Intelligent prediction method and system for spare parts of tobacco mechanical equipment

Abstract

The invention discloses a tobacco mechanical equipment spare part intelligent prediction method and system, and belongs to the technical field of tobacco mechanical equipment spare part management, and the method comprises the following steps: obtaining a spare part receiving time sequence data set and a spare part replacement time sequence data set in a selected data collection time interval; selecting a part of data in the spare part demand time sequence data set as a training data set; inputting the training data set into a random forest model to generate a random forest predictor; and adjusting parameters of the random forest model to enable a fitting error of the random forest model to a test data set to be smaller than an error threshold, and obtaining an optimized random forest predictor. The method can adapt to the demand fluctuations of different types of spare parts, and can effectively predict and manage the different types of demand fluctuations.

Description

Technical Field

[0001] The present invention relates to the technical field of tobacco machinery and equipment spare parts management, and more specifically, to a method and system for intelligent prediction of tobacco machinery and equipment spare parts. Background Art

[0002] Tobacco machinery and equipment are important equipment for the automated production of redrying, silk making, processing and packaging processes in cigarette production. Tobacco machinery and equipment have an important impact on the production scale, organizational mode, raw material consumption, energy, etc. of the tobacco industry, and to a certain extent determine the product specifications, production level, variety structure, product quality and economic benefits of cigarette production.

[0003] Tobacco machinery and equipment have a high degree of automation. Once a failure occurs, the shutdown and renovation will face significant losses. Therefore, tobacco machinery and equipment often strengthen the intensity of daily inspection and maintenance, pre-assess the risk of failure, and eliminate the hidden trouble of the maintenance strategy in advance. This requires that tobacco machinery spare parts have sufficient inventory reserves to meet the needs of daily inspection and maintenance of tobacco machinery and equipment. There are many types of tobacco machinery and equipment, and the technology of various equipment is constantly improving, resulting in a wide variety of parts, different specifications, and different adaptation models, resulting in a variety of spare parts inventory categories. For example, there are thousands of spare parts supporting the packaging machinery. In order to ensure the requirements of production, the existing spare parts management strategy has a large amount of redundant inventory in the spare parts inventory, which occupies a large amount of budget funds. Since tobacco machinery and equipment have a certain life cycle, after a device is eliminated, the spare parts inventory serving the equipment will also be eliminated. Therefore, the precise management of tobacco machinery spare parts inventory is not only for ensuring the production efficiency of tobacco machinery, but also for reducing maintenance costs. It is of great significance.

[0004] Parts are an important part of tobacco machinery and equipment. Each part plays a function in tobacco machinery and equipment. All parts are organically combined to form the whole machine of tobacco machinery and equipment. Spare parts are parts prepared in advance to ensure the normal and efficient operation of tobacco machinery and equipment. The more spare parts inventory of vulnerable parts, the higher the frequency of spare parts procurement; on the contrary, the less vulnerable parts have less spare parts inventory, or even zero. Most of the existing tobacco machinery and equipment spare parts inventory have information management systems to track spare parts procurement, warehousing, use, and scrapping processes. The spare parts inventory information management system has accumulated a large amount of historical data. There are many existing technologies that can use historical data to discover rules and predict future situations. For example, moving average method, exponential smoothing method, autoregressive difference moving average method, etc. In addition, nonlinear methods such as neural network method and random forest can also be applied. However, this type of method can only analyze and process the change data of spare parts inventory, but cannot track the reasons for the change of spare parts inventory, and lacks tracking of the actual use, replacement, and scrapping of parts.

[0005] The existing technology also monitors and detects the failures of tobacco machinery and equipment itself, collects the failure and fault data of related parts, and predicts the spare parts inventory demand data to reduce the risk of blindly purchasing spare parts inventory. For example: the invention patent "Cigarette Factory Pipeline Failure Warning Method" with the authorization announcement number CN111503530B discloses the use of digital twin technology to establish a cigarette factory power pipeline system, and locate, detect and predict faults. The system also includes a spare parts management module to support the timely elimination and repair of faults. The invention patent "A Cigarette Equipment Full Life Cycle Monitoring System and Monitoring Method" with the authorization announcement number CN108038626B discloses the establishment of a one-machine-one-file full life cycle digital management system for the entire life cycle of cigarette equipment from procurement, use, maintenance, repair, scrapping, and disposal, and reduces equipment maintenance costs and spare parts inventory by monitoring equipment failure frequency and repair parts data, thereby enhancing the long-term good, stable and efficient operation level of cigarette equipment. In addition, the invention patent "Spare Parts Management Device and Spare Parts Inventory Management Method" with authorization announcement number CN110555639B discloses a spare parts operating condition collection device, a spare parts inventory collection device and a spare parts full life cycle detection device, and adopts digital twin technology to intuitively provide equipment failure warning information and spare parts inventory warning information, thereby optimizing spare parts procurement decisions, improving spare parts inventory management efficiency, and saving costs.

[0006] However, the defects of the above-mentioned prior art are: the method of using spare parts inventory historical data to predict spare parts demand lacks tracking and identification of the causes of parts failure, the specific characteristics of spare parts collection, replacement, and scrapping, and spare parts inventory data alone cannot accurately predict spare parts procurement; on the other hand, increasing the detection of equipment failures requires a large amount of equipment and technical investment. Moreover, different types of tobacco machinery and equipment have different structures and working principles. Even the same model and different generations of products have different failure characteristics. Therefore, the technical path of improving the accuracy of spare parts demand prediction through equipment failure detection, monitoring, and early warning is difficult and not easy to maintain. In addition, in reality, in the process of spare parts collection and replacement, human factors often cause spare parts to be collected but not used, and accounts and materials do not match. It is also very difficult to track the whereabouts of spare parts in real time. Finally, for tobacco machinery and equipment, because the requirements for failure downtime and downtime are close to zero, preventive maintenance and maintenance are adopted in most cases to eliminate potential failures in the budding stage. Therefore, on the one hand, there is a lack of reliable basis for predicting future spare parts demand based on spare parts inventory historical data, and on the other hand, it is impossible to obtain data after the actual failure occurs, which reduces the accuracy of fault warning. Summary of the invention

[0007] An object of the present invention is to provide an intelligent prediction method and system for tobacco machinery and equipment spare parts to solve the aforementioned redundant inventory problem in the inventory management of tobacco machinery and equipment spare parts, as well as the lack of tracking of the actual usage of parts, high failure downtime rate, and inaccurate data caused by manual collection and replacement.

[0008] According to a first aspect of the present invention, there is provided a method for intelligent prediction of spare parts for tobacco machinery and equipment, comprising the following steps:

[0009] Select a specific time interval before the current time as the data collection time interval;

[0010] Acquire a spare parts collection time series data set and a spare parts replacement time series data set within a selected data collection time interval, and generate a spare parts demand time series data set based on the spare parts collection time series data set and the spare parts replacement time series data set;

[0011] Select a part of the data in the spare parts demand time series data set as the training data set;

[0012] Inputting the training data set into a random forest model to generate a random forest predictor;

[0013] Select another part of the spare parts demand time series data set as a test data set;

[0014] Adjusting the parameters of the random forest model so that the fitting error of the random forest model for the test data set is less than an error threshold, and obtaining an optimized random forest predictor;

[0015] Predict the spare parts demand in the period after the current time based on the optimized random forest predictor.

[0016] Optionally, the spare parts collection time series data set includes five basic element data of spare parts collection, and the spare parts replacement time series data set includes five basic element data of spare parts replacement. The spare parts collection time series data set and the spare parts replacement time series data set are related to each other. According to the mechanic maintenance personnel and spare parts material number, the spare parts collection time series data set and the spare parts replacement time series data set are merged to form a spare parts demand time series data set. Each record in the spare parts demand time series data set includes seven data elements: mechanic maintenance personnel, spare parts material number, quantity, collection time, replacement time, machine, and self-service spare parts shelf.

[0017] Optionally, the spare parts demand time series data set is divided into two parts according to the time sequence: the first 80% of the data is used as a training set, and the last 20% of the data is used as a test set.

[0018] Optionally, the random forest model uses multiple decision trees as base learners, and uses bagging technology to perform random sampling with replacement on the original training set to generate multiple different training subsets. Each training subset is used to train a decision tree. Random attribute selection is introduced in the training process of a single decision tree. The training is repeated multiple times to obtain multiple different decision trees, which are combined into a random forest predictor.

[0019] Optionally, the construction process of the random forest model includes the following steps:

[0020] Obtain the data set D of the first 80% of the spare parts demand time interval;

[0021] Using the random selection and replacement sampling technique, sample data D1 is randomly selected from the first 80% data set D of the spare parts demand time interval;

[0022] The randomly selected sample data D1 is input into the weak predictor K1, which uses the classification and regression decision tree (CART);

[0023] Make decision M1: until the weak predictor K1 outputs an evaluation criterion less than the mean square error threshold; repeatedly randomly extract sample data set Dn → input weak predictor Kn → make decision Mn, where n is the number of decision trees in the random forest model;

[0024] A basic random forest model was constructed, in which the selected feature variables included: historical requisition data, historical demand accuracy, mechanics, inventory location, and spare parts type.

[0025] Optionally, the Croston method is used to classify the spare parts demand according to the intermittent demand variation, using two indicators, namely: ADI and CV 2 , where ADI=ni / zi represents the average demand interval in a sequence, ni represents the number of cycles of the i-th spare part material number, zi represents the number of non-zero cycles of the i-th spare part material number, and CV 2 is the square of the coefficient of variation, which is the standard deviation of the non-zero demand periods divided by the mean of the non-zero demand periods.

[0026] Optionally, in the process of classifying the spare parts demand according to the intermittent demand variation using the Croston method:

[0027] The requirements are divided into four categories:

[0028] (1) Smooth demand: meeting the conditions and ;

[0029] (2) Random demand: meet the conditions and ;

[0030] (3) Intermittent demand: meeting the conditions and ;

[0031] (4) Block demand: meet the conditions and .

[0032] Optionally, the mean absolute percentage error (MAPE) is used as a criterion for judging the fitting error. The mean absolute percentage error is defined as the distance between the model predicted value F(t) and the true value A(t) divided by the average value of the true value A(t), that is: ;

[0033] Where N is the total amount of data in the test data set, and t is the subscript of each pair of predicted value and true value.

[0034] Optionally, the process of obtaining an optimized random forest predictor includes the following steps:

[0035] Start the loop and traverse the number of decision trees in the random forest model;

[0036] Start a loop to traverse the maximum depth of the decision tree;

[0037] Start the loop and traverse the minimum number of samples of leaf nodes;

[0038] Execution in the loop: read the 20% data set after the time interval, and calculate the MAPE evaluation index one by one with the predicted value;

[0039] Select the set of model parameters with the smallest MAPE as the optimal random forest model.

[0040] According to a second aspect of the present invention, there is provided a tobacco machinery and equipment spare parts intelligent prediction system, comprising tobacco machinery and equipment, the tobacco machinery and equipment comprising a plurality of machines, the machines being repaired and maintained in sections by a plurality of mechanics, each of which carries a mobile device; the mobile device is connected to a spare parts management system server via wireless communication, the spare parts warehouse of the tobacco machinery and equipment comprises two levels of inventory, namely an integrated spare parts warehouse and a plurality of self-service spare parts shelves, the integrated spare parts warehouse and the self-service spare parts shelves are connected to the spare parts management system server via a computer communication network, the self-service spare parts shelves have a unique self-service spare parts shelf QR code; the spare parts management personnel connects to and logs in to the spare parts management system server via a spare parts management system access terminal, a spare parts management system is installed on the spare parts management system server, and the spare parts management system stores a program for executing the above-mentioned tobacco machinery and equipment spare parts intelligent prediction method.

[0041] The intelligent prediction method and system for tobacco machinery spare parts disclosed in the present invention have the following technical effects:

[0042] Based on the optimized random forest predictor, the spare parts demand for the next period after the current time is predicted to solve the redundant inventory problem in the existing technology, reduce inventory capital occupation through accurate prediction, and optimize the inventory structure;

[0043] The spare parts demand time series data set is divided into two parts in chronological order: the first 80% of the data is used as the training set, and the second 20% of the data is used as the test set. The spare parts demand time series data set is generated based on the spare parts collection time series data set and the spare parts replacement time series data set to track the actual use of parts. By collecting spare parts collection and replacement data, the accuracy of spare parts prediction is improved.

[0044] The Croston method is used to classify the intermittent changes in spare parts demand. Through detailed classification and value analysis of spare parts, high availability and reasonable inventory control of spare parts are ensured. It can adapt to fluctuations in spare parts demand of different types and enhance the flexibility and adaptability of spare parts management.

[0045] Several mobile devices are respectively connected to the spare parts management system server through wireless communication, and the mobile devices scan the QR code of the machine to collect and bind the basic element data of spare parts use. The integrated spare parts warehouse is used to distribute spare parts to the several self-service spare parts shelves. The spare parts management system is accessed through the spare parts management system access terminal, and the automatic data collection is realized to reduce the registration errors caused by human factors, improve the accuracy and reliability of the data, and improve the work efficiency of spare parts management personnel;

[0046] The present invention reduces the failure downtime rate and improves production efficiency through an intelligent prediction system, adapts to demand fluctuations for different types of spare parts, and can effectively predict and manage different types of demand fluctuations; reduces spare parts procurement and inventory waste through accurate prediction, and reduces the data registration workload of mechanics when receiving and replacing spare parts through automated data collection by mobile devices and smart shelves, thereby improving maintenance efficiency.

[0047] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0049] Figure 1 It is an overall flow chart of the intelligent prediction method for tobacco machinery equipment spare parts of the present invention;

[0050] Figure 2 for Figure 1 The flowchart for building a random forest model is shown in the overall flowchart of ;

[0051] Figure 3 for Figure 1 Flowchart for optimizing random forest model parameters in the overall flowchart of ;

[0052] Figure 4 for Figure 1 A schematic diagram of time interval selection in the overall flow chart;

[0053] Figure 5 This is a structural diagram of the intelligent prediction system for spare parts of tobacco machinery and equipment of the present invention;

[0054] Figure 6 This is a schematic diagram of the relationship between tobacco machinery and equipment parts and spare parts inventory management;

[0055] Figure 7 It is a schematic diagram of the data collection process for spare parts replacement;

[0056] Figure 8 This is a schematic diagram of the data collection process for self-service spare parts shelves.

[0057] The following are marked in the figure:

[0058] 100, tobacco machinery and equipment; 101, 102: machine repair personnel; 103, 104: mobile equipment; 105, 106: machine platforms; 107, 108: self-service spare parts shelves; 109, integrated spare parts warehouse; 110, spare parts management personnel; 111, spare parts management system access terminal; 112, spare parts management system server;

[0059] 200, tobacco machinery and equipment; 202, machines; 203, self-service spare parts shelves; 204, parts; 205, spare parts;

[0060] 300. Tobacco machinery and equipment; 301. Machine; 302. Machine QR code; 303. Machine repair personnel; 304. Spare parts; 305. Mobile equipment;

[0061] 400. Self-service spare parts shelf; 401. Mechanic repair personnel; 402. Mobile equipment; 403. QR code of self-service spare parts shelf; 404. Spare parts. DETAILED DESCRIPTION

[0062] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless otherwise specifically stated.

[0063] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0064] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0065] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0066] like Figure 1 As shown, the overall flow chart of a data-driven intelligent prediction method for spare parts of tobacco machinery and equipment of the present invention includes the following steps:

[0067] Step 1: Start 500: The forecasting process is generally divided into three stages;

[0068] The first stage: model training, including the following steps:

[0069] Step 2: Select the time interval 501 of the model training data set: generally select a time interval before the current time, such as 12 months or 24 months;

[0070] Step 3: Generate spare parts demand time series data set 502: In the spare parts intelligent prediction system of the present invention, two time series data sets are collected, namely, a spare parts collection time series data set containing five spare parts collection basic element data, and a spare parts replacement time series data set containing five spare parts replacement basic element data; the two time series data sets are interrelated, that is, corresponding to the same mechanic, after collecting a spare part with a spare part material number, the spare part should be replaced within a time interval, so that the two time series data sets can be merged according to the mechanic and the spare part material number to form The spare parts demand time series data set is formed, and each record includes seven data elements (mechanic, spare parts material number, quantity, time of receipt, time of replacement, machine, and self-service spare parts shelf). In the actual process, there may be a situation where the parts are not replaced after being received. The way to deal with it is that if the corresponding replacement data cannot be found according to the mechanic and spare parts material number, the replacement time and machine data elements in the record in the spare parts demand time series data set are marked as invalid. In addition, in the actual process, there may be a situation where the receipt record is missing and there is only a replacement record. The way to deal with it is to remove the record.

[0071] Step 4: Select the first 80% of the data set in the time interval as the training data set 503: The data used for training the model is the first 80% of the data set in the time interval, rather than randomly selecting 80% of the data set, so that after training a basic model parameter, the later 20% of the data set is used to optimize the model parameter; dividing the data in chronological order can ensure that the data in the training set is ahead of the validation set and the test set in time, and can also avoid data leakage problems. By using the later 20% of the data set to optimize the model parameters, it can ensure that the model has better generalization ability when facing unknown data.

[0072] Step 5: Input random forest model 504: Random forest model is an algorithm that combines the ideas of bagging and random subspace. It is an integrated predictor model formed by multiple decision trees. The random forest algorithm uses multiple decision trees as base learners and uses bagging to generate different training sets. Random attribute selection is introduced in the training process of a single tree. After multiple trainings, multiple different decision trees are obtained, and these trees are combined into a random forest predictor. By combining the prediction results of multiple decision trees, random forest can reduce the error of a single model and improve the accuracy of the overall prediction.

[0073] The second stage: model optimization, including the following steps:

[0074] Step 6: Use the last 20% of the data set in the time interval as the test data set 505: The 20% data set in the time interval closer to the current time contains more accurate current information and can be better used to optimize the random forest model;

[0075] Step 7: Optimizing the parameters of the random forest model 506: Continuously adjust the parameters of the random forest model so that the model has the smallest fitting error for 20% of the test data set; preferably, the mean absolute percentage error (MAPE) is selected as the judgment indicator of the fitting error; the definition of the mean absolute percentage error is the distance between the model predicted value F(t) and the true value A(t) divided by the average value of the true value A(t), that is: ; Where N is the total amount of data in the test data set, and t is the subscript of each pair of predicted value-true value;

[0076] Phase 3: Spare Parts Forecasting:

[0077] Step 8: Predicting spare parts demand in the next period 507: Using the optimized random forest predictor to predict spare parts demand in the next period after the current time;

[0078] Step 9: End 508: The overall process ends.

[0079] The intelligent spare parts prediction method of the present invention can significantly improve the prediction accuracy of spare parts demand by introducing and optimizing the random forest model. Through accurate spare parts demand prediction, enterprises can make spare parts procurement plans in advance to avoid excessive or insufficient inventory, which helps to reduce the operating costs of enterprises and improve production efficiency.

[0080] like Figure 2 As shown, the flowchart of constructing a random forest model in the overall process of the spare parts intelligent prediction method of the present invention includes the following steps:

[0081] Step 1: The first 80% data set of the spare parts demand time interval D 600: is the spare parts demand time series data set after the spare parts collection time series data set and the spare parts replacement time series data set are combined, represented by the letter D;

[0082] Step 2: Randomly select set D1 601: Use the random selection and replacement sampling technique to randomly select sample data from the first 80% data set D of the spare parts demand time interval;

[0083] Step 3: Input weak predictor K1 604: weak predictor K1 adopts classification and regression decision tree (CART), and initial parameters include: maximum depth of decision tree, evaluation criteria, minimum number of samples for splitting nodes when classifying, minimum number of samples for leaf nodes, minimum standard for the sum of weights of all leaf nodes, judgment threshold for error reduction after splitting, judgment threshold for whether a node is split, etc.; preferably, the initial parameters are set as follows: maximum depth of decision tree is 10, mean square error is selected as evaluation criteria, minimum number of samples for splitting nodes when classifying is 2, minimum number of samples for leaf nodes is 1, minimum standard for the sum of weights of all leaf nodes is 0, judgment threshold for error reduction after splitting is 0, and judgment threshold for whether a node is split is 0.0000001;

[0084] Step 4: Make decision M1 607: until the weak predictor K1 outputs the evaluation criterion less than the mean square error threshold;

[0085] Similarly, repeat steps 2 to 4: randomly extract set D2 602 → input weak predictor K2 605 → make decision M2 608;

[0086] Similarly, repeat steps 2 to 4;

[0087] Repeat steps 2 to 4: randomly extract set Dn 603 → input weak predictor Kn 606 → make decision Mn 609, where n is the number of decision trees in the random forest model. As a preferred method, this parameter is initially set to 200;

[0088] Step 5: Constructing a basic random forest model 610: Select five characteristic variables, including historical collection data, historical demand accuracy, mechanics, inventory location, and spare parts type; the historical collection data refers to the spare parts collection time series data; the historical demand accuracy is calculated by the time interval between the collection time and the replacement time. For each mechanic, there are differences in risk habits and maintenance experience. Since tobacco machinery and equipment are mostly preventive maintenance, that is, routine maintenance and troubleshooting are performed to prevent the occurrence of failures before a failure occurs, experienced mechanics will have a higher prediction accuracy for failures. Spare parts will be replaced immediately after collection, so the time interval will be shorter, and the historical demand accuracy will be higher; the inventory location refers to the location of the self-service spare parts shelf to determine the prediction results of different inventory locations;

[0089] Finally, the spare parts types are classified according to the intermittent demand changes of spare parts demand using the Croston method, using two indicators, namely: ADI and CV 2 , where ADI=ni / zi represents the average demand interval in a sequence, ni represents the number of cycles of the i-th spare part material number, and zi represents the number of non-zero cycles of the i-th spare part material number. Therefore, the ADI indicator mainly reflects the intermittent degree of the occurrence of non-zero elements in the sequence, that is, the frequency of the occurrence of zero elements; and CV 2 It is the square of the coefficient of variation, that is, the standard deviation of the non-zero demand period divided by the average value of the non-zero demand period. As a preference, it is divided into four categories, namely: (1) Smooth demand: ADI≤1.32 and CV 2 ≤0.49, demand is relatively stable, and there are fewer zero-demand periods; (2) Random demand: ADI≤1.32 and CV 2 >0.49, the demand is unstable, the variability is large, and the demand occurs frequently; (3) Intermittent demand: ADI>1.32 and CV 2 ≤0.49, the demand is irregular and sporadic, and the demand is relatively stable; (4) Block demand: ADI>1.32 and CV 2 >0.49, the demand pattern appears randomly, with no demand in a large period of time and large demand differences in different periods, accompanied by a large number of zero demand stages. According to the intermittent changes in spare parts demand, different inventory management and procurement strategies can be adopted for different categories of spare parts. Using the Croston method for spare parts classification and forecasting can help companies better cope with uncertainties in the supply chain. By optimizing spare parts inventory management, companies can reduce the risk of production line stagnation caused by spare parts shortages, thereby enhancing the resilience and stability of the supply chain.

[0090] like Figure 3As shown, the flowchart of optimizing the random forest model parameters in the overall process of the spare parts intelligent prediction method of the present invention includes the following steps:

[0091] Step 1: Start the loop and traverse the number of decision trees in the random forest model, which is 700;

[0092] Step 2: Start the loop and traverse the decision tree to a maximum depth of 701;

[0093] Step 3: Start the loop and traverse the minimum number of samples of leaf nodes, which is 702;

[0094] Step 4: Execute in the loop: read the 20% data set after the time interval 703, and calculate the MAPE evaluation index 704 one by one with the predicted value;

[0095] Step 5: Select the set of model parameters 705 with the smallest MAPE as the optimal random forest model.

[0096] By traversing different parameter combinations, the embodiment of the present invention can find the parameter settings that make the random forest model perform best on the test set. By optimizing the random forest model parameters, the reliability and availability of the prediction results can be improved, thereby helping decision makers make more informed decisions.

[0097] like Figure 4 As shown, a schematic diagram of time interval selection in the overall process of the spare parts intelligent prediction method of the present invention is as follows:

[0098] A period of time before the current time includes a first 80% data set 800 and a second 20% data set 801, the first 80% data set 800 is used as a training data set, the second 20% data set 801 is used as a test data set, and the spare parts demand for the next period is predicted 802 at the current time;

[0099] In the next period, the selected time interval is shifted backward, including the first 80% data set 803 of the next period and the last 20% data set 804 of the next period. The first 80% data set 803 of the next period is used as the training data set, and the last 20% data set 804 of the next period is used as the test data set to predict the spare parts demand in the next period 805. By dividing the time series data into training data sets and test data sets, and using the rolling time window method for prediction, it can be ensured that the model can continuously learn the latest data trends and patterns. By continuously updating or retraining the prediction model, it can be better adapted to these changes, thereby improving the prediction ability and robustness of the model.

[0100] The present invention proposes an embodiment of an intelligent prediction system for spare parts of tobacco machinery and equipment. Specifically, Figure 5As shown, the following structures are included: tobacco machinery equipment 100, several mechanics 101, 102, several mobile equipment 103, 104, several machines 105, 106, several self-service spare parts shelves 107, 108, an integrated spare parts warehouse 109, several spare parts management personnel 110, a spare parts management system access terminal 111, and a spare parts management system server 112.

[0101] The tobacco machinery equipment 100 includes a first machine 105, a second machine 106 and several other machines, which are respectively repaired and maintained by a first mechanic 101, a second mechanic 102 and several other mechanics in sections. The first mechanic 101 and the second mechanic 102 carry a first mobile device 103 and a second mobile device 104 respectively. The first mobile device 103 and the second mobile device 104 are connected to the spare parts management system server 112 through wireless communication.

[0102] The spare parts warehouse for tobacco machinery and equipment is divided into two levels of inventory, namely: an integrated spare parts warehouse 109, and several self-service spare parts shelves 107, 108; the integrated spare parts warehouse 109 and several self-service spare parts shelves 107, 108 are interconnected with the spare parts management system server 112 through a computer communication network and exchange data; among them, several self-service spare parts shelves 107, 108 are deployed near the required machines 105, 106 according to the repair and maintenance needs of the tobacco machinery and equipment 100, so as to facilitate the use and replacement of mechanics 101, 102; among them, the integrated spare parts warehouse 109 is managed by several spare parts management personnel 110 who are responsible for spare parts procurement, warehousing, coding, archiving, storage, transfer and other operations. The spare parts management personnel 110 use the spare parts management system access terminal 111 to connect and log in to the spare parts management system server 112 to implement the above operations; the self-service spare parts shelves 107, 108 are replenished through the integrated spare parts warehouse 109.

[0103] The spare parts intelligent prediction system is used to collect spare parts collection and replacement data. The sources of these data collection include: the data generated by mechanics 101 and 102 in the process of replacing spare parts, and transmitted to the spare parts management system server in real time 103 and 104 through the mobile devices they carry with them; the data generated by mechanics 101 and 102 in the process of collecting spare parts, the spare parts collection data is automatically recorded by the self-service spare parts shelves 107 and 108, and is bound to the mechanics 101 and 102 who collected the spare parts, and the self-service spare parts shelves 107 and 108.

[0104] Due to the large variety of spare parts, complex specifications, and different replacement frequencies, the mechanics 101 and 102 often have data updates that are not timely and accurate during the process of receiving and replacing spare parts; for example, forgetting to register the spare parts after receiving them, forgetting to register the spare parts after replacing them, etc. For this reason, as a preferred embodiment, the present invention equips each mechanic 101 and 102 with a mobile device 103 and 104; when replacing parts with spare parts, the application installed on the mobile device 103 and 104 scans the QR code of the machine 105 and 106 to realize the collection and binding of the five basic element data (mechanic, machine, spare part material number, quantity, and replacement time) when replacing spare parts.

[0105] As a preferred embodiment, the self-service spare parts shelves 107 and 108 in the spare parts intelligent prediction system of the present invention are intelligent shelves with automatic data collection, which realize the collection and binding of five basic elements of spare parts collection (mechanic, self-service spare parts shelf, spare parts material number, quantity, collection time) after completing the spare parts collection. For example: Jinggong intelligent 6-row 4-column tool cabinet is preferred, which has its own intelligent control system, has face recognition, card recognition personnel authentication management, spare parts storage in different areas, self-service collection data automatic upload and other functions, which greatly improves the spare parts collection efficiency and collection data collection rate.

[0106] like Figure 6 Explain the relationship between tobacco machinery and equipment parts and spare parts inventory management. Tobacco machinery and equipment 200 is assembled from various parts 204. During use, parts 204 may experience wear, aging, failure, and other problems, which may cause tobacco machinery and equipment 200 to fail, and in serious cases, cause shutdown. Each machine 202 is monitored by a full-time mechanic 201 to monitor the working condition of the machine 202. If a failure is found, the machine will be shut down and repaired. In order to ensure the efficient and reliable operation of tobacco machinery and equipment 200, regular maintenance, preventive maintenance, and early replacement of parts are generally adopted to prevent failures. For this purpose, the mechanic 201 collects spare parts 205 from the self-service spare parts shelf 203 and replaces parts 204 in time to ensure the normal operation of the machine 202. It can be seen that when installed on the tobacco machinery and equipment 200, it is called a component 204, and when it is on the self-service spare parts shelf 203, it becomes a spare part 205, that is, when it is not replaced, it is called a spare part 205, and after it is replaced, it is called a component 204; the mechanic 201 analyzes and determines the component 204 with potential failure risks based on his own experience and the operating status monitoring data of the tobacco machinery and equipment 200, and collects the same type of spare parts 205 from the self-service spare parts shelf 203. The establishment of the spare parts management system makes the procurement, storage and collection of spare parts more standardized and efficient, avoiding the cost waste caused by insufficient or excessive spare parts. At the same time, by timely replacing and repairing parts, the service life of the equipment can be extended and the maintenance cost can be further reduced.

[0107] like Figure 7 As shown, a schematic diagram of the spare parts replacement data collection process is shown. Each machine 301 of the tobacco machinery equipment 300 is provided with a unique machine QR code 302. Before replacing the spare parts, the machine repair personnel 303 need to scan the machine QR code 302 with a portable mobile device 305, and bind the spare parts 304 that have been used in the spare parts management system. After completing the replacement and other replacement operations, the five spare parts replacement basic element data (machine repair personnel, machine, spare parts material number, quantity, replacement time) are collected and bound. The spare parts replacement data recorded in the embodiment of the present invention has completeness and accuracy, and can realize the full traceability of the spare parts replacement process. When the equipment fails or has quality problems, the cause of the problem and the person responsible can be quickly located by querying the spare parts replacement record. And it can count the inventory status and consumption trend of spare parts in real time, providing a scientific basis for spare parts procurement and inventory management.

[0108] like Figure 8 As shown, a schematic diagram of the self-service spare parts shelf collection process, each self-service spare parts shelf 400 has a unique self-service spare parts shelf QR code 403, before the mechanic 401 collects the spare parts 404, he first uses the mobile device 402 he carries with him to scan the self-service spare parts shelf QR code 403, enters the data of the spare parts 404, opens the self-service spare parts shelf 400, and takes out the spare parts 404. After the collection operation is completed, the five basic elements of spare parts collection (mechanic, self-service spare parts shelf, spare parts material number, quantity, collection time) are collected and bound.

[0109] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings, but the above is only a preferred embodiment of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched into a variety of equivalent schemes by those skilled in the art without departing from or changing the design ideas and technical effects of the present invention; therefore, the present invention is not limited to the scope of implementation shown in the drawings, and all changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the specification and drawings, should be within the protection scope of the present invention.

Claims

1. A method for intelligent prediction of spare parts for tobacco machinery and equipment, characterized in that: The following steps are involved: Select a specific time interval before the current time as the data collection time interval; Acquire a spare parts collection time series data set and a spare parts replacement time series data set within a selected data collection time interval, and generate a spare parts demand time series data set based on the spare parts collection time series data set and the spare parts replacement time series data set; Select a part of the data in the spare parts demand time series data set as the training data set; Inputting the training data set into a random forest model to generate a random forest predictor; Select another part of the spare parts demand time series data set as a test data set; Adjusting the parameters of the random forest model so that the fitting error of the random forest model for the test data set is less than an error threshold, and obtaining an optimized random forest predictor; Predict the spare parts demand in the period after the current time based on the optimized random forest predictor.

2. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 1 is characterized in that: The spare parts collection time series data set includes five basic element data of spare parts collection, and the spare parts replacement time series data set includes five basic element data of spare parts replacement. The spare parts collection time series data set and the spare parts replacement time series data set are related to each other. According to the mechanic maintenance personnel and spare parts material number, the spare parts collection time series data set and the spare parts replacement time series data set are merged to form a spare parts demand time series data set. Each record in the spare parts demand time series data set includes seven data elements: mechanic maintenance personnel, spare parts material number, quantity, collection time, replacement time, machine, and self-service spare parts shelf.

3. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 1 is characterized in that: The spare parts demand time series data set is divided into two parts according to the chronological order: the first 80% of the data is used as the training set, and the last 20% of the data is used as the test set.

4. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 1 is characterized in that: The random forest model uses multiple decision trees as base learners and uses bagging technology to perform random sampling with replacement on the original training set to generate multiple different training subsets. Each training subset is used to train a decision tree. Random attribute selection is introduced in the training process of a single decision tree. After multiple trainings, multiple different decision trees are obtained and combined into a random forest predictor.

5. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 4 is characterized in that: The construction process of the random forest model includes the following steps: Obtain the data set D of the first 80% of the spare parts demand time interval; Using the random selection and replacement sampling technique, sample data D1 is randomly selected from the first 80% data set D of the spare parts demand time interval; The randomly selected sample data D1 is input into the weak predictor K1, which uses the classification and regression decision tree (CART); Make decision M1: until the weak predictor K1 outputs an evaluation criterion less than the mean square error threshold; repeatedly randomly extract sample data set Dn → input weak predictor Kn → make decision Mn, where n is the number of decision trees in the random forest model; A basic random forest model was constructed, in which the selected feature variables included: historical requisition data, historical demand accuracy, mechanics, inventory location, and spare parts type.

6. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 1, characterized in that: The Croston method is used to classify the intermittent demand changes of spare parts demand, using two indicators, namely: ADI and CV 2 , where ADI=ni / zi represents the average demand interval in a sequence, ni represents the number of cycles of the i-th spare part material number, zi represents the number of non-zero cycles of the i-th spare part material number, and CV 2 is the square of the coefficient of variation, which is the standard deviation of the non-zero demand periods divided by the mean of the non-zero demand periods.

7. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 6 is characterized in that: In the process of classifying the intermittent demand changes of spare parts demand using Croston method: The requirements are divided into four categories: (1) Smooth demand: meeting the conditions and ; (2) Random demand: meet the conditions and ; (3) Intermittent demand: meeting the conditions and ; (4) Block demand: meet the conditions and .

8. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 1 is characterized in that: The mean absolute percentage error (MAPE) is used as the judgment indicator of the fitting error. The mean absolute percentage error is defined as the distance between the model predicted value F(t) and the true value A(t) divided by the average value of the true value A(t), that is: ； Where N is the total amount of data in the test data set, and t is the subscript of each pair of predicted value and true value.

9. The intelligent prediction method for tobacco machinery and equipment spare parts according to claim 1, characterized in that: The process of obtaining an optimized random forest predictor consists of the following steps: Start the loop and traverse the number of decision trees in the random forest model; Start a loop to traverse the maximum depth of the decision tree; Start the loop and traverse the minimum number of samples of leaf nodes; Execution in the loop: read the 20% data set after the time interval, and calculate the MAPE evaluation index one by one with the predicted value; Select the set of model parameters with the smallest MAPE as the optimal random forest model.

10. An intelligent prediction system for tobacco machinery and equipment spare parts, characterized in that: It includes tobacco machinery and equipment, and the tobacco machinery and equipment includes several machines. The machines are repaired and maintained in sections by several mechanics, and the several mechanics carry mobile devices respectively; the mobile devices are connected to the spare parts management system server through wireless communication, and the spare parts warehouse of the tobacco machinery and equipment includes two levels of inventory, namely an integrated spare parts warehouse and several self-service spare parts shelves, the integrated spare parts warehouse and the self-service spare parts shelves are connected to the spare parts management system server through a computer communication network, and the self-service spare parts shelves have a unique self-service spare parts shelf QR code; the spare parts management personnel connect to and log in to the spare parts management system server through the spare parts management system access terminal, and the spare parts management system server is installed with a spare parts management system, and the spare parts management system stores a program for executing the tobacco machinery and equipment spare parts intelligent prediction method described in any one of claims 1 to 9.

Images