type according to the standard mapping process; Step 7: In the linkage setting between the automatic dispatching device and the building facilities management system, verify the applicability of the work order triggering process according to the standard mapping process for the fault type.

2. The NLP-based automatic dispatching method for HVAC terminal equipment maintenance work according to claim 1 is characterized by: In the step 1, the extracted maintenance work order data set is divided into a training set and a test set in a ratio of 8:2, and the training set and the test set are labeled with data categories.

3. The method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP according to claim 1 is characterized by: In step 1, the automatic dispatching device for HVAC terminal repair and maintenance work includes a Chinese corpus for HVAC terminal equipment repair and maintenance, a stop word library, a fault classifier pre-training model, a fault category decision judgment table, a service process library, a dispatching linkage mechanism, a work order execution mechanism, a work order rule library, and a repair and maintenance evaluation unit; The Chinese corpus for HVAC terminal equipment repair and maintenance consists of a repair and maintenance vocabulary and word frequency. This corpus is created by segmenting HVAC terminal equipment maintenance work orders into Chinese words, and then using general maintenance vocabulary and HVAC terminal equipment-specific vocabulary to create the vocabulary. Stop word library: It consists of a list of stop words that can exclude unnecessary or meaningless words; Fault classifier pre-training model: This model consists of a classifier network model, a pre-training dataset, and a test dataset. Based on the results of HVAC terminal equipment fault classification data annotation, the HVAC terminal equipment fault classifier is pre-trained to generate a prediction model output for semantic feature recognition of HVAC terminal equipment fault repair report text. Fault category decision table: This table consists of fault code, fault description, fault type, and work team. Based on the decision table mechanism, the judgment logic for assigning work teams to HVAC terminal equipment is constructed, and service priorities are set. Service process library: It consists of fault code, service process code, and service process priority. It matches work order service processes based on the service priority identified by the fault category decision table and the fault type identified by the fault classifier. Dispatch linkage mechanism: It consists of service process code, work order process code, and trigger instructions, and performs work order allocation actions based on the dispatch rules set in the service process library; Work order execution mechanism: It consists of a work order workflow engine and a work order message notification module. It triggers the work order workflow engine actions based on the dispatch linkage mechanism message mechanism. Work order rule base: composed of service process code and maintenance team code, generates SMS notifications to maintenance personnel based on the work order rule base; Repair and maintenance evaluation unit: It consists of a work order processing unit, a work order evaluation unit, and an evaluation standard library. It conducts statistical analysis based on work order processing records to evaluate the execution of repair and maintenance.

4. The NLP-based automatic dispatching method for HVAC terminal equipment maintenance work according to claim 3 is characterized by: The work order processing unit includes: a processor for executing multiple instructions; a memory for storing multiple instructions; wherein the multiple instructions are stored in the memory and loaded and executed by the processor to perform the above-mentioned HVAC terminal equipment maintenance work order analysis method.

5. The method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP according to claim 3 is characterized by: The device should be configured with at least two linkage actuators for triggering the generation of maintenance work orders. Each linkage actuator can execute independently and in parallel to meet the dispatch processing requirements under different workload conditions.

6. The method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP according to claim 1 is characterized by: In the step three, after the pre-training model is adapted, the adapted inference model is packaged into the automatic dispatching device for deployment and calling.

7. The method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP according to claim 1 is characterized by: In step 6, the device fault response level is divided into three levels: Level 1: Extremely urgent, maintenance work requires a response within 15 minutes; Level 2: Emergency, maintenance work needs to be responded to within 1 hour; Level 3: Generally, maintenance work needs to be responded to within 4 hours.

8. The method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP according to claim 1 is characterized by: In step six, the maintenance service process rule base is set to include fault decision code, fault response level, expected maintenance time, fault escalation time, fault handling team, and fault handling operator.

9. The method for automatically dispatching repair and maintenance orders for HVAC terminal equipment based on NLP according to claim 1 is characterized by: In the step seven, the response mechanism of the linkage mechanism is tested for device robustness and pressure strength under a certain operating load.

Images