Enterprise resource management method and platform based on cloud computing

Abstract

The invention discloses an enterprise resource management method and management platform based on cloud computing, and the method comprises the steps: obtaining real-time operation data of various resources of an enterprise, and uploading the real-time operation data to a cloud platform; preprocessing the resource data, wherein the preprocessing comprises data denoising, format unification and structure modeling; based on the change characteristics of the resource data in the monitoring time period, calculating the change entropy of the resource state to evaluate the fluctuation degree of the resource state; determining a coherence index of the resource behavior based on the direction vector aggregation degree of the resource state change; dynamically calculating the size of a scheduling window of the resource according to the change entropy of the resource and a coherence index; generating a state threshold value according to the resource state mean value and the standard deviation in the scheduling window; identifying and grading the resources according to the state threshold to form resource classification and scheduling suggestions; and a scheduling result is provided for enterprise users through a visual interface for decision support.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a cloud-based enterprise resource management method and platform. Background Technology

[0002] As enterprise IT evolves towards cloud-native + microservices + data platform, the massive amounts of heterogeneous, multi-scale, and asynchronous data generated by systems such as manufacturing, human resources, supply chain, finance, customer relations, and operations monitoring need to be uniformly managed and scheduled. Existing technologies typically suffer from the following common problems:

[0003] Data fragmentation and inconsistent definitions: ERP / MES / WMS / CRM / IT operations and maintenance systems each form data silos, with inconsistent sampling periods, timestamp accuracy and indicator definitions, making it difficult to align and calculate resource status across systems under the same time benchmark.

[0004] Static thresholds and fixed windows: Many schemes rely on fixed statistical windows (such as daily / weekly) and manually set one-sided or single thresholds, which are prone to false alarms / false negatives when faced with sudden load changes, seasonal or rhythmic fluctuations, and have poor adaptability to non-Gaussian distributions and heavy-tailed data.

[0005] Fragility of rules and inseparability from trends: Anomaly identification based on simple rules is difficult to distinguish between directional changing trends and random fluctuations without direction. In noisy environments, short-term fluctuations are easily amplified, making it impossible to stably support scheduling decisions.

[0006] Unquantified upstream and downstream dependencies: Resource scheduling often ignores process orchestration and call chain dependencies, and fails to incorporate critical paths, substitutes and dependency depth into comprehensive evaluation, resulting in single-point optimization triggering system-level chain effects.

[0007] Alarms without action: Existing solutions mostly focus on monitoring and alarms, lacking the ability to close the loop of identification results into classification and grading, priority ranking, redundancy release and action suggestions, making it difficult to truly guide resource allocation.

[0008] High security and integration costs: Cross-system data exchange lacks a unified interface and permission model, transmission encryption and fine-grained authorization are costly to implement, the visual interface has weak interactivity, and it is difficult to meet the differentiated views and auditing requirements of multiple roles. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a cloud computing-based enterprise resource management method, comprising the following steps:

[0010] Acquire real-time operational data of various enterprise resources and upload it to the cloud platform;

[0011] The resource data is preprocessed, including data denoising, format unification, and structural modeling.

[0012] Based on the changing characteristics of resource data within the monitoring period, the entropy of resource status change is calculated to assess its degree of fluctuation.

[0013] Based on the degree of aggregation of the direction vector of resource state changes, the consistency index of resource behavior is determined;

[0014] The scheduling window size of resources is dynamically calculated based on the resource change entropy and consistency index.

[0015] A status threshold is generated based on the mean and standard deviation of resource status within the scheduling window.

[0016] Resources are identified and classified based on status thresholds to generate resource classification and scheduling suggestions; the scheduling results are provided to enterprise users through a visual interface for decision support.

[0017] Furthermore, the resource data includes:

[0018] Structured and unstructured data, including human resources data, equipment operating status, material inventory information, financial transaction information, and customer interaction logs.

[0019] Furthermore, the change entropy is expressed by the following formula:

[0020]

[0021] Where, p i Entropy represents the frequency of a resource state in the i-th state category, where n is the number of state categories. A higher entropy value indicates greater volatility.

[0022] Furthermore, the consistency index is calculated based on the consistency of the resource state change vector, including:

[0023] The state changes of resources at continuous time points are represented in the form of direction vectors, and the overall directional consistency is calculated by the horizontal and vertical components of the weighted aggregate vector.

[0024] Furthermore, the size of the scheduling window is dynamically adjusted based on the following indicators: change entropy value, consistency index, resource usage priority, upstream and downstream dependency degree, and historical availability rate.

[0025] Furthermore, the state threshold T is:

[0026] T=μ+k·σ

[0027] Where μ is the mean of resource status within the scheduling window, σ is the standard deviation, and k is the sensitivity coefficient.

[0028] Furthermore, the resource scheduling recommendations include: resource allocation, task priority ranking, recommendations for releasing redundant resources, and warnings of potential risks.

[0029] A cloud-based enterprise resource management platform, employing cloud-based enterprise resource management methods, includes: a data acquisition module, a preprocessing module, a resource status assessment module, a scheduling analysis engine, a result display module, a data processing module, a user access control module, an encrypted communication module, and an API integration interface module; the data acquisition module, preprocessing module, resource status assessment module, scheduling analysis engine, result display module, user access control module, encrypted communication module, and API integration interface module are respectively connected to the data processing module;

[0030] The resource status assessment module is used to calculate change entropy and consistency indicators;

[0031] The scheduling analysis engine is used to generate scheduling windows and classification suggestions based on the evaluation results;

[0032] The results display module provides the results to the user in the form of interactive charts and has an alarm prompt function. If the status of a certain resource exceeds the threshold, a graphic, sound or SMS prompt will be triggered.

[0033] The access control module is used for hierarchical management of access permissions and operation scope for users with different roles within the enterprise;

[0034] The encrypted communication module and API integration interface module are used to ensure data transmission security and interface with external systems.

[0035] The beneficial effects of this invention are: by aligning and standardizing multi-source heterogeneous data over time, it maps the data into resource state vectors for unified expression, solving the comparability problem across systems and scopes, and providing feasible computational anchors for subsequent evaluation and scheduling.

[0036] A sliding scheduling window is introduced for the monitoring period, and the window length is adaptively adjusted according to fluctuations and trends to balance short-term agile response and medium-term stability, thereby reducing alarm jitter and frequent rescheduling caused by improper window values.

[0037] By characterizing the fluctuation of resources within a window through changing entropy, and combining this with smoothing and discretization strategies for sparse / zero counts, the metric remains stable and comparable even in low-sample, asynchronous sampling scenarios. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating a cloud-based enterprise resource management method. Detailed Implementation

[0039] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0040] The features and performance of the present invention will be further described in detail below with reference to embodiments.

[0041] like Figure 1 As shown, the cloud-based enterprise resource management method includes the following steps:

[0042] Acquire real-time operational data of various enterprise resources and upload it to the cloud platform;

[0043] The resource data is preprocessed, including data denoising, format unification, and structural modeling.

[0044] Based on the changing characteristics of resource data within the monitoring period, the entropy of resource status change is calculated to assess its degree of fluctuation.

[0045] Based on the degree of aggregation of the direction vector of resource state changes, the consistency index of resource behavior is determined;

[0046] The scheduling window size of resources is dynamically calculated based on the resource change entropy and consistency index.

[0047] A status threshold is generated based on the mean and standard deviation of resource status within the scheduling window.

[0048] Resources are identified and classified based on status thresholds to generate resource classification and scheduling suggestions; the scheduling results are provided to enterprise users through a visual interface for decision support.

[0049] Specifically, cloud-based enterprise resource management methods acquire real-time or near-real-time operational data of various enterprise resources and connect and store them on the cloud platform;

[0050] The resource data is preprocessed, including at least: data denoising, format unification, time alignment, and structural modeling;

[0051] The preprocessed data is mapped into a numerical state vector of the resource at each time point. Based on the change characteristics of the state vector during the monitoring period, the change entropy used to characterize the degree of fluctuation is calculated.

[0052] Based on the consistency of the directional changes of the state vector at consecutive time points, a coherence index of resource behavior is calculated.

[0053] The size of the scheduling window is dynamically determined based on the change entropy and consistency index.

[0054] Generate state thresholds based on state statistics within the scheduling window;

[0055] Resources are identified and classified based on state thresholds to generate resource classification and scheduling suggestions;

[0056] The scheduling results are provided to enterprise users through a visual interface to support decision-making, and alarm prompts are issued when preset trigger conditions are met.

[0057] The resource data includes at least one of the following: structured and unstructured data such as human resources data, equipment operating status, material inventory information, financial transaction information, customer interaction logs, system operation logs, or business texts.

[0058] The acquisition of the resource state vector includes: performing time alignment and standardization on multi-source heterogeneous data, and mapping the aligned multi-dimensional features into a fixed-dimensional numerical state vector according to preset rules; wherein, the monitoring time period is used to carry out statistical analysis, and the scheduling window is a sliding sub-window within the monitoring time period, which is updated according to a preset step size.

[0059] The calculation of the change entropy is based on the discretization and statistics of the values ​​of the state vector within the scheduling window. The discretization is obtained through equal division or clustering. Additive smoothing is used for the zero-count case. The logarithmic operation uses the natural logarithm.

[0060] The coherence index is a measure of the consistency of state increment directions at adjacent time points. It is obtained by weighted aggregation of unit direction increments, where data from more recent times have higher weights. When only a two-dimensional scene is involved, this consistency can be equivalent to the weighted aggregation result of horizontal and vertical components.

[0061] The qualitative relationship between the size of the scheduling window and each indicator includes: when the change entropy increases, the scheduling window shrinks; when the coherence indicator increases, the scheduling window increases; when the upstream and downstream dependence increases, the scheduling window shrinks; when the historical availability rate increases, the scheduling window increases; when the resource usage priority increases, the scheduling window increases; the above factors are combined according to preset weights to determine the final window size.

[0062] The state threshold includes an upper threshold and a lower threshold, which are preferably determined based on the statistical mean and dispersion within the scheduling window and in combination with a sensitivity coefficient, wherein the sensitivity coefficient is within a preset range; when the data distribution within the window deviates from normality, the threshold can be replaced by an upper / lower limit based on percentiles.

[0063] The identification and classification include at least the following: Normal: the state is between the upper and lower thresholds; Attention: the state crosses the boundary no more than once and the duration is less than the preset threshold; Abnormal: the state crosses the boundary more than once or the duration of a single cross-boundary reaches or exceeds the preset threshold; and based on this, resource allocation, priority ranking, release of redundant resources and risk management suggestions are generated.

[0064] The upstream and downstream dependency degree is quantified based on process orchestration or system call chain to determine the dependency relationship between resources, and the scheduling window of resources is modified by combining dependency depth, critical path location and substitutability.

[0065] The visualization interface provides interactive charts and supports multi-dimensional filtering, drill-down, and backtracking, and can present differentiated views based on the permission policies of different roles.

[0066] A cloud-based enterprise resource management platform for performing the method includes:

[0067] The data acquisition module is used for multi-source access and supports streaming and batch acquisition;

[0068] The preprocessing module is used to denoise, standardize the format, and align the time of the collected data.

[0069] The data processing module is used for feature construction, resource state vector calculation, change entropy and coherence assessment, and scheduling window calculation.

[0070] The scheduling analysis engine is used to generate state thresholds, identify and classify results, and form scheduling and handling suggestions based on the evaluation results.

[0071] The results display module is used to present results in interactive charts and provide query, filter and backtracking functions;

[0072] The user access control module is used for hierarchical management of access permissions and operation scope for users with different roles;

[0073] An encrypted communication module is used to encrypt data transmission and manage keys.

[0074] The API integration interface module is used for interfacing and exchanging data with external business systems.

[0075] The above modules are respectively connected to the data processing module.

[0076] When any resource status exceeds the upper / lower threshold or meets the preset continuous out-of-bounds conditions, the result display module triggers a graphical or audio prompt and can push an alarm via SMS, instant message or email, supporting alarm level stratification and escalation.

[0077] The functional boundaries between the preprocessing module and the data processing module are as follows: the former completes the unification of data quality and time series standards, while the latter completes the calculation of status and indicators, generation of windows and thresholds, and scheduling analysis; the two are decoupled through a standardized data interface.

[0078] The API integration interface module supports providing external resource status subscription, threshold and alarm callback, and retrieval or push of scheduling suggestions, and implements access control through the user permission control module.

[0079] Example 1: Coordinated scheduling of equipment, manpower, and materials in discrete manufacturing production lines

[0080] An assembly manufacturing company (including multiple production lines, parallel workstations, and shared tooling) needs to consistently achieve its delivery and yield rate targets under conditions of fluctuating orders, frequent product changes, and limited manpower in work teams.

[0081] The platform is deployed in a cloud-native manner: the data acquisition module connects to PLC / SCADA and equipment logs through an industrial gateway, and also interfaces with MES (work order / process), WMS (inventory / batch), HRM (scheduling / skills), and QMS (quality control system), storing data in partitions within a data lake and time-series database. The access control module implements hierarchical authorization based on roles, workshops, and work sections; the encrypted communication module uses two-way authentication for external links.

[0082] Data collection items: equipment start-up / shutdown, production cycle time, yield, energy consumption, critical alarms; material inventory and ETA upon arrival; team on-duty status and skill matrix; work order priority and delivery commitment.

[0083] Preprocessing: Timestamps are standardized to the second level, and cross-system data is time-aligned and field definitions are mapped; metrics with missing data of no more than 5 minutes are interpolated using nearest neighbor, while those missing data of more than 5 minutes are marked as unreliable and reduced in weight in subsequent calculations.

[0084] A status vector is constructed at the production line-workstation level, including dimensions such as cycle time, work-in-process inventory, yield, equipment health score, material availability, number of employees and skill coverage, and work order urgency; it is updated on a rolling basis according to the workshop time zone.

[0085] The monitoring time period defaults to spanning multiple shifts (e.g., 8 hours), and the dispatch window updates by sliding within this timeframe.

[0086] Discrete events within the window (such as downtime type / frequency of short stops) and continuous quantities (cycle time / yield / manpower usage) are uniformly discretized and statistically analyzed. Natural logarithmic calculations and additive smoothing are used to suppress the impact of zero counts on the measurement. High entropy indicates unstable workstation status or frequent work condition switching.

[0087] By unitizing and weighting the state increments at consecutive moments with time decay, we can observe whether there is a directional tendency for continuous acceleration / deceleration and improvement / deterioration. High consistency indicates the existence of an interpretable trend (such as a ramp-up or aging period), which is beneficial for extending the observation window; low consistency allows for a short window response to sudden changes.

[0088] Based on qualitative monotonic relationships: increased entropy → smaller window; increased coherence → larger window; deep upstream and downstream dependencies, tight critical paths → smaller window; high historical availability, high priority → larger window. Boundary constraints: lower window limit 10 minutes, upper window limit 2 hours, step size 5 minutes.

[0089] Upper / lower thresholds are generated based on statistical results within the window; for heavy-tailed data, percentile thresholds (such as upper limit P95, lower limit P05) are used instead.

[0090] Tier rules: Normal / Attention / Abnormal (supports combined conditions for duration and number of consecutive out-of-bounds errors).

[0091] Resource allocation: When critical workstations are abnormal and there is a deep upstream dependency, prioritize the allocation of work teams with matching skills and increase the frequency of quality inspections; among parallel workstations, switch to the one with the higher health score.

[0092] Task prioritization: Prioritize work orders that are close to their promised deadline and have stable yield rates; merge work orders that require model changeover with similar process batches to reduce changeover losses.

[0093] Redundancy release: Reduce redundant manpower and maintain a minimum safe staffing level for workstations with low entropy and high continuity within two consecutive windows.

[0094] Material pull: Based on the material shortage warning and ETA of the downstream workstation, trigger the WMS replenishment or alternative material approval process.

[0095] Provides differentiated views for workshop managers / process engineers / team leaders: workstation heatmap, critical path Gantt chart, work-in-process and yield trends, short-stop Pareto; provides graphical / audio alerts and push notifications via IM / email when anomalies are triggered.

[0096] By integrating with MES to sort work orders, with WMS to trigger replenishment, and with HRM to adjust shifts, a closed loop is formed from identification to classification to handling.

[0097] Short-term shutdowns at key production line stations decreased year-on-year, and on-time delivery rates for work orders improved; alarm confirmation time and average handling and recovery time were shortened, and redundant manpower periods were recovered.

[0098] Example 2: Operation and Maintenance Scheduling of Cloud Infrastructure and Microservice Call Chain

[0099] A cloud-native microservice system (container orchestration + service mesh) of an internet company experienced latency jitter during peak hours. Traditional static alarms and fixed sampling windows resulted in frequent false alarms and delayed scaling.

[0100] The data collection module interfaces with Prometheus / OTel to capture service-level QPS, P95 / P99 latency, error rate, container CPU / memory / IO, node resources, and HPA events; it also collects call chain topology (upstream-downstream) and SLO targets.

[0101] The permissions module assigns the minimum permissions based on operations and maintenance, backend, and product; the encrypted communication module is used for cross-cluster indicator channels.

[0102] Each indicator is aligned and outliers are removed at the second level. When the node clock deviation exceeds the threshold, it is corrected and marked as untrustworthy.

[0103] Construct service instance-level state vectors that cover latency, error rate, resource utilization, instance health, SLO deviation, auto-scaling / scaling cooldown time, and traffic ingress and egress strength.

[0104] Change entropy is used to describe the overall instability of latency and error rate within a window; coherence index is used to identify directional trends such as continuous increase in latency and continuous decrease in error rate, thus distinguishing them from instantaneous spikes.

[0105] The monitoring period is set 2 hours before and after the peak, and the scheduling window is adaptively adjusted between 1 and 15 minutes: it is reduced when the entropy is high or the critical path of the call chain is under pressure; it is appropriately expanded when the continuity is high and the process is in the ramp-up phase to avoid jitter scaling.

[0106] Dependency is quantified based on the call chain graph: in-degree / out-degree, critical path location, availability of alternative services, and the effective status of the circuit breaker policy jointly determine the window and subsequent scheduling.

[0107] Upper and lower thresholds are set for latency, error rate, and resource utilization respectively; a quantile threshold is used for latency to adapt to long-tail requests; and the duration of continuous out-of-bounds errors is constrained for error rate.

[0108] The three levels of output—Normal, Attention, and Abnormal—are presented with different densities for different roles (Operations and Maintenance receive all outputs, while Product only receives aggregated outputs related to SLOs).

[0109] Scaling up and down and rate limiting: When upstream services continuously exceed limits and have high continuity, prioritize scaling up hot instances of bottleneck services; if downstream services are restricted, implement instantaneous rate limiting or gray-scale traffic distribution at the inlet.

[0110] Routing and Circuit Breaking: When a downstream service is abnormal and an alternative service is available, the service mesh routing weights are automatically adjusted; when the out-of-bounds route continues to reach a set threshold, a circuit breaker is triggered.

[0111] Resource reclamation: When multiple consecutive windows are stable and highly consistent, redundant instances are gradually reclaimed to avoid resource waste.

[0112] Change debouncing: The same service will not be repeatedly expanded or shrunk during the cooldown period to avoid oscillation.

[0113] High-voltage links are marked in real time on the topology heatmap; when the SLO deviation and latency exceed the limits of a certain service are met simultaneously and continue to exceed the threshold, a high-priority alarm is triggered; one-click location of the first upstream service that exceeds the limit is supported.

[0114] Submit scaling suggestions or automate work orders to the orchestration system; interact with the APM / log system to trace back related changes; and integrate with the work order system to achieve audit closure.

[0115] False alarm rate decreased during peak periods, and automatic scaling-up / scaling response time was shortened; SLO defaults decreased; resource utilization improved with no significant jitter.

[0116] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A cloud-based enterprise resource management method, characterized in that, Includes the following steps: Acquire real-time operational data of various enterprise resources and upload it to the cloud platform; The resource data is preprocessed, including data denoising, format unification, and structural modeling. Based on the changing characteristics of resource data within the monitoring period, the entropy of resource status change is calculated to assess its degree of fluctuation. Based on the degree of aggregation of the direction vector of resource state changes, the consistency index of resource behavior is determined; The scheduling window size of resources is dynamically calculated based on the resource change entropy and consistency index. A status threshold is generated based on the mean and standard deviation of resource status within the scheduling window. Resources are identified and classified based on state thresholds to generate resource classification and scheduling suggestions; The scheduling results are provided to enterprise users through a visual interface to support their decision-making.

2. The enterprise resource management method based on cloud computing according to claim 1, characterized in that, The resource data includes: Structured and unstructured data, including human resources data, equipment operating status, material inventory information, financial transaction information, and customer interaction logs.

3. The enterprise resource management method based on cloud computing according to claim 1, characterized in that, The change entropy is expressed by the following formula: Where, p i Entropy represents the frequency of a resource state in the i-th state category, where n is the number of state categories. A higher entropy value indicates greater volatility.

4. The enterprise resource management method based on cloud computing according to claim 1, characterized in that, The coherence metric is calculated based on the consistency of the resource state change vector, including: The state changes of resources at continuous time points are represented in the form of direction vectors, and the overall directional consistency is calculated by the horizontal and vertical components of the weighted aggregate vector.

5. The enterprise resource management method based on cloud computing according to claim 1, characterized in that, The size of the scheduling window is dynamically adjusted based on the following indicators: change entropy value, consistency index, resource usage priority, upstream and downstream dependency degree, and historical availability rate.

6. The enterprise resource management method based on cloud computing according to claim 1, characterized in that, The state threshold T is: T=μ+k·σ Where μ is the mean of resource status within the scheduling window, σ is the standard deviation, and k is the sensitivity coefficient.

7. The enterprise resource management method based on cloud computing according to claim 1, characterized in that, The resource scheduling recommendations include: resource allocation, task priority ranking, recommendations for releasing redundant resources, and warnings of potential risks.

8. A cloud-based enterprise resource management platform, characterized in that, The cloud-based enterprise resource management method according to any one of claims 1-7 is characterized by comprising: a data acquisition module, a preprocessing module, a resource status assessment module, a scheduling analysis engine, a result display module, a data processing module, a user access control module, an encrypted communication module, and an API integration interface module; wherein the data acquisition module, preprocessing module, resource status assessment module, scheduling analysis engine, result display module, user access control module, encrypted communication module, and API integration interface module are respectively connected to the data processing module; The resource status assessment module is used to calculate change entropy and consistency indicators; The scheduling analysis engine is used to generate scheduling windows and classification suggestions based on the evaluation results; The results display module provides the results to the user in the form of interactive charts and has an alarm prompt function. If the status of a certain resource exceeds the threshold, a graphic, sound or SMS prompt will be triggered. The access control module is used for hierarchical management of access permissions and operation scope for users with different roles within the enterprise; The encrypted communication module and API integration interface module are used to ensure data transmission security and interface with external systems.

Images