Failure prediction system

Abstract

To provide a technique capable of promptly and accurately detecting temperature control failure caused by various factors by using not individual calculation processing but integrated calculation processing.SOLUTION: A failure prediction system includes: an environment state management section 42 that generates an environment state value group of a temperature control accommodation device; an attribute state management section 41 that generates an attribute state value group of the temperature control accommodation device; a feature state management section 43 that generates a feature state value group on the basis of feature state data on temperature control; a device state calculation model 5 learned by using the environment state value group and the attribute state value group during a normal operation as a learning input parameter and the feature state value group during the normal operation as a learning output parameter; a difference level calculation section 51 that calculates a difference level between a calculated feature state value group output by inputting a current actual environmental state value group and a current actual attribute state value group to the learned device state calculation model 5 and an actual feature state value group; and a failure estimation section 52 that calculates failure possibility of the temperature control accommodation device on the basis of the difference level.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to a failure prediction system for a temperature-controlled storage device having an item storage chamber whose temperature is controlled by a temperature control device. [Background technology]

[0002] A temperature-controlled storage device comprises an item storage chamber and a temperature control device that supplies cold or hot air to the item storage chamber. The temperature control device includes a refrigerator, a heater, an air flow path, a valve, and a heat exchanger such as an evaporator. Examples of such temperature-controlled storage devices include refrigerated showcases and freezer showcases installed in stores such as supermarkets and convenience stores to store products in a refrigerated or frozen state. If a malfunction occurs in such a temperature-controlled storage device and it is no longer possible to control the temperature (refrigeration, freezing, or heating), the stored products will be damaged. For this reason, there is a demand for rapid detection of malfunctions in temperature-controlled storage devices.

[0003] The control device for a freezer showcase or a refrigerated showcase according to Patent Document 1 is equipped with a judgment unit that judges whether cooling is being performed normally or not based on the operation judgment result of a temperature control valve by an operation judgment unit that judges the operation of the temperature control valve, or that judges whether cooling is being performed normally or not based on the measurement result of a temperature sensor that measures the temperature inside the showcase. If the judgment unit judges that cooling is not being performed normally, recovery processing is carried out.

[0004] The control device in Patent Document 2 includes a refrigerant leakage determination unit that compares the compressor discharge temperature, based on the detection result of the discharge temperature sensor, with a predetermined reference discharge temperature, and compares the suction temperature, based on the detection result of the suction temperature sensor, with a predetermined reference suction temperature, and determines that there is a sign of refrigerant leakage from the refrigerant circuit if the discharge temperature rises relative to the reference discharge temperature and the rate of change is equal to or greater than a predetermined discharge temperature threshold, and if the suction temperature rises relative to the reference suction temperature and the rate of change is equal to or greater than a predetermined suction temperature threshold.If the refrigerant leakage determination unit determines that there is a sign of refrigerant leakage, a predetermined warning operation is executed. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2019-027714 [Patent Document 2] Japanese Patent Application Publication No. 2019-100569 Summary of the Invention [Problem to be solved by the invention]

[0006] The control device of Patent Document 1 determines poor cooling based on a malfunction of the temperature control valve, while the control device of Patent Document 2 determines refrigerant leakage based on the discharge temperature and suction temperature of the compressor. However, temperature control failures in temperature-controlled storage devices such as freezer showcases and refrigerated showcases can occur due to a variety of factors other than a malfunction of the temperature control valve or a refrigerant leakage. Detecting poor temperature control based on the individual detected values ​​of each factor requires multiple algorithms to detect poor temperature control, which creates the problem of a heavy computational load.

[0007] In view of the above circumstances, an object of the present invention is to provide a technology for quickly and accurately detecting temperature control failures caused by various factors by integrated arithmetic processing rather than individual arithmetic processing. [Means for solving the problem]

[0008] In order to achieve the above object, the failure prediction system of the present invention for a temperature-controlled storage device having an item storage chamber whose temperature is controlled by a temperature control device includes an environment state management unit that generates an environment state value group based on environment state data of the temperature-controlled storage device, an attribute state management unit that generates an attribute state value group based on attribute state data related to operation attributes of the temperature-controlled storage device, a feature state management unit that generates a feature state value group based on feature state data related to the temperature control of the temperature-controlled storage device, and a learning input parameter that uses the environment state value group and the attribute state value group during normal operation of the temperature-controlled storage device, and The system includes a learned device state calculation model using the group of feature state values ​​during operation as learning output parameters, a difference calculation unit that calculates the difference between a calculated feature state value group output by inputting a group of actual environment state values ​​generated based on the environmental state data of the current temperature regulating storage device and a group of actual attribute state values ​​generated based on the attribute state data of the current temperature regulating storage device into the learned device state calculation model, and an actual feature state value group generated based on the feature state data of the current temperature regulating storage device, and a failure estimation unit that calculates the possibility of failure of the temperature regulating storage device based on the difference.

[0009] In the present invention, important characteristic state data related to the temperature control of a temperature-controlled storage device are defined as a set of dependent variables, and factors (factors) affecting the characteristic state are defined as a set of explanatory variables, including environmental state data for the temperature-controlled storage device and attribute state data related to the operating attributes of the temperature-controlled storage device. Furthermore, a machine learning technique is used to construct a device state calculation model as a mapping that derives the set of dependent variables from the set of explanatory variables. During learning, the environmental state of the temperature-controlled storage device during normal operation, the attribute state during normal operation, and the characteristic state (correct value) during normal operation are used as training data. This enables the device state calculation model to calculate a characteristic state within the range of an optimum set (normal operating set). If a malfunction occurs in the temperature-controlled storage device, the output actual characteristic state will be outside the optimum set. Therefore, by comparing the actual characteristic state with a calculated actual characteristic state (included in the optimum set) calculated by inputting the actual environmental state and actual attribute state based on raw data of the current (operating) temperature-controlled storage device, the malfunction can be detected quickly and accurately through integrated calculation processing.

[0010] Note that failures of the temperature-controlled storage device include defects in the item storage chamber, failures of the temperature-controlled equipment, etc. Furthermore, raw data such as sensor detection values ​​and device setting values ​​are difficult to handle in the device state calculation model, so they are converted into a form that is easier for the device state calculation model to handle, for example, vectorization. Here, a group of environment state values ​​is generated based on the environment state data, a group of attribute state values ​​is generated based on the attribute state data, and a group of feature state values ​​is generated based on the feature state data. Of course, depending on the type of state data, they may be used as state value groups as they are.

[0011] Refrigerated storage devices (refrigerated showcases and freezer showcases) are widely installed in supermarkets, drugstores, convenience stores, and other locations. If they malfunction, the contents stored therein will be damaged, making it important to detect signs of malfunction. Such refrigerated storage devices operate with various specifications based on the type and capacity of the stored contents. However, it is preferable to adopt attributes common to refrigerated storage devices of various specifications as the operating attributes input into the device state calculation model. Such common attributes are preferably determined experimentally and empirically. For this reason, the present invention proposes that the temperature-controlled storage device be a refrigerated storage device, and that the operating attributes include model information (refrigerated case specifications, freezer identifiers such as freezer number), set temperature, cooling mode / defrost mode, etc.

[0012] The operating status of the temperature-controlled storage device changes depending on the temperature of the location where it is installed, as well as the season and time of day, i.e., the environmental status of the temperature-controlled storage device. It is difficult to obtain data on such environmental status using sensors attached to the temperature-controlled storage device, and it is preferable to obtain it from an external source. For this reason, the present invention proposes that the temperature-controlled storage device be provided with an environmental status acquisition unit that acquires the environmental status of the temperature-controlled storage device from an external server and converts it into environmental status data.

[0013] Some data indicating the environmental condition can be easily obtained from sensors located in or around the cooling storage device, so it is also preferable to use such data. For this reason, the present invention proposes that the temperature-controlled storage device is a cooling storage device, and that the environmental condition includes the temperature of the installation location of the cooling storage device, the humidity of the installation location, and date and time information.

[0014] The characteristic states that are important for the cooling performance of the refrigerated storage device and that are easy to obtain are determined through trial and error during the construction stage of the device state calculation model. That is, in the present invention, the temperature regulated storage device is a refrigerated storage device, and the characteristic state data includes the cooling operation time of the temperature regulated device per unit time, the temperature of the cooling media supplied to the item storage chamber, and the defrosting time.

[0015] The temperature control of a temperature-controlled storage device changes from moment to moment. Therefore, in order to more accurately predict temperature control failures, it is preferable to treat each of the environmental state, attribute state, and characteristic state as time-series data. Such time-series data is preferably vectorized, taking into account trends, periodicity, seasonality, etc., of the time-series data, and input to the device state calculation model. For this reason, the present invention proposes that the environmental state data, attribute state data, and characteristic state data include time-series data.

[0016] It is preferable to manage the multiple refrigerated storage devices (refrigerated showcases and freezer showcases) installed in each store by store, by region, or even by the entire region, rather than managing them individually.For this reason, the present invention proposes that the temperature regulated storage devices are multiple refrigerated storage devices installed in a store, and that a terminal control unit installed in the store is provided, and that the terminal control unit is connected to the multiple refrigerated storage devices by data communication lines and to a cloud server via a data communication line, and that the terminal control unit is provided with an operating attribute acquisition unit that acquires the attribute state data, an environmental state acquisition unit that acquires the environmental state data, and a feature state acquisition unit that acquires the feature state data, and that the environmental state management unit, the attribute state management unit, the feature state management unit, the device state calculation model, the dissimilarity calculation unit, and the fault estimation unit are constructed in the cloud server.

[0017] In a preferred embodiment of the networked failure prediction system of the present invention, the cloud server is provided with a learning management unit that, when the failure estimation calculated by the failure estimation unit is incorrect, re-learns the equipment state calculation model using the group of environment state values ​​related to the failure estimation error, the group of attribute state values ​​related to the failure estimation error, and the group of feature state values ​​related to the failure estimation error. With this configuration, the equipment state calculation model is updated based on the latest results, thereby improving the accuracy of failure prediction.

[0018] In the event of a serious malfunction, a large change occurs in the characteristic status data or the detection signal on which the characteristic status data is based. Such a large change must be reported immediately to the store where the device is installed, the management company, etc., and a corrective measure must be taken, without even needing to estimate the malfunction through the device status calculation model. For this reason, the present invention proposes providing an emergency abnormality notification unit that issues an emergency abnormality notification based on the characteristic status data or the detection signal from which the characteristic status data is based, and a preset threshold range for determining an abnormality in the storage compartment. [Brief explanation of the drawings]

[0019] [Figure 1] 1 is a schematic diagram of the overall configuration of a failure prediction system. [Figure 2] FIG. 1 is a cross-sectional view showing an example of the structure of a freezer / refrigerator showcase. [Figure 3] FIG. 2 is a functional block diagram showing a schematic diagram of the flow of information during learning of the device state calculation model. [Figure 4] This is a functional block diagram that schematically shows the flow of information in the failure prediction system when the learned device state model is put into practice. DETAILED DESCRIPTION OF THE INVENTION

[0020] An example of a failure prediction system of the present invention will be described using Figure 1. In this failure prediction system, a plurality of freezer / refrigerator showcases 1 (hereinafter simply referred to as showcases 1) installed in a store are managed via a network. The showcase 1 is an example of a temperature-controlled storage device, and is composed of an item storage room 13 and a temperature control device that supplies cold air to the item storage room 13.

[0021] A refrigerant is supplied to the multiple showcases 1 by a single refrigerator (a type of temperature control device) 2. Of course, the showcase 1 may have its own built-in refrigerator 2. Operation information for each showcase 1 and each refrigerator 2 is centrally managed by a terminal control unit 3 connected to the showcases 1 and the refrigerators 2 via data communication lines. Operation information (status data) collected by the terminal control unit 3 of each store is sent to a cloud server 4 via a data communication line such as the Internet. The cloud server 4 estimates the possibility of a malfunction in the showcases 1 and the refrigerators 2. The cloud server 4 generates malfunction prediction information including the possibility of malfunction and sends it to a service center SC that manages the showcases 1 and the refrigerators 2. If the service center SC detects signs of a malfunction in the showcases 1 or the refrigerators 2 from the malfunction prediction information, it reports the malfunction to the relevant store and dispatches a service technician if necessary.

[0022] A refrigerant circulation path (a type of temperature control device) 20 is formed between the refrigerator 2 and each showcase 1. The showcase 1 and the refrigerator 2 are provided with a measuring device group 9 consisting of sensors, setting switches, etc., and detection signals from the measuring device group 9 are sent to the terminal control unit 3 as operation information.

[0023] Next, the structure of the showcase 1 will be described with reference to FIG. 2. The showcase 1 includes an outer case 11 with an insulated structure and an open front, and an inner case 12 also open at the front. The interior space of the inner case 12 forms a temperature-regulated storage compartment 13. Shelves 14 mounted on multiple tiers of arm-like protruding brackets are arranged in the storage compartment 13. This showcase 1 is what is called an open-type case. Although not specifically shown, the showcase 1 also includes a panel for inputting the set temperature of the storage compartment 13 and displaying the current temperature of the storage compartment 13, as well as a controller for controlling the temperature based on the set temperature. The terminal control unit 3 may be provided with a function for setting the set temperature of the storage compartment 13, so that the temperature management of multiple showcases 1 can be managed collectively by the terminal control unit 3.

[0024] An air passage (a type of temperature control device) 15 is formed between the outer case 11 and the inner case 12. An evaporator (a type of temperature control device) 16 and a blower (a type of temperature control device) 17 are arranged in the air passage 15, and cool air (a secondary cooling medium) flows through the air passage 15. Furthermore, although not specifically shown, a heater (defrost heater) that heats and removes frost on the evaporator 16 is also arranged in an appropriate position. A cool air outlet 15A that blows cool air downward is formed at the upper end of the air passage 15, and a cool air inlet 15B is formed at the lower end of the air passage 15, with the space between the cool air outlet 15A and the cool air inlet 15B functioning as an air curtain. Furthermore, some of the blown cool air flows into the item storage chamber 13, thereby keeping the stored items cool. A case temperature sensor 91 belonging to the measuring device group 9 is disposed at the cool air outlet 15A, and the temperature inside the item storage room 13 is calculated based on the cool air outlet temperature, which is the detected value of the case temperature sensor 91.

[0025] A refrigerant (primary cooling medium) from the refrigerator 2 flows through a refrigerant circulation path 20 to an evaporator (a type of temperature control device) 16. A first solenoid valve (a type of temperature control device) 21A is provided in the refrigerant circulation path 20 upstream of the evaporator 16, and a second solenoid valve (a type of temperature control device) 21B is provided in the refrigerant circulation path (a type of temperature control device) 20 downstream of the evaporator 16. Furthermore, an expansion valve 22 for decompressing the liquid refrigerant is provided between the first solenoid valve 21A and the evaporator 16. The first solenoid valve 21A and the second solenoid valve 21B are controlled to open and close by the controller described above based on the detected temperature. The temperature of the showcase 1 is controlled by controlling the opening and closing of the first solenoid valve 21A and the second solenoid valve 21B, and therefore the first solenoid valve 21A and the second solenoid valve 21B are collectively referred to as temperature-controlling solenoid valves 21. When the temperature of the showcase 1 is to be lowered, the solenoid valve 21 is controlled to be open, and when the temperature of the showcase 1 is to be raised, the solenoid valve 21 is controlled to be closed. Furthermore, the defrosting operation control for defrosting the evaporator 16 at set time intervals by controlling the heater described above is also performed by the above controller, and the time setting used at that time may be configured to be set by the above panel or by the terminal control unit 3.

[0026] As shown in Fig. 3, the terminal control unit 3 is a computer system with data communication capabilities, and various functional sections are constructed using software, which is an application program, and hardware, which is an electronic circuit. The terminal control unit 3 receives detection values ​​(detection signals) from a group of measuring instruments 9 provided in the showcase 1 and the freezer 2, converts them into a format suitable for the functional sections constructed in the cloud server 4, and sends them to the cloud server 4. The terminal control unit 3 includes an operation attribute acquisition section 31, an environmental state acquisition section 32, and a characteristic state acquisition section 33. Note that environmental states such as the outside temperature may be obtained from an external server, and in such a case, at least some of the functions of the environmental state acquisition section 32 are provided in the cloud server 4.

[0027] The operation attribute acquisition unit 31 receives detection values ​​indicating the operation attribute states from an operation attribute detector 9A (one type of measuring device group 9) that detects attribute states related to the operation attributes of the showcase 1 and the freezer 2, converts them into data, and sends them as attribute state data to the cloud server 4. The attribute state data includes the operation mode of the showcase 1, i.e., data indicating whether it is cooling operation or defrosting operation, the set temperature of the showcase 1, a showcase ID indicating the type of the showcase 1, a freezer ID indicating the type of the freezer 2, etc. The attribute state data is data indicating the operation attributes of the showcase 1 and the freezer 2.

[0028] The environmental condition acquisition unit 32 receives detected values ​​and information indicating the environmental condition from an environmental condition detector 9B (one type of measuring device group 9) that detects the environmental condition in which the showcase 1 and the freezer 2 are installed, converts the detected values ​​and information into data, and sends the data as environmental condition data to the cloud server 4. The environmental condition data includes the temperature and humidity inside the store, the time, and the date. As described above, when environmental conditions such as date and time information and weather information are received from an external server via the Internet or the like, the environmental condition detector 9B serves as an information acquisition device that receives the environmental conditions such as date and time information and weather information from the external server.

[0029] The characteristic state acquisition unit 33 receives detected values ​​and information indicating the characteristic state from a characteristic state detector 9C (a type of measuring instrument group 9) that detects characteristic states related to factors important for determining whether there is a malfunction in the temperature regulation control of the showcase 1 or the freezer 2, converts the detected values ​​and information into data, and sends the data as characteristic state data to the cloud server 4. The characteristic state data includes the solenoid valve opening / closing ratio (corresponding to the cooling operation time), the blowing temperature, the defrosting time, etc., which are the states of the solenoid valve 21.

[0030] The terminal control unit 3 further includes an emergency abnormality notification unit 34. This failure prediction system is designed to predict failures in the showcase 1 or the freezer 2, but if a serious failure occurs in the showcase 1 or the freezer 2 or if a serious setting error occurs, the signal value of a specific measuring device in the measuring device group 9 will change significantly. In such cases, it is necessary to notify the store or service center SC of this before predicting the failure. The emergency abnormality notification unit 34 has the function of detecting an abnormal situation and issuing an emergency abnormality notification if the detection value from the measuring device group 9 (raw detection signal of characteristic state data) exceeds the item storage room abnormality judgment threshold range.

[0031] In this embodiment, the failure prediction of the showcase 1 and the freezer 2 installed in the store described above is substantially performed by the cloud server 4 based on various status data sent from the terminal control unit 3. The cloud server 4 is also a computer system with data communication capabilities, and various functional units are constructed by software, which is an application program, and hardware, which is an electronic circuit.

[0032] Fig. 3 shows the flow of data during the learning process of the equipment state calculation model 5, and Fig. 4 shows the flow of data when the learned equipment state calculation model 5 is put into practice. The cloud server 4 is provided with an attribute state management unit 41, an environment state management unit 42, a feature state management unit 43, an equipment state calculation model 5, a dissimilarity calculation unit 51, and a fault estimation unit 52. However, since there is no problem if the cloud server 4 during learning in Fig. 3 does not include the dissimilarity calculation unit 51 and the fault estimation unit 52, the dissimilarity calculation unit 51 and the fault estimation unit 52 are not shown in Fig. 3.

[0033] The attribute state management unit 41 generates a group of attribute state values ​​based on the attribute state data sent from the driving attribute acquisition unit 31. The environment state management unit 42 generates a group of environment state values ​​based on the environment state data sent from the environment state acquisition unit 32. The feature state management unit 43 generates a group of feature state values ​​based on the feature state data sent from the feature state acquisition unit 33.

[0034] The equipment state calculation model 5 is configured as a machine learning model. Here, the machine learning model is constructed as a mapping that derives the dependent variable group from the explanatory variable group, with the characteristic states of the showcase 1 and the freezer 2 set as the dependent variable group and the environmental state and attribute state of the showcase 1 and the freezer 2 set as the explanatory variable group.

[0035] The learning process of the device state calculation model 5 in FIG. 3 will be described.

[0036] An attribute state value group generated based on attribute state data during normal operation of the showcase 1 and the freezer 2 is input to the equipment state calculation model 5 as one of the learning input parameters (a collection shown by X0 in FIG. 3 and can be expressed as a vector), and an environment state value group generated based on environmental state data at the same time is input to the equipment state calculation model 5 as another of the learning input parameters (a collection shown by Y0 in FIG. 3 and can be expressed as a vector). Furthermore, a feature state value group generated based on feature state data at the same time is given to the model formula (shown by G() in FIGS. 3 and 4) of the equipment state calculation model 5 as a learning output parameter (a collection shown by Z0 in FIG. 3 and can be expressed as a vector), which is the correct answer value. As a result, the equipment state calculation model 5 calculates coefficients constituting a mapping that derives a feature state value group or feature state data expressing the feature state value group from the attribute state value group and the environment state value group. By performing this type of learning a predetermined number of times using attribute state data, environmental state data, and feature state data during normal operation at different points in time, a model formula G (X0 and Y0 are input to derive Z0) whose output has converged to a predetermined error, i.e., model parameters, is determined. In this embodiment, the equipment state calculation model 5 is constructed as a type of multiple regression model. This learning process is managed by the learning management unit 5A. The generated model formula is put into practice by the implementation management unit 5B and used for actual failure prediction.

[0037] The learning process of the equipment state calculation model 5 can be performed by a separate computer system. In this case, the attribute state data, environmental state data, and feature state data output from the terminal control unit 3 are given to this computer system to learn the equipment state calculation model 5. In this case, the learned equipment state calculation model 5 is moved (installed) to the cloud server 4 after the learning is completed.

[0038] The device state calculation model 5 that has completed learning is put into practice by the practice management unit 5 B. The process of putting the device state calculation model 5 into practice in FIG.

[0039] When the showcase 1 or the freezer 2 is operating, detection values ​​indicating the operating attributes (attribute states), environmental states, and characteristic states during operation are output from the operation attribute detector 9A, environmental state detector 9B, and characteristic state detector 9C to the terminal control unit 3. Accordingly, the operation attribute acquisition unit 31 generates attribute state data and provides it to the attribute state management unit 41. The environmental state acquisition unit 32 generates environmental state data and provides it to the environmental state management unit 42. The environmental state acquisition unit 32 generates characteristic state data and provides it to the characteristic state management unit 43.

[0040] Next, the attribute state management unit 41 generates a group of actual attribute state values ​​(a group indicated by Xr in FIG. 4 and capable of being represented by a vector) as a group of attribute state values ​​at the current time, and provides this to the device state calculation model 5, and the environment state management unit 42 generates a group of actual environment state values ​​(a group indicated by Yr in FIG. 4 and capable of being represented by a vector) as a group of environment state values ​​at the current time, and provides this to the device state calculation model 5. The feature state management unit 43 generates a group of actual feature state values ​​(a group indicated by Zr in FIG. 4 and capable of being represented by a vector) as a group of feature state values ​​at the current time, and stores this in a predetermined memory.

[0041] The device state calculation model 5 uses the real attribute state value group Xr and the real environment state value group Yr as input parameters, and calculates and outputs the calculated feature state value group (a collection shown as Zg in Figure 4, which can be expressed as a vector) as output parameters. Such a practice process is managed by the practice management unit 5B.

[0042] The output calculated feature state value group Zg is given to a dissimilarity calculation unit 51, which calculates the dissimilarity (shown as d in FIG. 4) between the calculated feature state value group Zg and the actual feature state value group Zr previously stored in memory. The dissimilarity calculation unit 51 is provided with a dissimilarity calculation algorithm (shown as D() in FIG. 4), which outputs the dissimilarity d between the given calculated feature state value group Zg and the actual feature state value group Zr read out from memory. The dissimilarity calculation algorithm D() may be Euclidean distance, Mahalanobis distance, cosine similarity, or the like.

[0043] Furthermore, the failure estimation unit 52 uses a failure estimation algorithm (denoted by F() in FIG. 4) to output a failure probability (denoted by Fp in FIG. 4) from the given dissimilarity. The failure estimation algorithm can be configured simply by threshold judgment, but more complex judgment methods may also be used. Alternatively, the failure probability may be determined based on the trend using the dissimilarity over time (dissimilarity calculation history) output from the failure estimation unit 52. In this case, the failure estimation unit 52 may be configured to display only the calculated data so that the final judgment is made manually. The output failure probability, together with the original data group (including detected values), is created as failure prediction information and sent to the service center SC. If the failure prediction information indicates a high failure probability, i.e., a failure sign, for a specific showcase 1 or freezer 2, the service center SC will issue a failure prediction report to the relevant store, and in some cases, a service technician will be dispatched.

[0044] When a failure symptom is found, a failure prediction report is sent to the relevant store, and an inspection reveals that the failure symptom was a false report. In this case, the failure estimation calculated by the failure estimation unit 52 is incorrect, and the device state calculation model 5, which is the primary cause of the error, is re-learned. This re-learning process is performed by the learning management unit 5A. In the re-learning process, the real attribute state value group Xr, the real environment state value group Yr, and the real feature state value group Zr involved in the erroneous estimation (failure estimation error) are read, and the device state calculation model 5 is re-learned using the real attribute state value group Xr and the real environment state value group Yr as re-learning input parameters and the real feature state value group Zr as the re-learning correct value. In other words, the device state calculation model 5 is re-learned so that the difference between the calculated feature state value group Zg output by inputting the real attribute state value group Xr and the real environment state value group Yr, which are the re-learning input parameters, and the real feature state value group Zr is small, i.e., so that the device state calculation model 5 is deemed to have no failure symptom.

[0045] The refrigerated showcase and the freezer showcase are not limited to the open type case exemplified in this embodiment, but may be any other type. - Closed type case with a front that can be opened and closed, - Flat case with open top, Dual case with a flat case and a closed case on top, A natural convection type case in which the storage chamber 13 is surrounded by a cooling plate to keep the stored items cool; - Non-defrost type case with multiple evaporators 16, A corner case having an L-shaped storage compartment 13 in a plan view, The present invention can also be applied to these showcases. For example, supermarkets and drugstores often have numerous showcases of different types (with different specifications), and service centers SC often monitor contracted stores across a wide area, such as the entire nation. Since failures of temperature-controlled storage devices such as showcases must be determined based on multiple factors, relying on human eyes for this monitoring can lead to variations due to individual differences and differences in experience, which can lead to significant losses for the store. However, the present invention, by using the device status calculation model 5, enables consistent, accurate, and rapid failure prediction.

[0046] Furthermore, the present invention can provide a failure prediction system that allows for comprehensive failure management by systematically predicting abnormal conditions (similar to a "pre-illness" state in medicine) that may lead to a failure in the future, allowing the service center SC to carefully assess the situation, while quickly notifying the store of any obvious failures.

[0047] [Another embodiment] (1) In the above-described embodiment, the temperature-controlled storage device is the freezer / refrigerator showcase 1, but it may also be a freezer / refrigerator storage cabinet. Furthermore, the temperature-controlled storage device may also be a constant temperature case.

[0048] (2) In the above-described embodiment, the first solenoid valve 21A and the second solenoid valve 21B are provided in the refrigerant circulation path 20 as the solenoid valve 21, but the second solenoid valve 21B does not necessarily have to be provided.

[0049] (3) In the above embodiment, the temperature of the installation location, the humidity of the installation location, date and time information, model information, the set temperature, the cooling mode / defrosting mode, the cooling operation time per unit time, the temperature of the cooling media, and the defrosting time were used as constituent factors of the environmental state data (environmental state value group), the attribute state data (attribute state value group), and the feature state data (feature state value group). However, this is just an example, and not all of these factors need to be used, or additional factors may be used. Furthermore, factors may be interchanged between the various state data (state value groups). Furthermore, time-series data of each constituent factor may be processed (statistically processed) over time.

[0050] (4) In the above embodiment, the equipment state calculation model 5 is constructed as a type of multiple regression model, and an example is shown in which a failure is predicted based on the degree of difference d. However, the equipment state calculation model 5 may use other algorithms. For example, SPC (Statistical Process Control) techniques, anomaly detection, MT method, guard bands, etc. may be used.

[0051] (5) The functional roles of the terminal control unit 3 and the cloud server 4 are examples, and at least some of the functions of the terminal control unit 3 may be provided in the cloud server 4, or at least some of the functions of the cloud server 4 may be provided in the terminal control unit 3.

[0052] (6) The functional units in the functional block diagrams shown in FIGS. 3 and 4 may be integrated or further divided.

[0053] (7) In the above embodiment, the temperature of the cooling medium is measured by measuring the temperature of the cool air blown out. However, the temperature of the item storage chamber 13 may be measured.

[0054] The configurations disclosed in the above embodiments (including other embodiments, the same applies below) can be applied in combination with configurations disclosed in other embodiments, as long as no contradiction arises. Furthermore, the embodiments disclosed in this specification are examples, and the embodiments of the present invention are not limited to these, and can be modified as appropriate within the scope that does not deviate from the purpose of the present invention. [Industrial Applicability]

[0055] The present invention is also applicable to a temperature-controlled storage device having an article storage chamber whose temperature is controlled by a temperature control device. [Explanation of symbols]

[0056] 1: Showcase (temperature-controlled storage device, freezer showcase, refrigerated showcase) 2: Freezer (temperature control equipment) 3: Terminal control unit 4: Cloud server 5: Device state calculation model 5A:Learning Management Department 5B: Practice Management Department 9: Measuring instruments group 9A: Driving attribute detector 9B: Environmental condition detector 9C: Feature state detector 11: Outer case 12: Inner case 13: Item Storage Room 14: Shelf 15:Ventilation passage 15A: Cool air outlet 15B: Cool air intake 16: Evaporator 17: Blower 20: Refrigerant circulation path 21: Solenoid valve 21A: First solenoid valve 21B: Second solenoid valve 31: Driving attribute acquisition unit 32: Environmental status acquisition unit 33: Feature state acquisition unit 34: Emergency and abnormality notification section 41: Attribute status management unit 42: Environmental Condition Management Department 43: Feature state management unit 51: Dissimilarity calculation unit 52: Failure estimation section 91: Case temperature sensor SC: Service Center

Claims

1. A failure prediction system for a temperature-controlled storage device having an item storage chamber whose temperature is controlled by a temperature control device, an environmental state management unit that generates an environmental state value group based on the environmental state data of the temperature-controlled accommodation device; an attribute state management unit that generates an attribute state value group based on attribute state data related to the operating attributes of the temperature control accommodation device; a characteristic state management unit that generates a group of characteristic state values ​​based on characteristic state data related to the temperature adjustment of the temperature adjustment accommodation device; a learned device state calculation model using the environmental state value group and the attribute state value group during normal operation of the temperature control storage device as learning input parameters and the feature state value group during normal operation of the temperature control storage device as learning output parameters; a difference calculation unit that calculates the difference between a calculated characteristic state value group output by inputting an actual environment state value group generated based on the environmental state data of the current temperature control storage device and an actual attribute state value group generated based on the attribute state data of the current temperature control storage device into the trained device state calculation model, and an actual characteristic state value group generated based on the characteristic state data of the current temperature control storage device; a failure estimation unit that calculates a failure probability of the temperature control storage device based on the degree of difference; A failure prediction system equipped with

2. The failure prediction system according to claim 1 , wherein the temperature-adjusted storage device is a cooling storage device, and the operating attributes include model information, a set temperature, and a cooling mode / defrosting mode.

3. The failure prediction system according to claim 1, further comprising an environmental condition acquisition unit that acquires the environmental condition of the temperature-controlled accommodation device from an external server and converts the environmental condition into environmental condition data.

4. The failure prediction system according to claim 3, wherein the temperature-controlled storage device is a cooling storage device, and the environmental conditions include the temperature of the installation location of the cooling storage device, the humidity of the installation location, and date and time information.

5. 2. The failure prediction system of claim 1, wherein the temperature-controlled storage device is a cooling storage device, and the characteristic state data includes the cooling operation time of the temperature-controlled equipment per unit time, the temperature of the cooling media supplied to the item storage chamber, and the defrosting time.

6. The failure prediction system according to claim 1 , wherein the environmental state data, the attribute state data, and the characteristic state data include time-series data.

7. The temperature-controlled storage device is a plurality of cooling storage devices installed in a store, A terminal control unit is provided in the store, the terminal control unit is connected to the plurality of cooling storage devices by data communication lines, and is also connected to a cloud server via a data communication line; The failure prediction system of claim 1, wherein the terminal control unit is provided with a driving attribute acquisition unit that acquires the attribute state data, an environmental state acquisition unit that acquires the environmental state data, and a feature state acquisition unit that acquires the feature state data, and the cloud server is configured with the environmental state management unit, the attribute state management unit, the feature state management unit, the device state calculation model, the dissimilarity calculation unit, and the failure estimation unit.

8. The failure prediction system of claim 7, wherein the cloud server is provided with a learning management unit that, if the failure estimation calculated by the failure estimation unit is incorrect, re-learns the device state calculation model using the group of environmental state values ​​related to the failure estimation error, the group of attribute state values ​​related to the failure estimation error, and the group of feature state values ​​related to the failure estimation error.

9. A failure prediction system as described in any one of claims 1 to 8, which is provided with an emergency abnormality notification unit that issues an emergency abnormality notification based on the characteristic state data or the original detection signal of the characteristic state data and a predetermined threshold range for determining abnormalities in the item storage room.

Images