Cartridge component damage prediction system

Abstract

To provide a cartridge component damage prediction system capable of accurately predicting the component damage of a cartridge.SOLUTION: The data storage unit 10 of the component damage prediction system 1 stores the number of times of regeneration, the transport information, and the component damage occurrence case as the learning data 12, and the processing unit 50 reads the learning data 12 from the data storage unit 10, performs machine learning of the read learning data 12, and generates the component damage prediction model 52 having the number of times of regeneration of the cartridge 100 and / or the transport information of the cartridge 100 as the input data 54 and the component damage time prediction information of the cartridge 100 and / or the component damage content of the cartridge 100 as the output data 56.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a parts damage prediction system for predicting damage to parts of cartridges filled with ion exchange resins or the like used in pure water production systems or the like. [Background technology]

[0002] Cartridges filled with ion exchange resins are often used in pure water production facilities that produce pure water. Patent Document 1 discloses a pure water production facility that includes a plurality of cartridges filled with ion exchange resins. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 4-66185 Summary of the Invention [Problem to be solved by the invention]

[0004] Cartridges filled with ion exchange resin are transported between users and regeneration plants, for example, during the process of regenerating the ion exchange resin. Cartridges include components such as a cylinder (i.e., a resin cylinder), a tank head, and a water collection pipe. These components are subject to damage due to aging, deterioration due to the number of regenerations, and the accumulation of various factors, such as vibrations during transportation. Until now, no system has been proposed that can accurately predict cartridge component damage. Therefore, an object of the present invention is to provide a cartridge component damage prediction system that can accurately predict cartridge component damage. [Means for solving the problem]

[0005] The part damage prediction system of the present invention is a part damage prediction system that predicts part damage of a cartridge that has a plastic tube, a water collection pipe, and a tank head and can be used repeatedly by recycling the filling material filled in the plastic tube, and is equipped with a data storage unit and a processing unit, and the data storage unit stores at least the number of times the cartridge has been recycled, transportation information for the cartridge, and examples of part damage occurrences for the cartridge as learning data, and the processing unit reads the learning data from the data storage unit, performs machine learning on the read learning data, and generates a part damage prediction model that uses the number of times the cartridge has been recycled and / or the transportation information for the cartridge as input data and predicts information on the time of part damage of the cartridge and / or the details of part damage of the cartridge as output data. [Effects of the Invention]

[0006] According to the present invention, it is possible to provide a cartridge part damage prediction system that can accurately predict damage to cartridge parts. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a cross-sectional view of the cartridge. [Figure 2] FIG. 2 is a functional block diagram of a part damage prediction system according to an embodiment of the present invention. [Figure 3] FIG. 3 is a flow diagram showing the flow of generating a part damage prediction model in the part damage prediction system according to the embodiment of the present invention. [Figure 4] FIG. 4 is a flow chart showing the flow of part damage prediction in the part damage prediction system according to the embodiment of the present invention. [Figure 5] FIG. 5 is a table showing an example of learning data in the part damage prediction system according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0008] An embodiment of a part failure prediction system 1 of the present invention will be described with reference to the drawings. The part failure prediction system 1 is a system that predicts part failure in a reusable cartridge. An example of a reusable cartridge is a cartridge filled with ion exchange resin used in pure water production equipment.

[0009] (cartridge) FIG. 1 shows an example of a cartridge. FIG. 1 is a cross-sectional view of a cartridge 100. As shown in FIG. 1, the cartridge 100 includes a cylinder, i.e., a resin cylinder 102, a water collection pipe 104, and a tank head 106. Region 110 in FIG. 1 illustrates the interior of the resin cylinder 102 for illustrative purposes. As shown in region 110, the interior of the resin cylinder 102 is filled with ion exchange resin 112. Arrows in FIG. 1 indicate the flow of water. As indicated by arrow 131, feed water 151 flows into the cartridge 100 from the outside of the cartridge 100 via the tank head 106. As indicated by arrow 133, the feed water 151 flows through the resin cylinder 102 from the upper tank portion 121 to the lower tank portion 122. During this flow, ions contained in the feed water 151 are ion-exchanged by the cation exchange resin and anion exchange resin contained in the ion exchange resin 112. This ion exchange converts the feed water 151 into pure water 152. After ion exchange, the feed water 151, i.e., the pure water 152, flows through the water collection pipe 104 from the tank lower part 122 to the tank upper part 121, as shown by arrow 135. The pure water 152 flows out of the cartridge 100 from the tank head 106, as shown by arrow 137.

[0010] 1 can be regenerated by regenerating a filler such as ion exchange resin 112 filled in a resin cylinder 102. The component damage prediction system 1 is a system that predicts component damage in such a regenerative cartridge 100. In the following description, the filler is assumed to be ion exchange resin 112.

[0011] (Cartridge Remanufacturing) In regenerating the cartridge 100, the ion exchange resin 112 in the used cartridge 100 is replaced with regenerated ion exchange resin 112. The replacement of the ion exchange resin 112 is usually carried out at a regeneration plant. Regeneration refers to restoring the ion exchange capacity of ion exchange resin 112 that has lost its ion exchange capacity. At the regeneration plant, work such as repacking the used ion exchange resin 112 that has lost its capacity with regenerated ion exchange resin 112 is carried out.

[0012] Furthermore, when recycling the cartridge 100, the resin cylinder 102, the water collection pipe 104, and the like are reused. Therefore, during the transportation process when the cartridge 100 is sent to and / or returned from a recycling plant, parts such as the resin cylinder 102 and / or the water collection pipe 104 may be damaged due to vibrations from the transportation vehicle, etc. If the resin cylinder 102 and / or the water collection pipe 104, etc. are damaged, water leakage and / or leakage of the ion exchange resin to the secondary side may occur. In addition to the above-mentioned transportation-related causes, damage to parts may also be caused by deterioration over time due to years of use and / or deterioration due to the number of recycling attempts.

[0013] (Parts damage prediction system) The component damage prediction system 1 of this embodiment is a system that predicts damage to components included in the cartridge 100 that is being regenerated as described above. Examples of components included in the cartridge 100 include a resin cylinder 102, a water collection pipe 104, and a tank head 106.

[0014] The part damage prediction system 1 of this embodiment will be described with reference to Fig. 2. Fig. 2 is a functional block diagram of the part damage prediction system 1. As shown in Fig. 2, the part damage prediction system 1 includes a data storage unit 10, a processing unit 50, a data input unit 82, and a data output unit 84.

[0015] (Data storage unit) The data storage unit 10 is a part that stores the learning data 12. The learning data 12 is data used for machine learning when generating the part damage prediction model 52. The learning data 12 includes the number of times the cartridge 100 has been recycled, transportation information about the cartridge 100, and examples of part damage occurring in the cartridge 100.

[0016] (Example of cartridge part damage) Examples of damage to parts of the cartridge 100 include cracks occurring in the resin cylinder 102 and the tank head 106, for example.

[0017] (Processing section) The processing unit 50 is a part that generates a part failure prediction model 52 and stores the generated part failure prediction model 52. The processing unit 50 includes a model generation unit 60 and a storage unit 62.

[0018] (Model generation part) The model generation unit 60 is a part that generates the part damage prediction model 52. The model generation unit 60 reads out the learning data 12 from the data storage unit 10, and performs machine learning on the read out learning data 12 to generate the part damage prediction model 52. The part damage prediction model 52 is a prediction model that receives as input data 54 the number of times the cartridge 100 has been reused and / or transportation information about the cartridge 100, and receives as output data 56 prediction information about the time when the cartridge 100 is damaged and / or details of the damage to the part of the cartridge.

[0019] The generated part damage prediction model 52 is stored in the storage unit 62.

[0020] When component damage prediction model 52 is used to predict component damage, input data 54 is input from data input unit 82 to processing unit 50. Processing unit 50 inputs input data 54 to component damage prediction model 52. Output data 56 output from component damage prediction model 52 is then output from data output unit 84.

[0021] The above-described component damage prediction system 1 can provide a component damage prediction system that can accurately predict component damage in the cartridge 100. The location where component damage occurs and the nature of the component damage are affected by vibrations and shaking applied to the cartridge 100 during transportation. In the component damage prediction system 1 of this embodiment, the learning data 12 includes transportation information about the cartridge 100. Therefore, the component damage prediction system 1 of this embodiment can generate a component damage prediction model 52 that can accurately predict component damage. Ultimately, the component damage prediction model 52 can be used to accurately predict component damage in the cartridge 100.

[0022] (Transportation Information) The transportation information can broadly include information relating to the transportation of the cartridge 100. The transportation information can include, for example, at least one of the transportation means of the cartridge 100, the transportation route of the cartridge 100, and the packaging condition of the cartridge 100 during transportation. By including the above-mentioned information in the transportation information, factors in the transportation condition that are particularly likely to affect part damage can be included in the transportation information. As a result, part damage can be predicted more accurately.

[0023] Specifically, the transportation means may include types of transportation means such as vehicles such as trucks, railways, airplanes, ships, etc. Furthermore, the transportation route may include the transportation distance, road traffic information and road construction information in the case of land transportation, wave information in the case of sea transportation, and air current information in the case of air transportation.

[0024] The transport information may also include information about the weather in general, the number of sudden braking of the vehicle if the transport is by land, etc.

[0025] Furthermore, location information related to transportation may be obtained from a global positioning system (GPS). The data included in the training data 12 is not limited to the data exemplified above. Training data 12 may also include various data that may be related to part failure, such as departure and arrival locations, transportation information divided into outbound and inbound journeys, the time of transportation, and the time required for transportation.

[0026] (Exception data) It is preferable that exceptional data be removed in advance from the training data 12. Exceptional data is exceptional data, such as data corresponding to a unique event, that is undesirable for use in machine learning when generating the part breakage prediction model 52. Examples of exceptional data include data about cartridges 100 that have human factors that affect the timing of part breakage in the cartridges 100, and data about cartridges 100 that have manufacturing defects and therefore are more likely to experience part breakage than cartridges 100 that do not have the defects.

[0027] By removing outlier data from the training data in advance, more accurate prediction of part damage becomes possible.

[0028] (output data) The part failure time prediction information for cartridge 100 in output data 56 output by part failure prediction model 52 may include the rate at which part failure is likely to occur in cartridge 100 and / or the remaining number of times that cartridge 100 will be reused until the number of times that part failure will occur in cartridge 100.

[0029] By including the above-mentioned information in the component failure time prediction information for the cartridge 100, it becomes easier to create a maintenance plan including component replacement, to suggest maintenance to the user, and to carry out component replacement in advance.

[0030] (output data) An example of output data will be described. The probability that part damage will occur in the cartridge 100, i.e., the damage risk, can be displayed as a percentage. In this case, it is also possible to determine in advance what percentage of damage is required to replace the target part. For example, assume that it is determined that the part will be replaced when the damage risk reaches 70%. In this case, if the output damage risk is 30%, it can be determined that the target part does not yet need to be replaced.

[0031] Furthermore, the remaining number of refurbishments until damage can be used to determine the timing of part replacement, for example, as follows. First, a graph is created that shows the correlation between the number of refurbishments and the risk of damage. If the graph is extrapolated, an estimate is made of how many more refurbishments are required to reach the risk of damage that would be a guideline for replacement. The risk of damage that would be a guideline for replacement is assumed to be 70%, for example. In this case, if the current risk of damage is 55%, extrapolating the risk of damage of 55% would result in a risk of damage of 70% after 35 more refurbishments. In this example, the time when the number of refurbishments will reach 35 can be estimated from past refurbishment frequencies, and the part can be replaced before that time.

[0032] If a factory uses multiple pieces of pure water production equipment, the replacement time can be predicted for each piece of pure water production equipment. If the replacement times are relatively close, the pieces of equipment whose replacement times are close can be replaced together.

[0033] (Generation of part damage prediction model) The flow of part damage prediction using the part damage prediction system 1 will be described with reference to Figures 3 to 5. First, the flow of generating the part damage prediction model 52 will be described with reference to Figure 3. Figure 3 is a flow diagram showing the flow of generating the part damage prediction model 52 in the part damage prediction system 1 according to an embodiment of the present invention. In Figure 3 and the following description, S1 indicates step 1. The same applies to the other steps. (S1) S1 is a step of collecting learning data, such as the number of times a cartridge has been recycled, cartridge transportation information, and instances of cartridge part damage.

[0034] (Learning data) An example of the learning data will be described with reference to Fig. 5. Fig. 5 is a table showing an example of the learning data 12 in the part damage prediction system 1. In the example shown in Fig. 5, the learning data 12 includes, as data at the time of part damage, transportation information for the outbound and inbound journeys for each target part. In addition to the data shown in Fig. 5, the learning data 12 may also include specific details of the part damage as one example of the part damage occurrence.

[0035] (S2) S2 is a step in which preprocessing is performed on the learning data 12. In S2, exceptional data is removed from the collected learning data. S2 is a step in which data to be used for learning is carefully selected in order to improve the prediction accuracy of the part damage prediction model 52. The data to be removed can be the exceptional data described above.

[0036] (S3) S3 is a step in which machine learning is performed on the preprocessed learning data to generate the part damage prediction model 52. With this, the flow for generating the part damage prediction model 52 ends.

[0037] (Parts damage prediction) Next, the flow of part damage prediction will be described with reference to Fig. 4. Fig. 4 is a flow chart showing the flow of part damage prediction in the part damage prediction system 1 according to the embodiment of the present invention.

[0038] (S11) S11 is a step of acquiring model input data, in which data to be input to the part damage prediction model 52, such as the number of times the cartridge has been recycled and cartridge transportation information, is acquired.

[0039] (S12) S12 is a step in which preprocessing of the model input data acquired in S11 is performed. In S12, exceptional data is removed from the acquired model input data. The data to be removed can be the exceptional data described above.

[0040] (S13) S13 is a step in which the model input data preprocessed in S12 is input to the part damage prediction model 52.

[0041] (S14) S14 is a step of obtaining an output from the part damage prediction model 52. The part damage prediction model 52 outputs part damage time prediction information and part damage details based on the data input in S13.

[0042] (S15) S15 is a step in which it is determined whether the content output in S14 indicates that an action such as part replacement is necessary. If the determination in S15 is NO, the flow ends. On the other hand, if the determination in S15 is YES, the flow proceeds to S16.

[0043] (S16) S16 is a step in which, based on the content of the output from part damage prediction model 52 in S14, a maintenance plan is created, the maintenance plan is proposed to the user, and parts are replaced in advance. In S16, an alarm may be issued from part damage prediction system 1 as necessary. This flow ends when a response based on the output from part damage prediction model 52 is taken in S16. As described above, the part damage prediction model 52 of this embodiment makes it possible to appropriately predict cartridge part failures and replace parts at appropriate times. You can do it.

[0044] Although the present invention has been described above with reference to the preferred embodiment, it is to be understood that the present invention is not limited to the preferred embodiment described above and that various modifications, variations, and combinations are possible.

[0045] (Variation) For example, in the above description, the part damage prediction model 52 was generated using the training data 12 from which exceptional data had been removed. The method for generating the part damage prediction model 52 is not limited to this. For example, the part damage prediction model 52 may be generated using training data 12 from which exceptional data has not been removed. This makes it possible to generate a part damage prediction model 52 that can predict part damage including exceptional damage occurrence cases that correspond to the exceptional data. In this case, when predicting part damage using the generated part damage prediction model 52, it is not necessary to remove the exceptional data from the input data 54. In other words, in this modified example, the preprocessing of the training data, which is S2 shown in FIG. 2, and the preprocessing of the model input data, which is S12 shown in FIG. 4, can be omitted.

[0046] [Contribution to the United Nations-led Sustainable Development Goals (SDGs)] This disclosure includes matters that contribute to achieving Goal 6 of the SDGs (Sustainable Development Goals), "Clean water and sanitation," and Goal 9, "Industry, innovation and infrastructure."

[0047] <1> A part damage prediction system for predicting part damage of a cartridge that includes a resin cylinder, a water collection pipe, and a tank head and that can be regenerated by regenerating a filler filled in the resin cylinder, A data storage unit and a processing unit are provided, the data storage unit stores at least the number of times the cartridge has been recycled, transportation information of the cartridge, and instances of damage to parts of the cartridge as learning data; The processing unit reading the learning data from the data storage unit, and performing machine learning on the read learning data; The number of times the cartridge has been used and / or the transportation information of the cartridge are used as input data; A part failure prediction system that generates a part failure prediction model that uses, as output data, prediction information on when the cartridge will fail and / or details of the part failure of the cartridge. <2> In the above-described part damage prediction system, the transportation information includes at least one of a transportation means, a transportation route, and a packaging state of the cartridge during transportation. <3> In the above-described part damage prediction system, the learning data stored in the data storage unit is Data about the cartridge having human factors that affect the timing of component failure of the cartridge; and A part damage prediction system, wherein data relating to cartridges that have manufacturing defects and therefore are more likely to experience part damage than cartridges that do not have the defects is removed from the data. <4> In the above-mentioned part failure prediction system, the part failure time prediction information for the cartridge output by the part failure prediction model includes the rate at which part failure will occur in the cartridge and / or the remaining number of times the cartridge can be reused until part failure will occur in the cartridge. [Explanation of symbols]

[0048] 1. Parts damage prediction system 10 Data storage unit 12 Training data 50 Processing section 52 Parts damage prediction model 54 Input Data 56 Output Data 60 Model Generation Unit 62 Memory section 82 Data Entry Section 84 Data output section 100 cartridges 102 Resin Cylinder 104 Water collection pipe 106 Tank Head 110 areas 112 Ion exchange resin 121 Top of the tank 122 Lower part of tank

Claims

1. A part damage prediction system for predicting part damage of a cartridge that includes a resin cylinder, a water collection pipe, and a tank head and that can be regenerated by regenerating a filler filled in the resin cylinder, A data storage unit and a processing unit are provided, the data storage unit stores at least the number of times the cartridge has been recycled, transportation information of the cartridge, and instances of damage to parts of the cartridge as learning data; The processing unit reading the learning data from the data storage unit, and performing machine learning on the read learning data; The number of times the cartridge has been reproduced and / or the transportation information of the cartridge are used as input data; A part failure prediction system that generates a part failure prediction model that uses, as output data, information predicting when part failure occurs in the cartridge and / or details of part failure in the cartridge.

2. 2. The part damage prediction system according to claim 1, wherein the transportation information includes at least one of a transportation means, a transportation route, and a packaging condition of the cartridge during transportation.

3. The learning data stored in the data storage unit is Data about the cartridge having human factors that affect the timing of component failure of the cartridge; and 3. The part damage prediction system according to claim 1, wherein data on cartridges that have manufacturing defects and are therefore more likely to experience part damage than cartridges that do not have the defects is removed from the data.

4. 3. A part failure prediction system as described in claim 1 or 2, wherein the part failure time prediction information for the cartridge output by the part failure prediction model includes the rate at which part failure is likely to occur in the cartridge and / or the remaining number of times the cartridge can be reused until part failure is likely to occur in the cartridge.

Images